Why Can't Marketing Prove It Drove Enrolment?

By Midya U · Midya U Advisory

In brief: Most institutions can't produce the one list that would settle the argument: who enrolled, and which of them marketing actually touched. The ad-platform data and the registrar data don't share a key, and AI-assistant visits are stripping out the referrer that used to help bridge the two. The fix isn't a better tracking tool. It's reporting what you can defend and adding one question to your forms that survives a stripped header.

Early in a marketing audit I ask for one thing: the list of students who enrolled in the last full intake, with a column showing which of them marketing touched.

I have not been handed that list yet. What arrives instead is two files: a platform export with conversions, and a registrar extract with enrolments. There is no shared key between them, and the person sending them usually says so in the email, in some version of "these don't quite line up."

Is this a tracking problem or a setup problem?

Because in most institutions nobody set it up to be provable.

That is the plain answer and it is usually the right one. The tracking was installed once, by someone who has since left, and never reconciled against what the registrar records. No field connects a click to a student file, because nobody was ever asked to create one. Marketing gets held to the outcome anyway, which is the same trap I described in Why Faster Drafting Didn't Make You Faster.

The structural reasons underneath are real enough. A domestic applicant may first hear about you in the autumn of grade 11 and deposit 18 months later, while ad platform attribution windows top out around 90 days. Safari deletes the cookies your analytics writes in JavaScript after seven days of browsing without a return visit, so for a seventeen-year-old cohort the trail is usually gone before they apply.

Are AI answers making attribution worse?

When someone asks an assistant about your program and clicks through, the visit often carries no referrer, so your analytics files it under Direct. The analytics vendor Clickport published an April 2026 sample of about 372,000 sessions and found roughly a third of its AI-assistant sessions arriving with no referrer at all. The header goes missing four different ways, and the one that matters most on a phone is the in-app browser dropping it when a user taps out. A second vendor, Loamly, looked at eight times as many AI visits in February 2026 and put the share closer to 70%.

Those two estimates disagree by a factor of two and I would not put either on a board report. Both are vendors with a product to sell, and neither reports a single institutional site in its sample. What they agree on is the direction: a real share of the visits that arrive ready to act now land with no origin recorded. That is the reporting-side version of what I covered in Where Your Search Traffic Went. The step from there to "AI answers are eroding higher ed attribution" is my inference. I would test it before I built a plan on it.

Do marketing teams know they have a measurement problem?

Most of them do not rank it as one of their problems. When EAB surveyed 121 higher ed marketing leaders in the summer of 2025, four in five named limited budget as a barrier to hitting their enrolment goals, and two in three said they were carrying higher goals against flat or shrinking staff and budgets. Only 26% named measuring marketing effectiveness and ROI as a challenge at their institution.

So four in five are fighting for money, and one in four counts the proof problem among their difficulties. Those two numbers sit badly together, because a budget argument is won with evidence about what the last dollar produced.

The same EAB work reports 61% of enrolment marketing dollars now supporting digital. More of the budget keeps moving into the channels that are hardest to attribute, and the budget argument keeps getting made in the room where that is least visible. The EAB figures are self-reported by marketing executives, fielded in the summer of 2025.

What should you report when you can't prove attribution?

Write down, on one page, what your team can defend with data it holds. Application volume by program against the same point last year. Inquiry-to-application conversion where the inquiry came through your own form and the conversion fired on your own domain. That page is thinner but every line on it survives a follow-up question.

Then retire the claims that do not. A report claiming marketing generated a third of this year's intake, when no join exists between the ad account and the student record, will eventually meet a CFO who asks how.

What would I do first?

Put a "how did you first hear about us" question on the inquiry form and on the application, with a short fixed list and no free-text box. Self-reported source data is imperfect and well documented as imperfect. It is also the one signal that survives a stripped referrer and an 18-month gap, because you are asking the person rather than reading a header.

Run it for a full cycle before you draw anything from it, and when you compare it against what the ad platforms claim, note how far apart they are.

If you want the wider version — what your team can prove, what it is claiming, and the distance between the two — that is where the Marketing Function Audit & Reset starts.