Is IPEDS ACTS a Stress Test or Just Stress?

26 Jul, 2026

A new IPEDS Technical Review Panel report finds the fragilities it exposed were already there — and the first cycle may measure reporting capacity more than admissions behavior.

What Did IPEDS ACTS Actually Measure?

When regulators stress test a bank, they are not trying to break it. The scenario they impose is deliberately harsher than anything the institution expects to encounter, and the purpose is to find out which weaknesses were already sitting in the balance sheet before the scenario arrived. A bank that fails a stress test has not been damaged by the test. It has been read by it, and the failure was legible only because something finally demanded enough of the institution at once to make it so.

Higher education has just been through one of these, and it was called the Admissions and Consumer Transparency Supplement. For the past year, almost every conversation about federal reporting has treated ACTS itself as the problem: the compressed timelines, the evolving guidance, the mid-cycle software revisions, the quality control scramble. Two earlier posts in this series tracked those struggles as they unfolded, first in What Happens When IPEDS Can't Be Maintained?, which raised structural concerns about the rollout, and then in The Quality Control Phase of IPEDS ACTS, which documented the scramble and the impossibly short institutional timelines. Both described the stress test while it was still running. A newly released Technical Review Panel report closes the loop, and it reverses the causal arrow the field has been using all year.

On June 2 and 3, 2026, RTI International, the contractor that operates the IPEDS web-based data collection system, convened Technical Review Panel #72 by videoconference to discuss practices and strategies for improving IPEDS data quality. Thirty-nine attendees participated, drawn from institutions, state system offices, research organizations, higher education associations, state and federal government, and elsewhere. To ground the discussion, RTI commissioned the Association for Institutional Research to conduct a background study bringing current practitioner voices into the room. Among that study's six key findings is the sentence that changes the conversation: ACTS "highlighted pre-existing fragilities in the IPEDS data ecosystem and also raised concerns about the reliability and validity of the data collected." Not created. Not introduced. Highlighted pre-existing fragilities. The panel agreed, and the summary records that the finding resonated with many panelists, who accepted that ACTS did not create the quality risks but made them more visible.

The lesson of ACTS, then, is not that federal reporting became harder. The lesson is that data governance became visible.

data-center

Christopher Bowns, Virginia Tech data center, 31 January 2008. Photograph originally published on Flickr and distributed via Wikimedia Commons under the Creative Commons Attribution-ShareAlike 2.0 Generic (CC BY-SA 2.0) license.

What Was Already Broken

If ACTS merely exposed the weaknesses, where did they come from? The report is unusually direct, and the catalogue will be familiar to anyone who has spent a reporting season inside an institutional research office.

Start with the ERP. Enterprise Resource Planning systems were designed to run business operations such as finance and human resources, and they were never designed for federal statistical reporting, so the way data are stored in them does not map to IPEDS definitions. Every cycle, IR professionals extract operational data and maneuver it into a framework the system was never built to produce, and that translation layer is where quality risk lives. Beneath it sits a deeper architectural mismatch. IPEDS is premised on reconstructable snapshots, the institution as it existed at a single point in time, while ERPs are transactional by design and continuously overwrite reality with the current state of the business. Institutions asked to reconstruct prior reporting periods are asking a system built to forget the past to remember it.

Staff turnover compounds all of it. When an IPEDS keyholder leaves, the queries survive but the reasoning often does not, and the report describes new staff inheriting code and running it without re-examination because the thought processes and judgment calls behind previous submissions were never written down. Outsourcing adds another layer, since many campuses depend on third-party vendors for IT or for the ERP itself, and those vendors understand their software without necessarily understanding IPEDS definitions and do not report to the IR office. Underneath everything sits the finding with the widest reach, which is the absence of reliable data governance. Staff in one department rename a variable or modify a field without realizing the change reverberates through every downstream office that depends on it, institutions migrate ERP systems without involving IR professionals, and departments keep changes to themselves because no one ever told them whom to notify.

