Most campuses already run an early-alert system
An early alert is a flag. An instructor notices missed assignments, or a data system notices a slipping grade, and a notice goes to an advising office so a person can reach the student before the course is lost. A 2012 national survey by the John N. Gardner Institute found early-alert systems at 93 percent of four-year institutions, and EDUCAUSE, the higher-education technology association, found an academic early-alert system deployed at 73 percent of institutions in its 2019 data.
Each bar carries its own year and provenance: the 93 percent figure is a 2012 national survey of four-year institutions by the John N. Gardner Institute, as cited in 2025 coverage; the 73 percent figure is EDUCAUSE data from 2019; and the community-college figure is approximate, from a 2019 Ruffalo Noel Levitz vendor survey, as cited by New America in 2022.
Source: Gardner Institute national survey, 2012, as cited in 2025 coverage; EDUCAUSE, 2019 data; Ruffalo Noel Levitz, 2019, vendor survey, as cited by New America 2022
Volume grew with adoption: faculty and staff raised 3.8 million flags in EAB's two systems in 2018-19, then 4.6 million in 2019-20, then 7.0 million in 2020-21.
Flags raised by faculty and staff across EAB's two systems, by academic year. The volumes are the vendor's own platform counts.
Source: EAB Navigate and Starfish systems; vendor-produced data, 2022 white paper
Of seven million flags, 51 percent led to outreach and 28 percent drew a response
EAB sells Navigate and Starfish, the systems behind those flags. In 2022 it published "Missed Connections," a white paper analyzing its own platform data. On average, 51 percent of alerts resulted in any outreach. The other 49 percent went unworked, flags that never turned into outreach. Roughly three-fifths of students who received outreach responded, which comes to 28 percent of the original alert pool. The paper names the cause: "Many advising offices do not have the capacity to follow up on every alert."
Source: EAB, 'Missed Connections,' 2022. The vendor's analysis of its own systems. No independent measurement of this funnel exists.
The white paper discloses neither an institution count nor a student sample size, and its recommended fix is added staff capacity to work more alerts.
Source: EAB, 'Missed Connections,' 2022; the about 3.6 million and about 2.0 million counts are computed from the vendor's published percentages
The vendor's paper is the only measurement of the funnel in print, and it discloses neither an institution count nor a student sample size; its recommended fix is added staff capacity to work more alerts. Whether anyone else publishes a flag-to-contact rate was checked directly for this piece: every record ERIC, the federal education research database, returns for a title search on "early alert," read one by one, plus the iPASS evaluation reports, the six-year MAAPS evaluation, the What Works Clearinghouse advising guide, and New America's 2022 report on alert systems. None of them publishes one.
How many flags go up depends on how hard faculty are asked
Before Cal State Fresno redesigned its early-alert process, faculty response rates to its university-wide progress surveys were "often less than 25 percent," so only a small subset of students ever received an alert. When the redesign sent targeted surveys about specific students, 35 percent were flagged in the first semester, against 6 percent of comparable students under the old process. Underlying risk was the same in both groups; the flag count followed how directly faculty were asked.
Augusta University published what a campaign does to the same number. Faculty response to progress reports averaged 65 percent in fall 2024 and 67 percent in spring 2025. After a concerted push in fall 2025, the rate reached 88 percent at week four, 87 at week seven, and 80 at week ten, against a goal of holding 80. The figures are the university's own reporting to Complete College Georgia.
The Cal State Fresno baseline is drawn as a band because the evaluation states only that faculty response rates to university-wide progress surveys were often less than 25 percent before the 2017 redesign. The Augusta University figures are the university's own reporting to Complete College Georgia. The two institutions ran different survey processes in different years, so the panel reads as two episodes rather than one series.
Source: MDRC/CCRC 2019; Complete College Georgia / Augusta University 2025 (institutional self-report)
Instructors raising the flags mostly cannot see what happens afterward. In Cal State Northridge's pilot, faculty "receive no feedback about resulting interventions for at-risk students." In EDUCAUSE's 2019 data, 39 percent of faculty users rated the systems very or extremely useful; among students the figure was 55 percent.
