Performance review template for Fintech

The performance review with the columns a regulated financial product needs: every line read on funded accounts rather than applications, approval rate stated beside the cost, and causes that include the two things advertising does not control — underwriting and the review cycle.

Everything else is the general review, unchanged. Planned from the plan, variance separated from cause, cause held to evidence, one decision with one owner and a date.

Run it monthly, and again after any change to a rate, a fee or a term, because that change resets the economics of every line in the plan.

Run it far enough after the spend that funded accounts have landed. A review read on applications will reach the opposite conclusion from one read on funded accounts, reliably and confidently.

The single decision that shapes this review is which number goes in the result column, and it should be the one the business counts.

An application is not revenue. A line that promises more than the product delivers, or reaches people further from qualifying, will produce applications cheaply and funded accounts rarely, and on an application-based view it is the best line in the account. Every report the platform produces will agree with that view.

So the result column reads funded accounts, or approved applications where funding is not a separate step, from the underwriting data rather than from the ad platform. That makes the review slower — funded outcomes lag — which is why the cadence is monthly rather than weekly and why the review is run far enough behind the spend for the cohort to have settled.

Every row carries the approval rate for its cohort, and it is the column that separates an advertising result from somebody else's.

A cost per funded account that rose while the approval rate held is an advertising signal: the line is reaching fewer or more expensive people. The same cost rising while the approval rate fell is usually not an advertising signal at all, and changing creative or audiences in response will change the wrong thing.

This is the most common misattribution in the category, and one extra column removes it.

A change to a published rate, fee or term resets the economics of every line at once.

Everything before the change and everything after it are different measurements. So the review records the date, splits the period at it, and compares like with like. A blended month spanning a rate change describes neither half, and it will be quoted later as though it described both.

The same change usually forces a creative replacement, because wording cleared against the old terms is stale from that day. That has its own consequences, which is the next cause.

When creative has to be replaced, there is a gap: days where the old assets were out of market and the new ones were still being cleared, or where a weaker fallback was running.

That gap shows up in the numbers as a dip, and from inside a platform it looks like fatigue or like a delivery problem. It is neither. It is an operational cost of the review cycle, and it belongs in the cause column with the dates.

Recorded consistently, it also produces something useful: a measure of what the review cycle actually costs in market presence, which is the evidence for whether the process is worth changing.

Retargeting audiences built from application starts almost always show the best cost per funded account in the account, and that number means something different from the others.

Those people had already applied. The line did not create the intent; it recovered a proportion of an intent that prospecting produced. Reading it as a standalone efficiency and moving money towards it will, over a quarter, starve the campaigns that generate the applications it recovers.

So the review states it plainly in the notes and treats the retargeting row as dependent rather than comparable. The decision that follows is almost never more weight.

Four decisions, and in this category they divide unusually cleanly.

Continue, where both the cost and the approval rate held. Change the creative or the audience, where the cost moved and the approval rate did not — that is an advertising problem. Investigate with underwriting, where the approval rate moved, because that is not an advertising problem and the fastest route to understanding it is a conversation. And stop, where a line's funded economics have never worked at any weight.

The second and third look identical on a dashboard and are completely different pieces of work, which is the whole reason for the approval-rate column.

The finished reviews are worth keeping alongside the review register, because together they answer a question that neither answers alone: which cleared claims have actually produced funded customers.

Over a year that becomes the most useful document in the account. It tells the next campaign what has worked, what was refused, and what was cleared and then failed in market — and the third category is the one nobody records anywhere else.

Demand for a financial product moves with things nobody in marketing controls: a rate announcement, a competitor's promotion, the end of a financial year, a news story about the category.

Each of those can move volumes and costs by more than any campaign decision in the period, and none of them is visible in an ad account. So the notes column records them with dates, at the time, while somebody still remembers.

The discipline matters because the alternative is worse than forgetting. A month distorted by an external event, with no note attached, sits in the record as evidence about a campaign and shapes the next plan. A month with the event recorded is a month everybody knows to read carefully.

Seven checks. Every row is read on funded accounts or approved applications, from the underwriting data. The approval rate for each cohort is stated beside the cost. Any change to a rate, fee or term is recorded with its date and the period is split at it. Gaps caused by the review cycle are named as causes with their dates. Retargeting rows are marked as dependent rather than compared with prospecting. Every decision is one of four, with an owner and a date. And the finished review is filed with the review register.