Which ads produced funded loans?

Two channels can deliver inquiries at the same cost while one produces borrowers who qualify and the other produces people who will not fund for years, if ever. Cost per inquiry is the number everyone optimizes and it is close to uninformative here.

Matching funded loans back to the inquiry that produced them ranks channels on commission earned. That comparison frequently inverts the apparent winner, because the cheapest inquiries usually come from the least qualified audience.

Volume in this business moves with rates as much as with marketing. Comparing this month with last month tells you about the rate environment, not about your campaigns.

Comparing the same period a year earlier, or holding a channel's mix constant across a rate move, is the only way to see what the advertising is contributing. Because nothing had to be installed at the time, you can run those comparisons over periods that have already closed.

Borrowers collect pre-approvals the way they collect quotes. One goes to the bank holding their current account, one to whoever answered the phone first, one to the broker their real estate agent recommended. Several files get opened, credit is pulled more than once, payslips are uploaded to more than one portal, and exactly one of them funds. Everything before funding is contested, and a borrower can sit in your pipeline as an active, well-progressed file while intending to close somewhere else entirely.

That makes every pre-funding metric a measure of being early rather than a measure of winning. Applications started rewards the channel that reaches people at the very beginning of their shopping, which is usually the cheapest traffic and the least decided. Conditional approvals issued rewards whichever team processes documents fastest, which is an operations result rather than a marketing one. Neither number tells a principal what a channel was worth, and both are available within days of the spend, which is precisely why they are the ones everybody reports on.

Ranking on funded loans and the commission actually collected removes the ambiguity in a single step, because a funding is the only event in the whole sequence that cannot happen twice for the same borrower. A source that produces a great many opened files and comparatively few fundings is losing borrowers somewhere between the pre-approval and the offer, to a bank or to another broker, and that is a concrete operational fact worth acting on rather than a suspicion the sales manager repeats every quarter without evidence.

A large share of most brokerages runs on real estate agent relationships. Coffee, closing gifts, joint open days, a contribution to a listing's advertising, a sponsored table at an office awards night. All of it costs money and time, all of it competes for the same budget as paid search, and none of it appears in any marketing report because no advertising platform can see it. It is managed on the loan officer's memory of who has been sending business lately.

A referral source only needs the borrowers it introduced, each with a phone number or email, tagged with the referring agent's name as the source. Add those rows to the lead file and every borrower introduced by each agent is ranked on commission collected, on the same axis and inside the same table as paid search, social and the brokerage's own database. Nothing has to be tagged in the CRM at the time, because the join happens afterwards on the borrower's contact details rather than on a field somebody filled in.

The distribution that comes back is usually severe. A handful of relationships account for most of the referred fundings while a long tail produces nothing measurable at all, and several of the agents everybody assumed were productive turn out to have introduced one borrower two years ago and nothing since. That reordering is what moves the co-marketing money, and because it rests on the brokerage's own funded files rather than on an estimate, it survives contact with the loan officer whose relationship is being questioned.

Basis points are not constant across a brokerage's book. A jumbo purchase, a small refinance, a government-backed loan and a second charge can all pay differently for broadly similar amounts of work, and lender paid and borrower paid arrangements diverge again on top of that. Ranking channels on volume funded therefore ranks them on average loan size, which is largely a property of the ZIP codes a campaign happens to reach rather than a property of the campaign's contribution to the business. The two can move in opposite directions for a whole year without anyone noticing.

The origination system holds everything the match needs and nothing it does not: the borrower's phone number and email address, the funding date, the loan amount and whatever compensation figure the brokerage records against the file. What it has never held is the keyword, the campaign or the name of the agent who made the introduction, and no amount of configuration will put them there, because those facts were created in a different system weeks or months before the file was opened.

So the export worth building is commission rather than principal, with a product column mapped as Product / service beside it and a funding date on every row. Substituting one column for the other can reorder the channel list outright, because the sources that produce the largest loans and the sources that produce the best-paying files are seldom the same, and only one of those orderings has anything to do with what the business actually earned. A brokerage that has never run the comparison should be ready for it to be uncomfortable.