On a Monday morning, an adviser opens the queue and starts choosing
On a Monday morning, an adviser at a campus running one of these systems opens a queue of alerts. EAB's paper calls the first phase of the work triage: sort by how many flags each student carries, by grade point average and academic standing, or by a model's risk score, and work down from the top, because the office cannot follow up on everything. Then come the calls. At Cal State Fresno, peer mentors phoning flagged students "were often unable to reach them by phone," and few students emailed or called back. At Northridge, almost half of alerted students could never be reached by an adviser after multiple attempts, and among 53 surveyed, six, or 11 percent, reported meeting one. A 2002 evaluation at Columbia College in California found the far end of the same funnel: of 370 students surveyed across 38 classes, 14 recalled ever receiving an early-alert letter.
11%
of alerted students at Cal State Northridge reported meeting an adviser, 6 of 53 surveyed
Marcal, 2019; peer-reviewed
about 50%
of alerted students could not be reached by an adviser after multiple attempts
Marcal, 2019; the same study
14 of 370
students surveyed across 38 classes recalled ever receiving an early-alert letter
Pfleging, Columbia College, 2002
The two Northridge tiles describe the same peer-reviewed pilot evaluation: the 11 percent is 6 of 53 surveyed alerted students, and about half of alerted students could never be reached after multiple attempts. The Columbia College tile is the far end of the funnel: the share of students who recalled receiving the alert letter at all.
Source: Marcal, Research in Higher Education Journal 37, 2019 (Cal State Northridge pilot; peer-reviewed); Pfleging, Columbia College program evaluation, 2002
The queue outnumbers the people working it. In Tyton Partners' annual advising surveys, a consultancy series, "caseloads for advisors are too high" has ranked as the top barrier to improving advising every year since 2020. Even at the 48 percent of institutions that require students to meet an adviser, advisers meet at most 85 percent of their caseloads.
The randomized trials found gains only where a campus worked every flag
A randomized controlled trial assigns students by lottery to a program or to business as usual, so any outcome difference is credited to the program. Alert-driven advising has two large ones, and both found no overall effect.
iPASS, evaluated by MDRC and the Community College Research Center, randomized 8,011 students at Cal State Fresno, Montgomery County Community College, and UNC Charlotte in 2017. The colleges saw more communication and a small increase in advising meetings, and the evaluation found "no discernible positive effects on students' academic performance." The final report also recorded a harm: credits earned fell slightly at one college, most likely because students could not register until they attended mandatory advising sessions.
MAAPS randomized 10,037 low-income and first-generation students at eleven public universities in fall 2016, pairing alert technology with proactive advising. Across the full sample the program had "no significant impact on either graduation or persistence" (persistence means returning the next year). First-generation students assigned to the program persisted at a rate 2 percentage points below their control group.
One site diverged. Georgia State ran the model at full fidelity, meaning it delivered the program as designed for the entire six years, with MAAPS advisers as students' primary advisers; the university's own materials describe an adviser reaching out the day after an alert fires. Program students there graduated 7 percentage points higher than control. Black students in the program graduated 15 points higher and persisted 11 points higher, and Georgia State does not use race as a factor in its alert models.
Effects are percentage-point differences between students assigned to the program and their control groups, from a randomized trial of 10,037 low-income and first-generation students who entered eleven public universities in fall 2016. The full-sample rows are drawn at zero because the evaluation found no significant impact on either graduation or persistence.
Source: Ithaka S+R six-year MAAPS evaluation, 2023
A 2025 study by Oklahoma State University's institutional-research team compared 3,014 students in courses where instructors raised alerts with 3,702 statistically matched students in similar courses without them; students in the alert courses were 30 percent less likely to withdraw, and 4 percent more students earned a grade of C or higher, as reported by Inside Higher Ed.
| Study | Design | Students (n) | Result |
|---|---|---|---|
| iPASS | RCT, 3 institutions, 2017 cohorts | 8,011 | No discernible academic effects; small credit reduction at one college |
| MAAPS | RCT, 11 universities, 2016 cohort | 10,037 | No significant impact on graduation or persistence |
| Oklahoma State instructor-alert study | Matched comparison, 2021-23 | 3,014 vs 3,702 | 30% lower course withdrawal; 4% more grades of C or higher (as reported by Inside Higher Ed) |
Scroll the table sideways to see every column.
