Abstract
Across 24 school-by-firm comparisons covering all eight Ivy League universities, Crimson Education, IvyWise and Solomon Admissions Consulting reported client rates averaging 32.93% versus 5.43% for their general-applicant benchmarks. The ratio of the unrounded averages was 6.06. PrepScholar published a further Ivy-wide comparison, and College Essay Guy and The Ivy Institute reported student-level portfolio outcomes. None identified a verified cohort applying without private consultants. The study presents the six providers separately, tests how weighting affects the collective result, and checks the company figures against independent records wherever possible.
Reading guide. The main comparison concerns one application to one university. The six company profiles also cover student-level outcomes across a list. The independent checks follow the complete company-only analysis.
The question a family is buying an answer to
A parent who asks whether a consultant improves the chance of admission usually has a particular teenager in mind. The relevant comparison is the outcome for that student after hiring help against the outcome the same student would have had with the support already available at home and at school. A firm can observe one of those paths. It can also count the successes of other clients and compare its records with a university’s published figures. These are worthwhile observations. The research question is how far a family can travel from the second kind of evidence toward the first.
This study begins with what the sellers themselves put on record. Its six cases are College Essay Guy, Crimson Education, IvyWise, PrepScholar, Solomon Admissions Consulting, and The Ivy Institute. A reader can use the first results table to locate a particular provider, the collective tables to compare like observations, and the verification section to see which figures can be checked against an outside publication. The unit changes between some of those sections. Whenever possible, the wording identifies whether the number counts applications, offers, or students who received at least one offer.
The choice of unit changes the apparent size of success. Someone who applies to eight universities and receives two offers contributes eight applications and two successes to an application ledger. In a student ledger that asks whether the person received an Ivy offer, the same person contributes one case and one success. In a ledger about enrolling at a first choice, either result is possible. Someone who turns down both offers because of cost still had two acceptances. Families who sense that one of these measures describes their priorities better than the others are asking an important measurement question.
The primary result here concerns the application ledger. Across the eight Ivy League institutions, three providers disclosed enough information to form a complete grid of 24 client rates and 24 corresponding institutional rates. The difference between the two averages was 27.50 percentage points, calculated before rounding. Each university received one vote at each firm. The general-pool average in this calculation is a baseline assembled from the companies’ own figures. It is different from the acceptance rate of a separately identified group of students who navigated the process without private advising. [1, 2, 3]
This distinction has practical consequences even for a reader who cares only about the numbers. If paid guidance is especially common among highly prepared applicants, a portion of any raw difference may already be present when those applicants first meet an adviser. Conversely, a consultant who accepts unusually difficult cases or encourages students to attempt ambitious applications could produce a lower observed rate while helping students. The public records cannot determine the relative contributions of selection, the advice, and the choices the advice changes. Both upward and downward influences are possible.
A second result is about transparency. An applicant can learn more about the scale and boundaries of some numbers than others. In one instance it was possible to count reported offers, reproduce application denominators, and compare the result with eight separate universities’ figures. In another, checking the provider’s published page code was necessary to recover the value of an animated percentage. In yet another, carefully reading a footnote changed a multiple into an expressly qualified historical maximum. [3, 4, 5]
Study design and evidence rules
This is a comparative public-record study of a purposive sample. Each firm offers individualized undergraduate application support in the U.S. selective-college market and appeared in the original six-firm series. The Ivy Institute commissioned the work and requested its inclusion. Keeping the existing six was a substantive choice. Replacing a company that reports limited data with one that advertises a striking rate would alter the result before the calculation began. The sample does not establish which consultants are best, how often all private consultants publish results, or a national average across an industry.
For the primary analysis, an eligible observation had to appear in a company publication, a results or services page, or a report that the company made publicly available. A consultant’s own description of another company did not qualify as evidence for that competitor. Selected testimonials could illustrate the distinction between a case and a cohort, but a successful testimonial could not supply the denominator of an acceptance rate. No admission office was asked to identify individual clients. The review date is November 8, 2026.
The extraction distinguished five types of observation. A count of offers tells how many positive decisions the provider reports. A client rate additionally identifies the group against which successes were counted, sometimes only in broad terms. A paired rate puts a client rate next to a named benchmark. A portfolio result follows one person across a set of schools. Finally, a comparative headline, such as a claimed multiple, may be available even when its underlying pair of rates is missing. The study retains evidence in each class. The demands on the evidence increase when two measures are to be divided or averaged.
For a school-specific pair, the extraction retained the provider, school name, reported client percentage, reported general percentage, the provider’s own multiple if available, and any counts explicitly connected with the row. Institution names were standardized only to connect the same college across displays. Penn means the University of Pennsylvania. Columbia in the verification section refers to Columbia College and Columbia Engineering, the divisions covered by the university’s class profile. The spreadsheet-readable supplement preserves the displayed precision of the sources through their links and distinguishes Crimson’s cumulative offer counts from Solomon’s cycle-specific totals. [1, 2, 3, 6]
All eight Ivies served as the primary common list. This set is externally recognizable, has a fixed membership, and does not change when a ranking moves a school across a numerical cutoff. The decision was also practical. All eight appeared in the tables of Crimson, IvyWise, and Solomon. The same companies supplied a wider common set of 13 U.S. institutions, used for a sensitivity analysis. Different years or rounds remain a limitation even on the fixed list. Equal school weighting standardizes one part of the comparison while leaving those other features visible. [1, 2, 3]
The main rate ratio is calculated by averaging the 24 client percentages and dividing by the average of the corresponding 24 general percentages. A second statistic averages the 24 individual ratios themselves. Those formulas give slightly different answers because the first gives higher-baseline rows more influence on the final ratio than the second. The percentage-point gap is the difference between the average rates. All calculations use the published rate cells before rounding the resulting summaries. Where a company publishes counts, separate calculations use the counts directly. Inputs displayed to one or two decimal places limit the precision of any derived result.
Missing information remained missing. The calculations did not impute an applicant count from a rounded percentage, assign a zero effect to an undisclosing firm, turn a firm’s maximum claim into its mean, or derive a person-level rate from the sum of offers across colleges. These decisions matter to the result. Several firms advertise strong portfolio outcomes, and an average that silently treated those outcomes as per-application probabilities would produce a substantially different headline. There is no statistical basis for that substitution.