What the draft consensus in the field has tended to miss, and what the panel put on the record, is that a second family of risks originates not at institutions but inside IPEDS itself. Several panelists agreed with the background paper's finding that IPEDS is designed around four-year baccalaureate institutions, which means that a college enrolling mostly part-time or adult learners may be reporting data that describe only a small fraction of its students, and that institutions with rolling admissions or quarter-based calendars are structurally obliged to leave many of their students out. Definitions have shifted over time and some remain ambiguous enough to be applied differently across campuses, with the result that even help desk guidance cannot always be perfectly consistent. The Prior Year Revision System permits corrections only one year back, so a mistake propagated through several cycles cannot be fully repaired. Panelists also noted that cross-survey validations are opaque, leaving reporters unsure which figures must reconcile before an error message fires.

None of this arrived with ACTS. All of it predates ACTS by years and in some cases decades. ACTS simply applied enough pressure, across enough offices, that the cracks could no longer be attributed to any single survey component.

The People Were Not the Problem

One sentence in the report deserves particular emphasis before anything else is built on top of it. Both the background study and the panelists, the summary says, were clear that these risks "stem from institutional business structures and systems, not from a lack of effort or care on the part of staff at institutions who are responsible for reporting IPEDS data."

That sentence is worth sitting with, because throughout the ACTS rollout many practitioners felt blamed by their institutions, by the discourse, and sometimes by themselves for outcomes they could not have prevented. Missed edge cases, irreproducible prior-year numbers, and definitions applied inconsistently across offices read, in the moment, as personal failures. The panel says otherwise. The systems were the problem, and for a profession that spent the past year absorbing an extraordinary amount of stress, that official recognition relocates the conversation from individual performance to institutional capacity, which is exactly where it belongs.

A Diagnostic Instrument, Not a Descriptive One

What follows goes somewhat further than the report does, and the seam is worth marking clearly. The panel's third key finding has two clauses, and the second one says that ACTS raised concerns about the reliability and validity of the data collected. The body of the summary develops the first clause at length and leaves the second largely where it found it. The argument below is an interpretation of that unexplored clause, not a conclusion the panel reached.

Consider what the report establishes and then apply it to ACTS specifically. Institutions struggled to reconstruct numbers for earlier years because the methodology behind prior submissions was undocumented and most campuses have no data warehouse. The preparation window was too short. There was no time to convene a technical review panel beforehand, to settle definitions and reporting guidance thoroughly, or to test the collection tools properly, and the panel agreed that all of this contributed to data quality problems. Coordination and data sharing were required across offices to an unprecedented degree, which is another way of saying that the handoffs the background study identifies as the single greatest risk to data quality were multiplied at precisely the moment institutions were least prepared to manage them. Institutions supplied numbers, and those numbers were assembled under exactly the conditions the report says produce quality risk.

The conclusion is not that anyone did the work badly. It is that the resulting collection is fit for one purpose and unfit for another. As a diagnostic instrument, ACTS worked extraordinarily well, telling us a great deal about which institutions can reconstruct history, which can coordinate across offices under load, and which have governance that survives contact with a novel requirement. As a descriptive dataset, one that supports confident comparison between institutions, the first cycle should be treated with real caution. Researchers, journalists, rankings compilers, and state systems will be tempted to read year-one ACTS figures as measurements of admissions behavior. They are better understood as measurements of reporting capacity, and the variation between institutions may say more about data infrastructure than about anything happening in an admissions office.

The panel's own instincts ran partly in the other direction, and any honest account has to say so. Some panelists supported expanding the student-level collection methodology that ACTS introduced to other survey components, on the grounds that having NCES perform the aggregation would produce consistency across institutions and remove a whole class of calculation and data entry errors at the campus end. The design, in other words, attracted support even as the first year's output attracted concern, and there is no contradiction in holding both positions. A well-built instrument can still produce noisy readings the first time it is fielded under compressed conditions and unsettled definitions.

What the Panel Wants Next

The recommendations read like a data governance curriculum, and their appearance in an official document of the IPEDS ecosystem rather than a consultant's deck is the newest thing about them. Panelists urged institutions to keep data in a centralized shared location, to document every change to systems and variables and communicate it to everyone downstream, to make snapshots standard practice even if that means storing flat Excel files in a shared folder, to involve IR professionals when leadership decides to change ERP systems, and to write real standard operating procedures with step-by-step guides for pulling data and mapping variables to survey components. Some panelists reported that their institutions had already begun reevaluating governance practices during ACTS itself, updating documentation and looking into data warehousing. The stress test was already producing repairs before anyone had written the report on it.

