Most higher education data analytics misses the file where transfer credit decisions are actually worked out: a spreadsheet one person maintains, usually on the associate registrar’s desktop. The student information system holds what was decided. The spreadsheet holds why. So the reports can be accurate about every outcome and still explain none of them. Most campuses run part of their student record this way. Last August I sat across from an associate registrar working through a stack of transfer appeals two weeks before term. Fourteen years in that chair. She could tell me, without looking anything up, which sending institution’s second-year statistics course never maps clean, which program advisor approves everything that comes in from one particular college, and which three course codes have quietly meant different things since a curriculum change nobody flagged to her office. None of that was in the system. Some of it was in the spreadsheet. Most of it was in her. Why does higher education data analytics miss the registrar’s spreadsheet? Analytics projects on a campus start where the data is easy: enrolment counts, credit hours, persistence rates, term-over-term retention. All of it is real, and all of it describes outcomes after the fact. The decisions that produced those outcomes happened one at a time, in an office, with a syllabus and a course outline and somebody’s judgement about whether these two things are the same thing. The system records the result of that judgement. A credit was granted, or substituted, or denied. It does not record the condition the decision was made under: which version of the syllabus, what the advisor was told about the student’s program plan, whether this was the third exception granted for the same sending course this month, or that the last four students who received this substitution did not count toward the degree at the audit two years later. There is a difference between context and history. Context is what you can hand a system: the policy, the articulation agreement, the catalogue, the transfer guide. It is written down, it is shareable, and any institution can have it. History is what actually happened here, over years, in this office, with these sending institutions and these programs. It cannot be downloaded from anywhere. It exists only in the people who were there, until somebody writes it down. Right now most campuses have plenty of context and almost no history. The history is in a spreadsheet and in a person, and the person is going to retire. What is in a registrar’s transfer credit spreadsheet When you finally open one of these spreadsheets, the columns show you what the official systems have no field for. The same things keep showing up: Exceptions and the reason for them. The substitution that was granted, and one line of plain text explaining why, written by the person who granted it. Sending institutions that need a second look. Not a policy, just a list of the ones where the printed course title no longer matches what the course covers. People. Which advisor to call for which program, which department chair will actually answer in August, who has signing authority when the dean is away. Timing. Which appeals have to clear before the fee deadline, and which ones can wait until week three without hurting the student. Reversals. The decisions that came back. A credit granted in first year that did not count toward the major at graduation is the most expensive kind of error an office makes, and it is almost never in the analytics stack. That last column, when it exists at all, is usually the shortest one in the file. It is also the only one that measures whether the office is getting the decisions right. An afternoon audit of your exception decisions Here is something worth doing this week. It needs a spreadsheet and about three hours. Pull last term’s exceptions. Transfer credit substitutions, course equivalency waivers, prerequisite overrides, anything that required a human approval rather than a rule. Take twenty at random. Do not pick the interesting ones. For each one, answer four questions. Who made the call. What evidence they had in front of them. Where that reasoning is written down today. What happened to the student afterward. Count how many times the third answer is a person’s name. That number is how many decisions walk out the door when that colleague does. If eleven of twenty decisions can only be explained by asking one colleague, the problem is bigger than documentation. Those eleven decisions leave the institution on the day that colleague does. Check the fourth answer against the degree audit. Of those twenty students, how many had the credit hold all the way through to graduation requirements? Each one that did not is a rule the office should write down. Write the five clearest rules as plain sentences. Not a policy memo. One sentence each, in the language the office actually uses. Attach each rule to the thing it governs. The articulation record for that specific sending course, so it appears the next time that course code arrives. A rule filed in a shared drive only helps people who already know it is there. Then run the same twenty-decision sample next term and see whether the exception rate on those five moved. That is a measurable outcome from a data exercise that cost you one afternoon and no budget. Most offices do not like what they find in step three. That number is the most useful thing the audit gives you. What changes once transfer credit decisions are written down The registrar’s office stops depending on one person’s memory. A new advisor can answer a transfer question in September without escalating it. The appeals that come in during the two weeks before term get resolved on the same basis they were resolved on last year, which is what students think is happening already. And the analytics finally have something to work with. A retention model that knows a student’s transfer credits were granted by exception, and that similar exceptions failed at degree audit twice before, is reasoning about your institution. A model that sees only the credit count is describing higher education in general, which is not what you run. Somewhere on your campus there is a person who knows which transfer credits never reconcile, and that knowledge is currently a favour they do the institution for free. When they retire, the students who lose are the ones transferring in with twelve credits from a college three hours away, who will be told to retake a course they already passed. They will pay tuition for it. They will graduate a semester late, or not at all. Nobody will file that as a data problem, because nobody will be able to see what was lost.