Six companies and six ways to read their results
| Provider | Client result or claim | Benchmark and usable comparison |
|---|---|---|
| College Essay Guy | 98% student outcome among applicants to its selective group of 50; 300+ annual top-college offers. | Portfolio and count. No comparable nonuser rate or school-by-school multiplier. |
| Crimson Education | 33.31% equal-school Ivy client mean. Company also reports approximately 7× across top colleges and 7.4× across Ivies plus Stanford in 2025–26. | 5.29% equal-school Ivy general mean. Recalculated fixed-list ratio 6.30×. |
| IvyWise | 34.41% equal-school Ivy client mean; above 98% to one or more best-fit colleges over five years. | 5.60% equal-school Ivy general mean. Fixed-list ratio 6.14×; company’s broad headline 6×. |
| PrepScholar | 31% at Ivies; 38% at top-ten institutions; 98% into one or more of a student’s top ten choices. | 4.8% and 4.9% group benchmarks. Printed rates imply 6.46× and 7.76×. Ivy headline 6.1×. |
| Solomon Admissions Consulting | 56 reported Ivy offers in 165 applications in 2024–25; 31.06% equal-school client mean. | 5.41% equal-school general mean; 5.74× ratio of means. Provider’s mean school factor 5.8×. |
| The Ivy Institute | 92% all-eight Ivy applicant portfolio; 98% top-20 portfolio for relevant 2025–26 applicants; up to 12× internal historical claim. | No published matched portfolio benchmark in sources examined. Historical maximum lacks the inputs for recalculation. |
Table sources are the firms’ own results and services pages and Solomon’s company-published exhibits. Institutional verification appears later. [4, 2, 7, 1, 8, 3, 9, 5]
The first table is an inventory. A firm that has an entry in the client-result column may still be absent from the three-firm calculation. For example, PrepScholar supplies both a client percentage and a benchmark for the Ivy League as a group. Giving that single group rate the same role as a complete set of eight school-specific rates would assume a weighting rule the company has not specified there. College Essay Guy and The Ivy Institute provide relevant evidence about obtaining at least one offer. Their student-level results belong in a different column from one institution’s decision on one application. [8, 4, 9]
College Essay Guy and the difference between an offer count and a student outcome
College Essay Guy publishes a figure that is easily missed in plain-text views of its one-on-one counseling page. The site uses animated counters. In the underlying public HTML, the counter for annual acceptances to top colleges has a target of 300 with a plus sign. The counter describing students who applied to the 50 most selective schools has a target of 98 with a percentage sign. The zeros visible before the animation runs are starting values. The college counseling FAQ also describes at least half a dozen admissions each year to each Ivy+ college. These are first-party claims with different units. [4]
The annual offer count speaks to the scale of activity. It can help distinguish a provider reporting a single notable result from one describing repeated placements. It does not tell how many assignments involved any particular school. A firm could obtain the same number of offers by serving more students, by supporting more applications per student, by achieving a higher rate at the same volume, or by combining all three. The figure also cannot be allocated to Ivy League schools simply because the text mentions Ivy acceptances alongside other destinations. Its wording includes a wider range of institutions. [4]
The 98% figure answers a closer question for families seeking a broadly selective outcome. The page ties it to students who applied to schools in its group of 50. A student receiving three offers and a student receiving one can each satisfy that measure. It would make a substantial difference whether most clients applied to one institution in that group or to a dozen. The reviewed claim did not supply each student’s list, the corresponding number of students without an offer, or an equivalent group of unaided applicants. A probability for one named college cannot be recovered from it. [4]
Consider two hypothetical seniors using the same writing service. One wants an engineering program at a single research university and would otherwise attend a local college. The other wants the option to choose among several selective universities and submits a much broader list. Both can count as students who applied to a top-50 school. Their definitions of a satisfactory outcome differ, even before any differences in preparation are considered. A persuasive portfolio rate for the second household might give the first household very little information about the specific decision it cares about.
The FAQ adds an important piece of context by defining success more broadly than where a student is admitted. The firm describes intentional choices and the experience of going through the process. A study of acceptance alone cannot evaluate whether a writing coach helped a student work more independently or present an experience in their own words. Equally, satisfaction with that process cannot be entered as an additional successful admission in a quantitative study. Both records can be kept without making either one carry the evidentiary burden of the other. [4]
The most useful additional disclosure for this provider would be a small portfolio table. It could identify a cycle, the list used to define the 50 institutions, the number of students who applied within the list, the distribution of applications per student, and the number receiving one or more offers. A separate tabulation could distinguish essay-only clients from comprehensive counseling clients. That would let a reader identify a cohort closer to the service they intend to buy. Until then, the published evidence supports an annual offer count and a reported selective-school portfolio percentage. It supplies no defensible success multiple against nonusers.
Crimson Education and the importance of the admission round
Crimson supplies school-level percentages in a U.S. results table and separately describes particular admissions cycles. In the table, the eight Ivy client rates range from 23.8% at Brown to 54.1% at Cornell. Across those eight institutions the equal-school mean is 33.31%, compared with 5.29% for its displayed general rates. Their ratio is 6.30. The company also publishes a wider top-college headline of about seven times. The narrower number in this study comes from a fixed school list and a specified calculation. Neither number requires assuming a client would have faced the general rate without Crimson. [2]
One reason to prefer the individual rows is that they reveal where the aggregate comes from. Cornell contributes a higher general rate than Harvard, as well as a higher client rate. Harvard nevertheless has a larger ratio because its benchmark is smaller. A household interested in how many additional positive decisions are observed per hundred applications will focus on the difference between rates. A reader interested in relative selectivity will often start with the ratio. For both readers it helps to have the original percentages beside any summary. [2]
The results page also distinguishes cumulative offer counts from admissions rates. A column of all-time offers is a count over a long window. Other parts of the page describe offers in the 2025–26 cycle and rates in the regular round. Combining a cumulative offer count with a rate from a different time window could produce a plausible-looking application denominator that never existed. That is why the data supplement keeps the count under its own label. No Crimson application volume was reconstructed from those cumulative counts. [2]
The company provides a second comparison with a different statistical design. In its public description of 2025–26 results, it reports a 7.4-fold figure across the Ivy League and Stanford. It describes weighting by Crimson’s application volume and using each university’s published overall rate for that cycle. It also reports 388 Ivy offers in the cycle. The school set, weighting rule and time period for that figure are explicitly different from the eight-school, equal-weight definition used in this study. Its 388 is a count of offers. A person admitted to multiple Ivies can contribute more than once. [7]
Why does weighting matter here? If the applications of a firm are concentrated in schools where its ratios are higher, the firm’s clients collectively experience a different mix of decisions than the one in an equal-school index. Volume weighting can therefore be appropriate for describing the firm’s workload. Equal weighting is useful when three firms with different client distributions are compared on a common menu of colleges. Neither weighting rule fixes differences between applicants within each college. The weighted and unweighted figures are both meaningful if the question and the period are identified.
The distinction between rounds is especially clear at Penn. Crimson highlights 95 admits for the 2025–26 round and a 40% regular-decision client rate. An all-rounds offer count cannot be divided by a regular-decision percentage to establish how many regular applicants it served. Some of those offers may have arrived earlier. A family applying only in the regular round would need a regular-round client numerator and denominator and an equivalent regular-round comparator. The broader table remains useful for understanding the firm’s general claim, while the highlighted example defines a more specific event. [2]
An older company publication issued for Singapore provides a historical cross-check on interpretation. It pairs Crimson and general rates for the 2021–22 round and flags some general rates as approximate. For Harvard, that earlier publication lists 17% for Crimson and 3% for the general population. The contemporary U.S. table displays 32.3% and 3.7%. These are different published observations. The earlier brochure’s audience, sources, and reporting window do not justify treating their difference as an improvement in counseling quality. The historical pair demonstrates why a family should retain the date as carefully as the statistic. [10, 2]
For an applicant weighing Crimson today, the most informative public questions concern which service the numbers cover, what the latest cycle shows for the intended round, and whether the provider can separate applicants with comparable starting records. A clear record of school, year and round would help quantify the first two. The third would require information about students’ circumstances before they enrolled. The school-by-school rates make Crimson suitable for the collective index. The material examined here did not identify a separately measured group of otherwise similar families who purchased no private guidance.
