Education & TrainingFree
Responding to Institutional Data Requests
Send the request and your programme records, and get every figure back with the definition and population it was computed on.
The request says retention rate and it looks like one number. Every guide to responding to institutional data requests treats it that way: find the figure, format it, send it, log the submission. What none of them does is ask what population the requester means, and that is the whole difficulty. One programme, one set of student records, and four defensible answers, which is why the same institution reports a metric three ways in a single year and cannot explain it afterwards.
The BS in Environmental Science enrolled 214 degree-seeking students in fall 2024: 128 first-time full-time, 41 first-time part-time, 45 transfers in. On the federal basis, which counts certificate- or degree-seeking, first-time, full-time undergraduate students, 103 of the 128 re-enrolled somewhere in the institution, which is 80.47 percent. Ask the same cohort a slightly different question, whether they stayed in this programme rather than at this institution, and 96 of the 128 did. That is 75.00 percent, from the identical records.
The provost's office counts all 214 and any re-enrolment, so 176 gives 82.24 percent. The accreditor excludes the 9 who graduated early and asks about the programme, so 158 of 205 gives 77.07 percent. Four numbers, one dataset, and a spread of 7.24 points between the highest and the lowest. The programme's accreditor threshold is 78 percent, so two of the four clear it and two do not. Nothing about the students changed between them.
The definition is not a technicality, it is the answer
Requesters differ on purpose rather than by accident, and the regulation is explicit about it. An accreditor's standards must set expectations for success with respect to student achievement in relation to the institution's mission, as established by the institution. That is a deliberate grant of latitude: the accreditor is meant to measure something the institution defines, so its retention figure is not supposed to match the federal one. A response that quietly sends the federal number has answered a different question.
Which means the fix is not a canonical number, it is a stated one. Every figure leaves with four fields attached: the definition in a sentence, the population and its size, the date the data was extracted, and the system it came from. The 82.24 percent and the 77.07 percent stop being a contradiction the moment both carry their denominators, and next year's comparison is against a definition rather than against a memory of one.
Then keep the register. Four definitions of retention, three of headcount and two of completion is nine rows, and the second request of the year is a lookup rather than a rebuild. It is also the one artifact that survives staff turnover, which is the usual reason a figure cannot be reproduced two years later. Where the figures feed a full review, the programme review pack holds the analysis and this register holds the numbers underneath it.
How it works
Read the request
Every metric it names, the definition it gives, and the terms it leaves undefined.
Fix the populations
The exact denominator each metric needs, with every inclusion and exclusion written out.
Compute and label
Each figure with its definition, population size, source system and extraction date attached.
Reconcile the history
This cycle's figures against what you sent before, with any difference explained by definition.
What you get
- Every figure stated with its definition, its population, its size and its extraction date
- The same metric computed under each requester's definition, side by side with both denominators
- A definition register that turns the next request into a lookup rather than a rebuild
- The caveats a figure needs, written as part of the response rather than added later
- Any term in the request that carries no definition, listed as a question back
- Prior submissions reconciled to this one, so a changed number has a stated reason
Common questions
Why not just pick one definition and use it everywhere?
Because the requesters are not asking the same question. The federal collection specifies first-time, full-time, degree-seeking students, and an accreditor is required to measure student achievement against the institution's own mission. Sending one number to both means one of them received an answer to something it did not ask, and that is worse than reporting two figures.
What if the request does not define the term at all?
Then it comes back as a question rather than a guess, and the run drafts it. An undefined term is the single most common cause of a figure that cannot be reproduced next cycle. Where you have to answer before you can ask, the response states the definition used and flags it as an assumption in the covering note.
How does this stop the same metric being reported three ways?
It does not stop it, and it should not. Four legitimate definitions of retention for one programme produced 75.00, 77.07, 80.47 and 82.24 percent. What the register stops is reporting three of them without labels, so a year later nobody can say which was which or why the accreditor's figure was lower.
Does it matter when the data was pulled?
Enough that it gets its own field. A census-date extraction and a same-question extraction three weeks later disagree, because late registrations, withdrawals and grade changes all keep moving. Two figures with the same definition and different extraction dates look like an error and are not, and the date is what settles it.
What about the caveats?
They travel with the figure rather than being appended to a cover letter nobody reads. A cohort of 128 is small enough that seven students move it 5.47 points, and that belongs next to the number. The accreditation self-study pack is where the narrative around these figures gets written.
Can this handle the request for course-level outcome data?
Yes, and it is where definitions get slipperiest, because a course outcome and a programme outcome are often the same words at different grain. The outcome assessment evidence pack carries the mapping, and item analysis is what tells you whether the underlying assessment measured what the figure claims.
Does it produce the submission file itself?
It produces the figures, the definitions and the response document. Uploading into a collection portal is a separate act, and it should be, because someone at the institution has to own what was submitted. The register is what makes that person able to answer a follow-up question three months later.
Responding to Institutional Data Requests
Fill in the form and your workspace opens with the work already underway.