Google Ads for mortgage brokers: the build order

For mortgage brokers the conversion that matters is a funded loan in the LOS, with the revenue it earned you, and it lives in your LOS and mortgage CRM (ICE Encompass, ARIVE or Byte, with Total Expert, Shape or Bonzo) — not in the ad platform. Google can only optimize toward what you send it, so whatever you pick here decides what the algorithm spends the next ninety days chasing. Pick the wrong one and every later step in this playbook is executed perfectly in the wrong direction.

This is the step we get called about most. If you do not have a developer for it, that is the normal situation rather than the exception — it is most of what we do, and it is worth fixing before the spend rather than after.

Google enables these on every new Search campaign, and none of them are keyword decisions — they are switches on the settings screen that decide where the money goes before a single keyword applies. Everything else in this playbook operates on whatever is left after they have taken their cut, so a reader who skips this step spends the rest of it optimizing a remainder.

For Mortgage · Target licensed states, then counties or cities; never ZIP codes or demographics for credit offers. Keep purchase, refinance and government loans in separate campaigns so each has its own budget.

A broker can lend only where it holds licenses, and Google's US rules for credit offers mean the account cannot target by ZIP code, age, gender, parental status or marital status. So the map starts with licensed states and narrows to counties or cities where your branches and realtor partners are, not the ZIP-level radius a local trade would draw. Purchase borrowers lean toward the loan officer their realtor trusts; refinance borrowers will close with anyone licensed in their state.

Borrowers call when the scenario is complicated: self-employed income, a VA entitlement question, a contract that has to close before the rate lock runs out. Those calls are worth more than any form. The same number also rings with existing borrowers asking about payoffs and escrow, realtors checking on a file, and lead sellers. Count a call only when a new borrower describes a scenario, and let the started application or pre-approval in the LOS be what bidding learns from.

One flat keyword list is why most accounts in this vertical cannot tell which half of the budget works. Sorting terms by the intent behind them means you can bid, write and judge each group on its own terms — a problem-aware search and a competitor search convert at rates that have nothing to do with each other, and averaging them hides both.

Exact match includes close variants — misspellings, plurals, reordered words, dropped function words. Phrase match now behaves roughly like the old broad-match-modified. Broad match reaches as far as Google judges useful, which with a good conversion signal is genuinely valuable and without one is an open tap. Start narrow, prove the signal, then widen — in that order, because widening is reversible and a mistrained bidding algorithm is not.

For Mortgage · Broad match on 'mortgage' drifts into payment, login, servicing and job searches, and into national lender brand names. Keep purchase and VA terms on exact and phrase until applications import from the LOS.

Every vertical has a set of searches that look commercial and are not. In mortgage brokers those are job seekers, students, people looking for free help, and DIY researchers. Broad and phrase match will find all of them for you. Add this list at the account level on day zero — the first week of an unfiltered account is the most expensive week it will ever have.

Ad groups exist so the ad can match the search. The moment a group holds two intents, the ad has to be vague enough to cover both, and vague ads lose the auction on relevance before budget ever enters the picture. Build these groups, each with its own ads and its own landing page.

Google rewards assets that match the search and the page behind it. It penalizes — and increasingly disapproves — copy that promises something the page does not deliver. Every line below is written to be provable on arrival. Character counts are shown because assets over the limit are rejected at upload, which is where most copied-from-a-blog ad copy dies.

Mortgage brokers and lenders have no Local Services Ads category in the US, so no pay-per-lead unit sits above your Search ads; national lenders and rate tables sit there instead. LSA still reaches your business through referral partners. Google's list includes real estate services, so the agents who send you purchase borrowers may be buying LSA leads, and the borrowers they pass on will search your name when told to call you.

Mortgage demand runs on two clocks. Purchase follows the home-buying season, building in late winter, strongest through spring and summer and slowing toward the holidays. Refinance follows rates, not the calendar: a drop brings a wave of refinance searches within days, and a rise ends it just as fast. A flat budget overspends on refinance in a quiet rate month and runs out in a busy one.

Position is not bought, it is earned and then paid for. Google ranks you on your bid multiplied by how good it expects your ad and landing page to be, plus your assets — so a relevant advertiser at half the bid outranks a vague one, and then pays less per click at the same position. This is the mechanism underneath everything above: tight ad groups and honest ad copy are not tidiness, they are the discount.

For Mortgage · Rate tables, marketplaces and national lenders bid on every rate term. Compete on scenarios and local relationships, and keep rates and payments out of ad copy unless compliance has approved the disclosures.

Start on Maximize Conversions with 'application started' in the borrower portal, or a qualified call, as the Primary goal, never the rate-quote form or the calculator. Carry the click ID into Floify, Blend or ARIVE and onto the loan file, import 'pre-approval issued' and 'application submitted' from the LOS while they are inside the 90-day import window, and send funded loans with your revenue as value. Purchase loans often fund after that window closes, because the borrower still has to find a house, so the earlier milestones are what bidding really learns from. Keep purchase and refinance in separate campaigns: refinance volume swings with rates and would drag the purchase bids with it. Smart bidding is a data problem before it is a strategy problem: with thin or wrong conversion data it will confidently optimize toward noise, and it does that faster than a manual account can lose money.

For Mortgage · Bid on application started or qualified calls first, add pre-approval issued from the LOS, and pass funded-loan revenue as value once enough loans close inside the import window.

Most accounts are killed or rescued on evidence that could not have meant anything yet. The bidding algorithm is fitting a model, and a model fitted on two weeks of data is fitted on noise. This is the shortest honest schedule: what to look at each stretch, and — more usefully — what to ignore.

For Mortgage · Purchase loans fund slowly, so judge the first month on applications and pre-approvals, and read funded loans by campaign only once a full purchase cycle has passed.

An account left alone drifts toward the cheapest clicks it can find, because that is what the system is built to do when nobody corrects it. This is the shortest routine that keeps it honest. It takes about an hour and it is the difference between an account that compounds and one that plateaus at month three.

For Mortgage · Compare Google's conversions with applications started in the portal for each campaign. A campaign that fills quote forms but starts no applications is buying rate shoppers.