IvyWise and what a five-year average reveals
IvyWise labels its school-specific figures five-year averages. It reports both a general and an IvyWise percentage at each of 22 institutions and describes a broad result for getting into a student’s best-fit schools. In the eight-Ivy subset, the client mean is 34.41%, the general mean is 5.60%, and their ratio is 6.14. Across all 22 institutions, the corresponding means are 44.75% and 7.48%, producing a ratio of 5.99. That latter calculation is a close numerical match to the firm’s sixfold headline. It demonstrates consistency with one reasonable reading of the table without establishing the firm’s internal weighting method. [1]
A five-year view has a real advantage for a family investigating a school to which relatively few clients apply each year. One surprising decision can dominate a single small cohort. Combining several cycles can broaden the experience represented. Yet the term “five-year average” leaves open another consequential question. An average of five annual percentages, with each year weighted equally, can differ from five years of offers divided by five years of applications. The public display examined for this study supplies percentages rather than the annual counts needed to distinguish those constructions. [1]
For example, imagine that two students apply in one year and twenty in the next. If the firm receives one offer in the first year and five in the second, an average of the two rates is 37.5%. Pooling six offers over twenty-two applications instead gives 27.3%. Neither approach changes a decision. The analyst changed what received equal weight. The example is hypothetical. It explains why it is possible to reproduce IvyWise’s published table as a descriptive index while being unable to infer the actual probability experienced by one randomly chosen historical client application.
The university mix also matters. The firm publishes rates for schools outside the Ivy League, including Colgate and Amherst. Its 95.65% Colgate client rate and 15.04% general rate describe a different level of absolute selectivity from the 11.94% and 4.73% pair at MIT. Both observations make sense in a provider’s published table. If the question is access to engineering at MIT, however, the overall firm average at a broad list of schools carries less information than the relevant row. Identifying the intended school is the first meaningful adjustment a family can make. [1]
IvyWise’s other widely visible measure is the share of students getting into at least one best-fit school over five years, reported as above 98%. The outcome depends on how the best-fit list is constructed. Lists designed to contain a mix of possibilities and realistic alternatives can serve families well. Their success can reflect the work of identifying a college the student likes and can afford. That result should be judged on its own terms. An at-least-one best-fit outcome and an acceptance at a single highly selective college track different parts of the advising process. [1]
A subtle reason to preserve the greater-than sign is that the reported threshold is the result. Writing exactly 98% invents a precision that the company did not claim. Averaging the threshold with other firms’ exact-looking percentages adds another error. Some of those percentages refer to a top-ten personal list, others to a national ranking band, and another to applying to all eight Ivies. Although these claims use a similar vocabulary, the event need not have the same difficulty or meaning across firms.
The reporting horizon makes IvyWise especially useful as a contrast to one-cycle data. At a college where the current cohort changed substantially, a five-year general rate can legitimately differ from the rate released this spring. A single-year discrepancy is not evidence that a five-year figure is erroneous. To independently verify the precise five-year rate one would need the five cycles in question and the aggregation convention. To verify the client percentage one would also need the client application and offer records for those cycles. The website provides enough for the common-school calculation without resolving either of those additional questions.
PrepScholar and a headline that does not reproduce exactly
PrepScholar’s main admissions page reports that its clients are admitted at a rate of 31% at Ivy League colleges against a general benchmark of 4.8%. These figures give a 26.2-point absolute gap. Dividing the printed values gives 6.46. The site also describes an Ivy advantage of 6.1 times. Readers should be able to see both values. A company-reported multiple belongs in a record of the claim the company made. The result obtained by dividing the adjacent numbers belongs in a record of what a reader can reproduce. [8, 11]
The gap is large enough to examine the common explanation that the calculation used unrounded rates. If 31% is conventionally rounded to the nearest percentage point, the underlying value would normally fall between 30.5% and 31.5%. If 4.8% is rounded to the nearest tenth, its underlying value would lie between 4.75% and 4.85%. Even the smallest quotient compatible with those rounding intervals is approximately 6.29. The largest is about 6.63. Ordinary rounding alone, under these assumptions, does not recover 6.1. A separate period or weighting convention could account for it. The materials reviewed do not establish the explanation. [8]
The same page reports 38% at its group of top-ten U.S. colleges against 4.9%. Here the quotient is about 7.76 and the absolute gap is 33.1 points. As with the Ivy group figure, the arithmetic can be checked. Rebuilding the underlying benchmark would require a school list, period, and weights. The phrase “top ten” in this comparison refers to an institutional group. Elsewhere on the same page, getting into a student’s top ten refers to personal choices. A family can encounter both in a few seconds of reading and reasonably assume they mean the same thing. They describe different events. [8]
PrepScholar also reports a typical list of 16 applications and a 98% outcome for admission to at least one of the student’s ten preferred schools. That pair provides valuable context on the portfolio outcome. A family might ask whether an applicant’s ten preferences were recorded before decisions and whether every client in the published group applied to ten schools. It would also be helpful to know whether a family who enrolled for guidance on seven applications was counted on the basis of those seven or their entire eventual list. The questions are about the definition of exposure to advising. They do not invalidate the company’s reported outcome. [8, 12]
PrepScholar’s own comparisons additionally describe an association between starting counseling before senior year and a higher Ivy admission rate than starting in twelfth grade. The two groups in that claim both used the firm. The comparison can motivate a serious question about timing, but it cannot supply the missing group for a study of consulting versus no consulting. Students who start early may differ in family resources, planning habits, academic trajectory or the length of service purchased. Separating the effect of starting early would require measuring those starting differences or assigning timing independently. [11]
The company’s Ivy figure can be shown alongside the school-specific index as a sensitivity exercise. Giving equal weight to four firms’ Ivy summaries, with the first three defined as each firm’s ratio of equal-school means and PrepScholar contributing its printed-rate quotient, gives 6.16. Using its headline of 6.1 instead gives 6.07. The two answers are close, but neither is the main estimate. The first three contributing numbers use the same eight-school rule. PrepScholar’s aggregate school mix is undisclosed. Reporting this broader summary acknowledges the fourth firm’s evidence while displaying the cost of relaxing comparability. [1, 2, 3, 8]
Solomon Admissions Consulting and the value of a countable denominator
Solomon supplies the most readily reconstructible application ledger in the sample. It publishes an accountant’s report that lists, by university, client applications, client acceptances, the resulting percentage, and an overall-school benchmark. The 2024–25 Ivy records sum to 165 applications and 56 reported offers. That is a client application rate of 33.94%. This is the rate obtained when the student submitting an application to Cornell contributes exactly as much to the denominator as the student submitting one to Dartmouth. It is not the share of the firm’s students admitted to an Ivy. [3]
The equal-school average for the same eight Solomon rows is lower, at 31.06%. Cornell accounts for 32 applications and 17 offers, while Dartmouth accounts for eight applications and two offers. Equal-school weighting gives Dartmouth and Cornell the same influence. Counting each application gives Cornell four times the influence of Dartmouth. A family looking for the performance of the average application in the reported ledger would choose the second construction. A researcher comparing Solomon with firms that do not disclose those application counts can use only the first construction consistently across all three. [3]
Counts also permit a useful calculation of how much the school mix alone changes the benchmark. For each university, multiply Solomon’s application count by its published general rate. Add those eight products. The sum is 9.154 expected offers under the hypothetical assumption that each school’s general rate applies to the firm’s applications. Compare 56 observed offers with 9.154 benchmark-expected offers and the ratio is 6.12. Dividing the expected count by 165 gives a school-mix-adjusted benchmark of approximately 5.55%. Expected offers can be fractional because they are a mathematical reference, not observed decisions. [3]
The word “expected” has a narrow meaning here. This calculation assumes that a Harvard application would succeed at the overall Harvard rate and a Cornell application at the overall Cornell rate. It does not condition on school attended, academic record, intended program, application round, athletic recruitment, or any other applicant characteristic. Nor does it make the observed school list independent of advising. The 9.154 figure cannot be interpreted as the number of offers these particular students would have received on their own. Its purpose is to align the universities in the comparison, an adjustment that the published records make possible.
