What we're watching.
What it means for your CU.

Correlation is easy. You run an email campaign in March. Funded loans go up in April. You write in your board report that the email campaign drove loan growth. Everyone nods. Nobody pushes back. The number is real. The cause-and-effect claim is not.

Maybe loan volume goes up every April because tax refunds drive home improvement projects. Maybe the branch team ran a rate promotion at the same time. Maybe two hundred members who got the email would have applied anyway because their lease was expiring. You do not know. You correlated. You did not prove causation.

What causation actually requires

To move from correlation to causation, you need three things: identity resolution (the same person who received the campaign touch is the same person who funded the loan), journey mapping (you can see every touch between the campaign and the funded loan, not just the first and last), and counterfactual logic (you have a basis for knowing this loan would not have happened without that campaign touch).

None of those three things are possible when your campaign data lives in HubSpot, your funded loan data lives in your LOS, your member identity lives in your core, and none of them talk to each other.

Why this matters right now

AI tools are making it easier to produce confident-sounding attribution reports from disconnected data. The language gets sharper. The dashboards look cleaner. The underlying problem -- disconnected systems, unresolved member identities, no journey continuity -- does not go away because the presentation improved.

A report that looks authoritative is not the same as a report that is right. Your leadership will eventually figure out the difference. Better that the discovery happens on your terms, with a real solution in place, than during a budget review.

Attribution fails at the identity layer more often than at any other point in the process. Not because the technology for identity resolution does not exist -- it does, and it is quite good -- but because the data that feeds it is messier than most marketing teams expect when they start the project.

Consider what actually happens when a member touches your CU across multiple systems. They open a marketing email via HubSpot under the address they used when they joined fifteen years ago. They click through to your website and fill out a loan inquiry form using a different email -- their work address. They call your call center to ask a question, and the call center rep looks them up by member ID. They eventually apply through the LOS portal. They fund the loan. Your core records the transaction.

What we assess before anything is built

When Attrivix engages with a credit union, identity resolution complexity is one of the first things assessed. How fragmented are your member records across systems? What shared identifiers exist between your core, LOS, and CRM? The answers shape the attribution model and set realistic expectations about coverage from day one versus six months in, as the identity graph matures over time.

A credit union with clean, consistent member data gets to full attribution coverage faster than one with years of fragmented records. The identity layer is where most attribution efforts break down. It is also where Attrivix starts.

Every month, somewhere in a credit union marketing department, someone opens a spreadsheet they did not want to open. They know what the next several hours look like. Exports from the LOS. A data request to IT. Call center numbers pulled separately. All of it landing in Excel, where they will spend the better part of a week trying to stitch together an answer that leadership will receive with polite skepticism.

The answer they produce is not wrong, exactly. It is just not defensible. It is the best possible output from a process that was never designed to answer the question being asked.

The real cost is not the hours

The labor cost is real -- thirty hours a month at fully loaded rates is money leaving the building for a report nobody fully trusts. But the larger cost is strategic. When attribution is unverifiable, budget decisions default to inertia. Last year's allocation becomes this year's allocation. Channels that do not work keep getting funded because nobody can prove they do not work.

That is not a reporting problem. That is a lending performance problem.

What changes when attribution is provable

When attribution is provable, the monthly report stops being a project. It arrives. Your leadership reviews a document that names the loans, names the campaigns that drove them, and names the cost per loan by channel.

Instead of opening a spreadsheet, you forward an email. Your leadership reads it. Nobody asks follow-up questions.

That is what provable looks like. The question is whether what you are spending today is getting you there.