The two randomized trials evaluate whole advising programs built around alert technology. The Oklahoma State study compares students in courses where instructors raised alerts with statistically matched students in similar courses without them, so the three designs answer related but different questions. The Oklahoma State percentages are as reported by Inside Higher Ed.
Source: Mayer et al., MDRC/CCRC, 2019; Miller, Cohen, Yang & Pellegrino, MDRC, 2020; Ithaka S+R, 2023; Han, Meints, Burns & Loper, Studies in Higher Education, 2025
The federal What Works Clearinghouse, which reviews education research for rigor, issued its advising guide in October 2021 with four recommendations; a standalone early-alert system appears in none of the four.
The unworked share shows up in retention
In EAB's own data, flagged students who meet with an adviser are retained more than 8 percentage points above flagged students who get no outreach, and 12 points above students who get outreach and never respond. The same paper reports that at example institutions students of color received 51 percent of alerts while making up 33 percent of enrollment, so the students flagged most often are the most exposed to the unworked share. Every figure in this paragraph is the vendor's.
EAB, 2022; vendor-produced.
EAB publishes the gaps rather than the underlying retention rates, so the bars are drawn as percentage-point deficits against flagged students who met with an adviser. The no-outreach gap is stated by the vendor only as more than 8 points.
Source: EAB, 'Missed Connections,' 2022; the vendor's analysis of its own systems
As cited by New America, 13 percent of community colleges called their alert system a very effective retention strategy for the full student body. New America's 2022 report found "very little empirical evidence to validate the use of these programs" in national impact studies.
13%
of community colleges call their early-alert system a very effective retention and completion strategy for the whole student body
2019 data, as cited by New America 2022
about 40%
of four-year practitioners report improved retention and graduation rates
2020 Gardner Institute survey, as cited by New America 2022
Both figures are carried as cited by New America in 2022: the community-college share from 2019 survey data, and the four-year practitioner share, which is approximate, from a 2020 Gardner Institute survey.
Source: New America, 2022, citing 2019 survey data and a 2020 Gardner Institute survey
An institution can compute its own funnel from its system logs: flags raised, contacts made, responses drawn, and the time between steps. The published benchmark for comparison is one white paper, from one vendor, covering 7.0 million flags.
Methodology
Every figure in this piece comes from a dedicated fact base compiled for this research, whose sources were fetched and confirmed on August 21, 2026, either directly or through dated Internet Archive snapshots where publishers block automated access. Every number traces to a source in the Sources section. Vendor-produced and self-reported figures are labeled as such in the body and figure captions: the flag funnel and retention gradient come from EAB's analysis of its own systems (no institution count disclosed); the Augusta University response rates are the institution's own reporting to Complete College Georgia; the Georgia State next-day-outreach description comes from the university's marketing page, while the Georgia State outcome effects come from the independent Ithaka S+R evaluation. The absolute funnel counts in Figure 3 (≈3.6M, ≈2.0M) are computed from EAB's published percentages and labeled as computed.
The claim that no journal, federal evaluation, or association publishes a flag-to-contact rate rests on a documented search: an ERIC title search on "early alert," each record reviewed individually (search run August 21, 2026), both iPASS evaluation reports, the six-year MAAPS evaluation, the What Works Clearinghouse advising practice guide, and New America's 2022 report. Figures excluded as untraceable or misattributed: the circulating "45% of institutions deploy alerts on course decline (EDUCAUSE)" claim, which appears in no verified EDUCAUSE publication; the Purdue Course Signals 21.5% graduation claim, challenged on selection-bias grounds and never independently verified; and any standalone "early alerts reduce dropout by X%" estimate, which no rigorous evaluation supports. The 93% adoption figure is four-year institutions only, from 2012, and survives as cited because the primary survey report could not be retrieved. Baylor University's progress-report response rates, which surfaced in search snippets during research, were left out because the source page could not be retrieved; Augusta University covers the same point with verifiable reporting.