A prior Solomon report permits a second cohort to be reconstructed. For 2023–24, the eight rows yield 62 offers in 198 applications, or 31.31%. Multiplying that cohort’s application counts by its own displayed general rates gives 10.838 benchmark-expected offers, or a ratio of 5.72. Combining the two cohorts gives 118 offers in 363 applications, a 32.51% client application rate. The sum of the two expected counts is 19.992, producing a two-cycle observed-to-expected ratio of 5.90. This is an application-level, school-and-cycle-standardized comparison for one provider. It is distinct from the collective three-company index. [3, 13]
The historical comparison illustrates a common trap in interpreting improvement. Solomon recorded fewer Ivy offers in the later year, yet its application-weighted acceptance rate increased. Both statements are true because the later denominator was smaller. It would take further evidence to conclude that the firm accepted fewer clients, supported fewer applications per client, or advised different kinds of students. Similarly, a higher ratio can be driven by the mix of colleges or a change in the universities’ own selectivity. Two cohorts are useful context for a one-year claim, but they do not establish a trend in the effectiveness of the service. [3, 13]
The same exercise can be carried beyond the Ivies. Across the 34 schools in Solomon’s 2024–25 overall exhibit, there are 205 reported offers in 566 application records. The client rate is 36.22%. Applying the school benchmarks to those application counts gives 55.082 expected offers, for an observed-to-expected ratio of 3.72. A broader list produces a higher absolute client acceptance rate and a smaller comparative multiple. Institutions where a larger proportion of general applicants are admitted can bring both results about at once. These figures describe the exhibit, which is not necessarily the entire application history of every Solomon client. [3]
Students often care most about a dream school where the firm has comparatively few applications. At Dartmouth, changing just one decision among eight applications would move the reported client rate by 12.5 percentage points. Among 32 Cornell applications the same change would move the rate by 3.125 points. That is a material difference in the stability of the figures even if every recorded decision is correct. Small cells can be informative about a past cohort and still be poor stand-alone forecasts. Publishing the counts allows readers to see this directly. [3]
The Ivy Institute and the probability of receiving at least one offer
The Ivy Institute reports three distinct types of evidence relevant to this question. Its most precisely defined Ivy outcome is a 92% success rate among Ivy League Track students who applied to all eight Ivy universities, with success defined as admission to at least one. Its 2025–26 top-20 measure is 98% among students applying to nationally ranked top-20 universities who received at least one qualifying offer. Separately, its pricing page advertises up to a twelvefold acceptance advantage and attributes that comparison to internal historical information. The primary results table preserves all three statements and their different evidentiary status. [9, 14, 5]
The Ivy result has a stronger list definition than many portfolio statistics. The identity and number of Ivy universities are known, and a reader can tell that the denominator concerns participants who applied to all eight. It also narrows the population to which the figure applies. A student applying only to Penn, a family that would decline to attend several of the other Ivies, or a participant in a different track is asking a materially different question. The published fraction cannot identify the particular probability for any of them. [9, 14]
The top-20 group has a different boundary. Applicants can apply to different numbers of universities, and a ranking list can change from one year to the next. For meaningful comparison, a researcher would record the ranking authority, edition, treatment of ties, and which schools each student actually applied to. A public 98% portfolio rate by itself does not show how often applicants were admitted to their first preference within the group or whether the same individuals are also in the Ivy League Track denominator. The two published rates cannot be added, subtracted from each other, or averaged into a new outcome. [9, 14]
The site also displays a 100% top-30 headline. That number is retained as a broader provider claim. The more explicit denominator language for the top-20 and Ivy measures makes them better suited to the main comparison. A 100% observed proportion especially benefits from a cohort count. Five successes in five observations and one hundred successes in one hundred observations produce the same printed percentage but convey different amounts of historical experience. Providing the cohort size would improve its interpretation without requiring the school applications of individual students to be disclosed. [9]
The twelvefold figure presents a separate interpretive issue. “Up to” ordinarily describes the upper end of an experience or comparison, while a mean requires an averaging rule and a complete set of eligible observations. The public qualification identifies an internal historical comparison. The page reviewed did not identify the pair of rates, comparison group, dates, and aggregation rule necessary to calculate the number independently. Another firm-authored article expressly cautions against treating the 92% all-eight portfolio result and a university’s single-school acceptance rate as equivalent. This study follows the stronger definition and does not reverse-engineer a baseline by dividing 92 by twelve. [5, 14]
The firm’s senior-year Acceptance Commitment is relevant because it defines a family-centered outcome in contractual terms. Families designate eight colleges under the applicable terms, and a qualifying acceptance triggers the second payment. The designated eight can differ completely from the eight Ivy League institutions. An offer could satisfy the contract and fall outside the top-20 metric. Another offer could be celebrated by the student but fall outside the designated list. It follows that the performance of the admissions service, the result for an Ivy-only subgroup, and the incidence of a contingent payment require separate ledgers. [5]
What would let the Institute contribute a school-specific multiplier to the next edition? For each school and cycle, the firm could publish counts of eligible first-year client applications, offers, withdrawals and unresolved decisions, with a consistent rule for early admits and deferrals. For its distinctive all-eight question, an additional record could count comparable seniors who applied to all eight without private consulting and obtained at least one offer. That second disclosure would answer a question the school-specific analysis cannot. The available Institute evidence should therefore be read as strong in its definition of one student-level event and limited in its capacity to identify a measured unassisted baseline.