Two forward-looking suggestions in particular have not received the attention they deserve. The first is that NCES might approach ERP vendors directly and open a conversation about making their products more conducive to IPEDS reporting, which would attack the translation layer at its source rather than asking every campus to solve the same problem independently. The second is that a few panelists favored expanding the ACTS Aggregator Tool to other survey components to improve consistency and reduce tabulation errors, which is a quietly significant signal that the infrastructure built under duress for a disliked requirement is now being eyed as reusable.

Buried in the practices discussion is an unexpected detail. Among the innovative approaches the background study identified, alongside ticketing systems that store scripts and notes, parallel verification in which two offices independently run the same data, peer consistency groups that compare similar institutions to detect anomalies, and saved population snapshots reused across components, is the practice of AI-narrated standard operating procedures, in which an IR professional narrates their reporting process in real time to an AI tool that turns it into documentation. Only a few years ago AI barely appeared in IPEDS conversations. Now it shows up without fanfare in a federal review panel's inventory of promising data quality practices, as a way to capture exactly the kind of undocumented institutional knowledge the report warns about, and the panel was appropriately sober about the barriers, noting that some institutions prohibit AI tools outright and that solo practitioners have no time to document anything.

Running through all of it is a recommendation that may prove the most consequential in the long run, which is that senior leadership needs to understand IPEDS. Panelists suggested that NCES communicate directly with presidents and chancellors, reframed in the language leaders actually respond to, such as risk and rankings, because leadership engagement determines whether IR professionals can get access to data, whether governance policies have teeth, and whether reporting is treated as an institutional priority or a compliance formality. State coordinators observed that some institutions have a strong data culture supported by leadership while others treat reporting as a box to check, and that the difference shows up directly in data quality. One panelist's suggestion was almost poignant in its specificity, which is that senior leaders may be more likely to read a physical letter than an email.

A Familiar Pattern

Readers of this site's AI governance writing will notice a parallel, and it is unlikely to be a coincidence. When organizations began deploying agentic AI, the technology exposed weaknesses that had always been there: undocumented processes, unclear data ownership, systems nobody fully understood, knowledge that lived in one person's head. Agentic AI did not create those weaknesses. It revealed them, because it was the first thing to depend on all of them at once. ACTS did precisely the same to institutional data governance, arriving as a demand for coordination, reproducibility, and documentation at a level institutions had never actually been tested on, and the institutions that had quietly deferred that investment discovered all at once how much of their reporting rested on memory, spreadsheets, one expert, undocumented SQL, and institutional folklore. Any sufficiently demanding new dependency is a stress test of the governance underneath it.

One part of this story has gone almost unremarked. In May, NCES announced that it had completed procurement for the next collection cycle and that beginning in 2026-27 institutions will be working with a new contractor, AEM Corporation, with RTI continuing to support the 2025-26 collection year into early 2027. Read that against the report and the symmetry is hard to miss. A panel convened to warn that institutional knowledge evaporates when the person who holds it leaves, that queries outlive the reasoning behind them, and that undocumented judgment calls are a quality risk, was convened in the same year the system itself changes hands. Everything the TRP asks of a one-person IR office — write the procedures down, centralize the record, do not let the reasoning walk out the door with the individual — applies with considerably more force to a national reporting infrastructure mid-handover. The transition is not a crisis, and NCES has said continuity is the priority, but it is a governance event, and it deserves to be treated as one rather than as a procurement footnote.

ACTS has absorbed more criticism than any IPEDS development in memory, much of it deserved, and some of it leveled in these pages. History may nonetheless remember it differently: not as the reporting requirement everyone disliked, but as the stress test that arrived early enough to expose the weaknesses before something more important broke. Federal reporting requirements come and go, and technical debt does not. What ACTS will be remembered for is less the data it collected than the questions it forced institutions to ask about their own reporting infrastructure. Who owns this data? Where is the documentation? Could we reproduce last year's numbers if the person who built them left tomorrow?

The TRP #72 summary is open for public comment through August 14, 2026. Comments may be sent to Kaitlin Lilienthal, IPEDS Technical Review Panel Task Leader at RTI International, at ipedsTRPcomment@rti.org. If the past year taught the field anything, it is that these processes go better when practitioners weigh in early.

My blog