Collective findings across the same eight universities
| Provider | Client mean | General mean | Rate ratio |
|---|---|---|---|
| Crimson | 33.31% | 5.29% | 6.30× |
| IvyWise | 34.41% | 5.60% | 6.14× |
| Solomon | 31.06% | 5.41% | 5.74× |
| Three-firm average | 32.93% | 5.43% | 6.06× |
The central calculation gives 32.93% for the client side and 5.43% for the general side. The ratio of the unrounded averages is 6.06. Taking the mean of the 24 individual multiples instead gives 6.19. The median of those individual multiples is 6.00. All 24 client rate cells are above their paired school benchmarks. That consistency is an empirical characteristic of the published records. It does not reveal whether other companies, other colleges, or other unreported client cohorts would show the same pattern. Providers have a commercial reason to publish results their clients will find salient. [1, 2, 3]
The simplest way to understand the 6.06 statistic is to imagine a ledger with three equally sized blocks, one per provider. Every block has eight lines, one per Ivy. In one column are the firm’s eight admissions rates and in another are its eight general rates. Average the first column across the sheet and then do the same to the second. The quotient compares the two sides of this artificial but explicit ledger. It gives Brown the same weight as Harvard even if the actual applicants at any firm submitted far more applications to one than to the other.
A family may reasonably prefer a ledger with one line per person. That would require a different dataset. In the present records one student can contribute a Princeton application and a Brown application. If both succeed, adding two offers counts the decisions accurately while overstating the number of distinct successful people. If only one succeeds, a per-person at-least-one measure counts a full success but the two-application ledger records one success in two trials. Neither perspective is intrinsically superior. The study makes explicit which of the two is available for its collective calculation.
The distinction matters for judging the size of the absolute gap as well. A difference of 27.50 percentage points describes a difference between averages of reported acceptance rates. It cannot be scaled into a forecast of 275 additional acceptances caused by consulting for every thousand families who purchase it. That interpretation would require thousand-family cohorts with comparable initial qualifications, the same conditions of application, and observed outcomes with and without the intervention. No such cohort is contained in the company comparison. The group average is nevertheless worth knowing because it quantifies how different the firms’ reported records are from their chosen public reference points.
| University | Crimson | IvyWise | Solomon |
|---|---|---|---|
| Brown | 23.8 / 5.4 | 34.62 / 5.67 | 17.6 / 5.7 |
| Columbia | 31.5 / 4.2 | 40 / 4.27 | 52.4 / 4.9 |
| Cornell | 54.1 / 9.2 | 48.39 / 8.47 | 53.1 / 8.4 |
| Dartmouth | 34.3 / 5.8 | 32.2 / 6.52 | 25 / 6 |
| Harvard | 32.3 / 3.7 | 32.5 / 3.79 | 29.2 / 4.2 |
| Penn | 40.3 / 5.8 | 36.55 / 6.35 | 45.8 / 4.9 |
| Princeton | 24.6 / 4 | 25.49 / 4.94 | 11.8 / 4.4 |
| Yale | 25.6 / 4.2 | 25.56 / 4.82 | 13.6 / 4.8 |
| University | Client mean | General mean | Rate ratio |
|---|---|---|---|
| Brown | 25.34% | 5.59% | 4.53× |
| Columbia | 41.30% | 4.46% | 9.27× |
| Cornell | 51.86% | 8.69% | 5.97× |
| Dartmouth | 30.50% | 6.11% | 4.99× |
| Harvard | 31.33% | 3.90% | 8.04× |
| Penn | 40.88% | 5.68% | 7.19× |
| Princeton | 20.63% | 4.45% | 4.64× |
| Yale | 21.59% | 4.61% | 4.69× |
What the university rows add to the average
Brown offers a compact example of the gap between a relative and an absolute measure. Averaging the three providers gives a client rate of 25.34% and a general rate of 5.59%. The gap is 19.75 points. A reader using the ratio gets 4.53. A reader using the two percentages learns something the multiplier alone does not convey: the reported client outcome is still far below certainty. This is also a school at which the label “general rate” can refer to different stages of the same admissions cycle, a question examined in the verification section. [1, 2, 3]
At Columbia, the client mean is 41.30% and the general mean is 4.46%. Their ratio is 9.27, the highest of the eight school-level ratios of means. The three providers’ client figures are 31.5%, 40.0%, and 52.4%. That spread is substantive. Someone quoting only the average would hide the fact that the firms did not report one interchangeable Columbia experience. Cohorts may differ in size, period and preparation. [1, 2, 3]
Cornell has the highest client mean among the eight at 51.86%, but its ratio is smaller than Columbia’s because its general mean is also higher at 8.69%. The 43.17-point gap is the largest in the school-level table. An analyst looking for the largest difference in absolute chances and an analyst looking for the largest multiple would therefore name different universities. The choice depends on what the family means by advantage. A student deciding where to spend an application fee may find the actual percentages more informative than either ranking. [1, 2, 3]
At Dartmouth, the three provider rates average 30.50%, against a general mean of 6.11%. The ratio is about 4.99. A single number near five hides the sizes of the client cohorts. Solomon identifies eight Dartmouth applications. The other firms’ table cells do not supply equivalent application counts. The result is consequently a mean of three provider perspectives, with equal weight for each. No reader should infer from three firms being represented that three comparably large samples have been observed. [1, 2, 3]
Harvard has the lowest mean general rate of the eight at 3.90% after rounding. The client mean is 31.33%, which produces an 8.04 ratio. The three company client percentages sit within a narrow interval, from 29.2% to 32.5%. Agreement in these published percentages makes the descriptive finding particularly easy to see. It still does not resolve whether the students began as unusually competitive candidates for Harvard. Similar selection mechanisms operating at multiple firms can make their observed rates similar without identifying the effect of the service itself. [1, 2, 3]
Penn shows why portfolio and school-level marketing can speak past one another. Its three client rates average 40.88%, with a general mean of 5.68% and a ratio of 7.19. These numbers count a Penn decision. A family whose primary goal is admission to Penn’s particular undergraduate program would additionally want to know about the intended division, admissions round, and applicant category. A family interested in any one of eight universities should look for a portfolio measure instead. [1, 2, 3]
At Princeton the client mean is 20.63%, the lowest institution-level client mean in the grid, against 4.45% for the mean general rate. The ratio is 4.64. One feature of the row is the distance between Solomon’s two offers in seventeen reported applications and the higher percentages on the other provider pages. With that small Solomon denominator, one additional offer would materially change its percentage. The individual numbers provide a concrete reason to refrain from ranking firms by their Princeton cell alone. [1, 2, 3]
Yale has a three-firm client mean of 21.59% and a general mean of 4.61%, for a 4.69 ratio. Crimson and IvyWise report nearly identical client percentages, while Solomon’s figure is lower. The equal-weight blend places the smaller stated Solomon cohort on the same footing as the other rows because the other denominators are unavailable. Observing that the combined Yale result is lower than Harvard’s says something about these specific records. It does not establish that Yale is intrinsically harder for a similarly qualified applicant who buys guidance. [1, 2, 3]
How much does the collective answer depend on the calculation?
| Set and weighting | Client mean | General mean | Rate ratio |
|---|---|---|---|
| 8 Ivies, all 3 firms | 32.93% | 5.43% | 6.06× |
| 8 Ivies, omit Crimson | 32.74% | 5.51% | 5.94× |
| 8 Ivies, omit IvyWise | 32.19% | 5.35% | 6.02× |
| 8 Ivies, omit Solomon | 33.86% | 5.45% | 6.22× |
| 13 shared schools, 3 firms | 33.71% | 5.23% | 6.44× |
Leaving out one provider at a time tests whether a single collection of numbers dominates the primary result. The ratio of means is 6.02 without IvyWise, 5.94 without Crimson, and 6.22 without Solomon. In all three cases the common list remains the same eight universities. The general result is therefore fairly stable to which two of the three disclosed matrices are retained. This exercise does not address missing firms. A provider that reports only portfolio outcomes could have very different school-level results, and the leave-one-out calculation offers no principled way to fill those cells. [2, 1, 3]
Leaving out one university at a time gives a wider range, approximately 5.69 to 6.28. Removing Columbia produces the lowest value, while removing Brown produces the highest. This is compatible with their positions in the school table. Removing an institution changes both the client mean and the general mean, so the size of the shift depends on the levels of both rates. Describing these as sensitivity results is intentional. They are not confidence intervals and carry no claim about the probability that an unobserved population parameter lies inside them.
The wider overlap adds Duke, Johns Hopkins, MIT, Stanford, and the University of Chicago to the eight Ivies. All thirteen U.S. institutions have company-reported pairs at all three firms. Across these 39 pairs, the client mean is 33.71% and the general mean is 5.23%. Their ratio rises to 6.44. This is a useful test of what happens when the definition of selective college changes in a specific, reproducible way. Extending the list changes the headline without adding a single new client record to the source publications. [2, 1, 3]
An even broader comparison uses each company’s own U.S. table, allowing the school lists to differ. At Crimson the seventeen U.S. rows produce a client mean of 36.65% against 5.71%, a 6.42 ratio. At IvyWise the twenty-two rows produce 44.75% against 7.48%, a 5.99 ratio. Solomon’s thirty-four overall exhibit rows produce 39.92% against 11.29%, a 3.53 ratio. Crimson’s combined Oxford and Cambridge row is outside the U.S. scope. These are valid summaries of three different menus. They cannot support a fair ranking of the firms, because changing the schools changes the calculation. [2, 1, 3]
The broadest example also exposes a mathematical ceiling. A school that admits four in ten general applicants can never have a client-to-general ratio greater than 2.5, even if it accepts every client. A school that admits one in twenty has much more room for a large ratio. A firm that helps clients enter a broad mix of excellent institutions may therefore have a lower average multiple by construction. A parent who prioritizes fit, scholarship eligibility, a specialized program, or living near home can reasonably prefer a less selective destination and derive no additional value from a large marketing multiplier.
The median offers another way to describe the distribution. The middle of the 24 individual ratios lies near six, while the highest and lowest single cells occur in much smaller slices of the evidence. A median is less affected by the most extreme cell, but it gives the same importance to a percentage based on eight applications as to a percentage based on an unspecified, possibly much larger group. A median also throws away the absolute level of the underlying rates. For these reasons the medians and arithmetic mean of the ratios appear as companion statistics. The primary result keeps both rates visible.
It may be tempting to calculate a standard error across the 24 cells and call the result a confidence interval for the industry. Such an interval would not have a credible sampling interpretation. The firms were intentionally selected, the schools are a fixed group, reporting windows overlap, and students may appear in multiple school cells. The national general rates for a school also involve overlapping applicant populations across providers. Reporting the sensitivity to a changed school set is more honest and more informative than a narrow interval computed as though the cells were independent.
What can be learned about applying without a private consultant?
A published general acceptance rate is calculated for all applicants included by the institution. It can include families who hired the named firm, families who hired another firm, applicants advised only by their school, participants in nonprofit access programs, and applicants who prepared largely on their own. A family that declines to purchase private help is not necessarily an applicant with no help. A comparison that aims to isolate the effect of the purchase must define what the alternatives include. None of the six firms’ records examined here provided both a verified no-private-consultant cohort and equivalent client outcomes for the same target population. [7, 1, 3, 8, 4, 14]
The missing information can be described precisely. Suppose one university knew the share of applicants who had hired any private consultant. Its overall acceptance rate would be a weighted average of the admission rates for private-consulting users and nonusers. To recover the nonuser rate from that relationship, a researcher would need the consulting share and a representative rate for all of the users in the same applicant pool. A high-profile firm’s published client rate is not generally representative of everyone using private advice. Substituting it would silently replace one unobserved group with another.
Even a university that could identify nonusers might face a second problem. At the point when a family hires help, students can already differ in grades, course opportunities, resources, geographic location, confidence in their school counseling, and preferred application strategy. Some might have been admitted whether or not they purchased additional help. Others might never have considered the institution without the adviser. Comparing the two observed groups would answer a more specific question than using the whole applicant pool, but it would still combine initial differences with any gains or losses attributable to the intervention.
It is possible for two firms with the same quality of advice to publish very different results. Suppose one is often hired years in advance by families already determined to apply early at a small set of universities. The other handles late-season applications after a disappointing early result. Both could improve a student’s materials in the same ways. The second group starts with less time and a different pattern of past decisions. A comparison of raw offers cannot distinguish those differences from the effect of the counselor. Documenting when students entered the engagement would address one piece of that problem with relatively little burden on a provider.
There is also a choice about whether to hold the college list fixed. A fixed-list study asks what difference consulting makes for chances at institutions the family initially intended to pursue. That is useful when a student already knows the goal. An open-list study asks whether the student ultimately obtained a desirable, affordable offer anywhere, including newly discovered colleges. That question permits an important form of consulting to work. It also requires defining a desirable outcome before the offers arrive so that any eventual acceptance cannot be relabeled a top choice after the fact.
For a student who gets into a school discovered through an adviser, the role of the counseling may be obvious in one sense: without hearing about the college, the student might never have applied. It is more difficult to infer whether the counselor improved the probability of an offer conditional on applying. Conversely, a family might learn from an adviser that a college is a poor fit and decide against applying. That may be a successful counseling outcome while removing an observation from an acceptance-rate dataset altogether. The effect of advice on submitting an application needs its own record.
An especially misleading shortcut is to multiply one national acceptance rate by a firm’s national headline and present the result as a prediction for an individual. A published ratio depends on the school set, years and clients from which it came. It does not transfer automatically to a different university, a different degree program, or a student with different academic preparation. In some combinations the calculation could even exceed one hundred percent. The underlying issue is the absence of a model calibrated to that particular student and institution, not merely the need to cap the result at one hundred.
Why the all-eight comparison cannot be reconstructed from university rates
For the Ivy Institute’s all-eight portfolio, the missing baseline is the share of comparable applicants who applied to all eight without paid guidance and received one or more offers. Eight published admissions rates cannot reveal that quantity. Each university-wide rate concerns its own applicants, many of whom did not apply to all the other seven. Restricting the denominator to all-eight applicants changes the population before any arithmetic is performed. The Institute’s page defining its portfolio outcome offers a better account of this distinction than any attempt to divide its 92% by a general one-school figure. [9, 14]
Even if eight separate rates for the correct all-eight subgroup were known, the overlap of admissions across students would still matter. In an illustrative group of ten applicants to two schools, suppose each school admits three. If all three offers at the second school go to people already admitted at the first, only three people receive any offer. If the two sets are disjoint, six people receive an offer. Both patterns produce exactly the same per-school percentages. That is why the joint outcome requires student-level information about which decisions belong to the same student.
The popular shortcut of subtracting the product of eight rejection rates from one assumes independent decisions with appropriately defined probabilities. The same transcript, achievements and personal circumstances accompany several applications, and the student’s choice of schools already relates them. Without independence or evidence about the joint distribution, the calculation is a scenario rather than an observed comparison. Introducing a hypothetical all-eight rate into the company evidence would give the Institute an apparent multiplier with no measured baseline. Leaving that cell unestimated preserves the value of its genuine portfolio statistic.
The same logic applies to personal preference lists. Two applicants can each report a top-ten list yet have only one college in common. The probability of at least one offer can differ even for equally prepared applicants because the portfolios differ. A better comparison would preserve both the number of schools and their broad levels of selectivity, alongside the student’s criteria for fit and affordability. Some questions call for a list defined by a researcher, while others call for the family’s list. The outcome label should make it possible to tell which was used.
Independent checks after the company analysis
The following checks were performed after completing the company-only calculations. University figures did not replace the company benchmarks in the primary index. An institution can verify how many first-year applicants it admitted. It generally does not identify which applicants hired a particular company. An accounting firm can trace contracts and reported decisions according to its engagement. That engagement may establish that a published table follows stated rules without supporting a causal claim. Finally, research with an actual comparison group can test advising in a different population. These are three levels of corroboration with different subjects.
| University | Offers / applications or source | Outside rate | Solomon benchmark |
|---|---|---|---|
| Brown | 2,418 / 42,765 | 5.65% | 5.7% |
| Columbia | 2,946 / 59,616 | 4.94% | 4.9% |
| Cornell | 6,065 / 72,281 | 8.39% | 8.4% |
| Dartmouth | 1,699 / 28,230 | 6.02% | 6.0% |
| Harvard | 2,003 / 47,893 | 4.18% | 4.2% |
| Penn | 3,570 / 72,544 | 4.92% | 4.9% |
| Princeton | 1,868 / 42,303 | 4.42% | 4.4% |
| Yale | Fall profile; 50,266 applicants | 4.8% as published | 4.8% |
| MIT | 1,334 / 29,281 | 4.56% | 4.6% |
Eight Ivy benchmarks in the Solomon report could be checked at the displayed precision against institutional publications or published Common Data Set figures, with a secondary transcription used for Penn’s full admitted count. Brown illustrates why the release date belongs beside the figure. Its March announcement reported 2,418 offers from 42,765 applicants, or 5.65%. Rounded to one decimal place, the underlying quotient gives 5.7%, matching the company table. A final fall count can differ after waitlist activity. This verifies a plausible benchmark for that reporting stage and nothing about which of the successful students used Solomon. [15, 3]
Columbia’s published class profile lists 59,616 applicants and 2,946 admitted students for its two undergraduate divisions, about 4.94%. Cornell’s Common Data Set lists 72,281 applicants and 6,065 admits, about 8.39%. Dartmouth gives 1,699 admits from 28,230 applicants, about 6.02%. These figures reproduce the order and magnitude of the three company benchmarks. They also show why precision should not be exaggerated: numbers presented to one decimal place in a provider table need not match the quotient of official counts to the last hundredth. [6, 16, 17, 3]
Harvard’s class reporting provides a particularly straightforward verification. The university reported 2,003 offers and 47,893 applications, yielding about 4.18%. The 1,675 students accepting their places are enrollees in that report. Using that figure as the numerator would answer a different question and understate the admit rate. Princeton’s Common Data Set provides 1,868 admits out of 42,303 applications, or 4.42%. Yale’s fall class profile reports 4.8% directly. Taken together, these examples show how an outside reader can distinguish an acceptance benchmark from an enrollment or yield measure. [18, 19, 20, 3]
At Penn the university’s counselor newsletter identifies 72,544 applicants, including its early and regular applicant counts. The public transcription of Penn’s 2025–26 Common Data Set gives 3,570 admissions, producing 4.92%. The university’s Common Data Set index lists the corresponding report; the underlying hosted PDF could not be inspected directly for this extraction. The distinction is recorded because a reader should know whether a figure was taken from the institution itself or from a transcription of its return. At MIT, an additional non-Ivy check, the institutional facts page lists 1,334 offers and 29,281 applicants, yielding 4.56%. [21, 22, 23, 24, 3]
The independent evidence is strongest for Solomon’s own numerator where BPM describes the procedures it actually performed. The 2024–25 report says the records came from the company and decisions were reported by clients. The accountants inspected client contracts, traced totals and recalculated institutional rates and comparative factors. Unconfirmed results were counted as denials, while withdrawn applications were excluded. Those choices affect the reported rates. A family can decide whether the treatment of missing and withdrawn decisions is appropriate for its purposes because the report documents the rules. [3]
The engagement also states its limits: it did not express an examination opinion or a review conclusion on the rates. An agreed-upon-procedures report identifies the steps performed and findings from those steps. It cannot independently certify that every student’s reported decision was correct unless a procedure actually checked that fact. Nor could recomputing a quotient establish that the service caused the numerator to be higher. The most accurate verdict for Solomon is that specified numbers and procedures have outside documentation, and the historical comparison can be recalculated. [3, 13]
Crimson states that EY independently verified its results and labels its 2025–26 school-weighted comparison accordingly. This study located the statement in Crimson’s own materials. Searches for a public EY-signed report specifying the cohort, reporting period, verification procedures and conclusion did not locate such an engagement document. EY’s published entrepreneur profile establishes a public connection to the founder but does not provide those verification procedures. Accordingly, the paper attributes the verification claim to Crimson. It does not treat press coverage or an award as independent confirmation of the company’s reported 7.4 figure. [7, 25]
For IvyWise, the published rate pairs can be recomputed, and the five-year label makes the intended reporting horizon visible. Recomputing them independently is an arithmetic check, not access to the underlying five years of client files. For PrepScholar, the published inputs enable the numerical discrepancy to be documented, while a verified untreated applicant rate remains unavailable. For College Essay Guy, inspecting the published counter values verifies what the company displays, not which students made up the 98%. For The Ivy Institute, its explanatory pages help resolve the eligibility and outcome definitions, but the historical records underlying the upper-bound multiple were not available for independent reconstruction. [1, 8, 4, 14, 5]
Independent research with a comparison group
Avery’s randomized pilot addresses a related question more directly. The study followed 107 high-achieving seniors from relatively low-income families, of whom 52 were randomly offered individual counseling. The estimated increase in the probability of being admitted to at least one college in the most competitive category was 1.6 percentage points and was statistically uncertain. The available uncertainty was wide enough to admit both positive and negative effects. The study helps a reader understand what a genuinely comparative student-level design looks like. Its result is not a verification or rejection of any six-firm commercial estimate. [26]
Phillips and Reber studied two virtual-advising interventions in a randomized field experiment. Students applied more broadly to four-year colleges and reported feeling better supported. The authors did not find an overall increase in acceptance or enrollment. Here a measured improvement in the process coexisted with little evidence of movement in a final outcome. That matters to interpretation of the company evidence: an adviser can change application behavior even where an admissions advantage has not been demonstrated. The populations, intensity and services in the experiment differ from the private programs profiled here. [27]
A broader reading of selective-admissions research gives another reason for caution. Chetty, Deming and Friedman found differences in admissions linked to factors beyond standardized test scores, including recruitment, legacy status and nonacademic credentials in the highly selective private colleges they studied. A comparison that matches students only on test scores would therefore leave important sources of initial advantage unmeasured. That finding concerns the admissions environment. It does not quantify how much of the company-to-general gap comes from any of those factors, and it cannot be used to subtract a guessed selection adjustment from the observed sixfold ratio. [28]
A university confirmation, an accountant’s engagement and a randomized trial can all be correct at the same time. The university may genuinely admit about four percent of its applicants. A consultant may accurately record a much higher rate among its clients. A trial may still find a much smaller or uncertain change when similar students are offered a particular service. There is no contradiction to resolve, because the three sources measure different comparisons. Independent research is most valuable here when it disciplines the interpretation of claims while preserving the specific descriptive results that can be verified.
Reporting an outcome that a family can actually compare
A useful annual disclosure starts when the student enrolls. Record the type of engagement, its start date, the student’s initial preferences and the ordinary counseling available through school or another program. Preserve that record before decisions are known. After the season, distinguish a student who never filed an application from a student who applied and withdrew and from one who received a denial. Publish the number in each category. A small table containing these distinctions can be more informative than an elaborate dashboard whose denominator changes silently as the season progresses.
The college list also deserves a version history. A family’s first list captures what the student would initially have attempted. A later list may reflect better information about fit, cost or competitiveness. When a student adds an unexpected college and receives an offer, a study should be able to recognize that as a change in search behavior as well as a successful application. When a student drops a college that would have been a poor financial fit, the advising may also have served its purpose. Recording the list before and after advice lets the data represent both possibilities.
At the end of the cycle, publish at least two tables. One should present applications and decisions by institution, with early, regular and waitlist rules stated. The other should count people against a clearly defined portfolio outcome, such as at least one offer from a list chosen before applying. If the intended question concerns a top-ten personal list, record whether the student actually applied to ten schools and how the cutoffs were handled. If it concerns a ranked top-twenty group, identify the edition and treatment of ties. Those small details determine whether a later reader can reproduce the result.
A genuine no-private-consultant comparison requires additional evidence. One approach would recruit eligible families before purchase, record their initial goals and preparation, and follow both those who buy the service and those who do not. The study would need to determine whether nominal nonusers later hired another consultant and whether purchasers actually used the program they bought. It should publish attrition separately for both groups. Even with these precautions, an observational study would still need to assess how strongly unmeasured motivation or resources could influence its conclusions.
If a firm offers more funded places than it can immediately fill, a lottery among willing eligible families could identify the effect of an offer of advising with stronger internal validity. Follow everyone originally assigned, publish the rate of using the offered help, and define admitted outcomes before the draw. A lottery studies its own applicant pool, so its result would not automatically describe paying clients elsewhere. But it would give families something the present public records cannot: a transparent estimate of what changed when comparable students gained access to a specified service.
Conclusion
For a reader comparing published outcomes, the most useful collective result is an equal-school, equal-provider comparison covering all eight Ivy universities at three firms. Client rates averaged 32.93% and the paired general rates averaged 5.43%. Their ratio was 6.06. The sensitivity analyses show what happens when one firm is removed, when more institutions are included, and when actual application counts are available at one provider. The six company profiles preserve additional information about offer volume, success on a student’s personal list, and explicitly qualified headline claims. [2, 1, 3, 8, 4, 9]
For the household deciding whether to purchase help, the outstanding number is the admission probability of comparable students who went through the same cycle without private consulting. None of the company datasets reviewed identifies that probability. The defensible conclusion is therefore specific: several firms report client cohorts with substantially higher acceptance rates than their selected general-applicant benchmarks, and several publish high rates of obtaining at least one preferred offer. The multiplier that can be calculated from those observations describes the historical comparison. The incremental benefit of hiring one of these firms for a particular student remains unmeasured.
Disclosure and data availability
The Ivy Institute commissioned this series and is a firm examined in it. The references identify the material used, and the series supplement provides the underlying coding tables.
References
- IvyWise. Five-year overall and client admission rates by university.
- Crimson Education. U.S. student results and school-level comparison table.
- BPM LLP (May 14, 2026), prepared for Solomon Admissions Consulting. Independent accountants’ report on agreed-upon procedures for the 2024–25 admission cycle, exhibits A and B.
- College Essay Guy. College admissions counseling packages and FAQs.
- The Ivy Institute. Costs, package scope and Acceptance Commitment.
- Columbia Undergraduate Admissions. Columbia College and Columbia Engineering, Class of 2029 profile.
- Crimson Education. Published 2025–26 outcome definitions and verification claims.
- PrepScholar. College admissions counseling, client rates and benchmarks.
- The Ivy Institute. Ivy League Track outcomes and denominator definition.
- Crimson Education Singapore. Student success in 2021–22, historical brochure.
- PrepScholar. Firm’s 2026 comparison article, own results and early-start definition.
- PrepScholar. Complete admissions package, terms and prices.
- BPM LLP, prepared for Solomon Admissions Consulting. Agreed-upon procedures and exhibits, 2023–24 cycle.
- The Ivy Institute. What Value Do College Admissions Consultants Add? Definitions of student and application outcomes.
- Brown University. Class of 2029 March admission offers.
- Cornell University. Common Data Set 2025–26, C1, first-year applications and admissions.
- Dartmouth College. Class of 2029 profile.
- Harvard Gazette. Class of 2029 yield and admitted class, October 23, 2025.
- Princeton University. Common Data Set 2025–26, section C1.
- Yale University. Yale College Class Profile, fall 2025 edition observed.
- University of Pennsylvania. Penn Class of 2029 counselor newsletter.
- Penn Institutional Research and Analysis. Common Data Set archive.
- CollegeData.fyi. Transcription of University of Pennsylvania 2025–26 Common Data Set.
- Massachusetts Institute of Technology. MIT Facts, undergraduate admissions, entering fall 2025.
- EY New Zealand. Entrepreneur profile of Jamie Beaton, 2023.
- Avery, C. (2010). NBER Working Paper 16359. The Effects of College Counseling on High-Achieving, Low-Income Students, Table 5 and appendix regressions.
- Phillips, M., and Reber, S. (2022). American Economic Journal: Economic Policy, 14(3), 198–234. Does Virtual Advising Increase College Enrollment?.
- Chetty, R., Deming, D. J., and Friedman, J. N. (2026). Quarterly Journal of Economics, 141(1), 51–145. Diversifying Society’s Leaders? The Determinants and Causal Effects of Admission to Highly Selective Private Colleges.
