Industries · Business lending
Value-Based Bidding for Business Lending: Bid on Funded Loans, Not Applications
Applications are easy to generate and most never fund. Use your own history of what funded, and what it earned, to estimate the contribution each new application is likely to bring, so bidding learns to find the borrowers who close.
Why lead counts fall short
Why a count of applications misses what a funded loan is worth
Lending funnels lose most of their volume between application and funding. An enquiry from a business with two years of trading and steady revenue, and one from a startup with no revenue yet, arrive as the same lead, and bidding on the count pays for whichever is cheaper to generate.
Your loan origination system knows what each kind of application became. Loan type, amount requested, time in business and industry all show up in funding rates and in the fee or contribution a funded loan earned. That is the signal the ad platforms need.
Funding rate and fee move separately
Small working-capital loans fund often for a small fee; large equipment or term loans fund less often for far more. Expected value weighs both.
Contribution, not principal
The lender earns a fee, a margin or a commission, not the loan amount. Price on the contribution your system records on a funded loan and the values mean what the business earns.
Credit data stays out
Personal credit scores and consumer credit data are refused by column name. Loan type, amount requested, time in business, industry and revenue band are the fields to use, and they are facts about the business, not the person.
The funnel
From application to funded
- Application
- Documents received
- Approved
- Funded
Known when the lead arrives
What a new lead can be priced from
- Loan type
- Amount requested
- Time in business
- Industry
- Annual revenue band
Learned later
What the history is trained on
- Whether documents came in
- Whether the loan was approved
- Whether it funded, and the fee or contribution
The loan type, the amount and the facts about the business on the application are what price a new lead. Documents, approval and funding are what the history is trained on.
How it works
From your sales history to what the ad platforms see
- 1
Bring in your history
A CRM export or a HubSpot connection: leads with their outcomes, won, lost or still open.
- 2
Find what relates to value
Each field known when a lead arrives is tested against your closed deals. Only the ones that clearly moved outcomes are kept.
- 3
Estimate each new lead
Its expected value: how likely leads like it are to become customers, times what those customers were worth.
- 4
Send the value
To Google Ads or Meta, as a conversion value on the lead. Sending changes no bid by itself.
- 5
Judge it on real outcomes
Compare closed deals before and after your campaigns start optimising on the values, never the values we sent.
Worked example
Three leads, three different values
Expected lead value = the estimated chance of becoming a customer × the expected value of that customer.
Value basis: Fee or contribution on the funded loan. The origination fee, broker commission or contribution margin your system records when a loan funds. Use one basis across the export; the product does not model interest earned over the life of the loan.
Equipment loan, $250,000 requested
- Five years in business
- Revenue band: $1M to $5M
14% × $7,500
$1,050
estimated value of this lead
Working capital, $60,000 requested
- Two years in business
- Revenue band: $250k to $1M
20% × $2,400
$480
estimated value of this lead
Startup, $100,000 requested
- Under one year in business
- No revenue yet
2% × $3,000
$60
estimated value of this lead
Illustrative example, not customer results.
The equipment loan is worth about twice the working-capital application and seventeen times the startup enquiry. The startup asks for more than the working-capital borrower and is worth a fraction of it, because it almost never funds.
How the product actually estimates this
It does not score each lead on its own. It starts from your overall close rate times your average won deal, with any deal above three times the median counted at that cap, then applies a multiplier for each field that clearly moved outcomes in your history, and rescales so the average estimate matches what your leads were actually worth. Fields that do not clear the evidence thresholds are left out and listed. The formula above is the same idea in one line.
What data you need
Your CRM already has most of it
- Application dateWhen the application arrived, not the funding date.
- Funded, declined or still openDeclined and abandoned applications teach the model as much as the funded ones.
- Fee or contribution on funded loansOne basis for the whole export: origination fee, commission or contribution margin.
- Fields from the application formLoan type, amount requested, time in business, industry and revenue band. No personal credit data.
- Matching informationGoogle's click ID or the applicant's email for Google; email or phone, plus Meta's cookies from your site, for Meta.
What matters most is reliable outcomes and enough of them. The report checks your own file and names anything it could not price, rather than applying a one-size minimum.
How activation works
Estimating, sending and optimising are three steps
The product estimates each lead's value and sends it. Your campaigns use it only once you set them to optimise for value, a change you make in the ad platform, and the platform decides when an account qualifies.
Google Ads
Sends each application's estimated value to a conversion action in Google Ads, matched on the click ID or on email. Your campaigns use it only once they bid on value, for example with Maximize conversion value.
Switching a live campaign, step by stepMeta
Sends a ValuedLead event with the value through the Conversions API. Website-form ad sets bid on values once they qualify for value optimisation; instant-form ad sets can learn from the documents and approval stages through Conversion Leads.
How the Meta route worksThe full method, from pricing a lead to measuring the result, is in the complete guide. Leads that reach your CRM without a click ID are covered in the click ID guide.
How to judge success
Measured in outcomes, not in the values we sent
Funded loans per application
By loan type, before and after the switch.
Cost per funded loan
Ad spend against loans funded, not against applications.
Contribution per application
Recorded fees or margin divided by the applications that produced them.
Reported conversion value going up is not the result; it repeats what we sent. The product's evaluation compares funded loans before and after the switch, with applications from other sources as a control.
Questions
Business lending, specifically
Can we price on the applicant's credit score?
No. Consumer credit data is sensitive, the ad platforms restrict bidding on it, and the product refuses credit columns by name. Time in business, revenue band, industry and loan type are facts about the business and carry the signal.
We are a broker, not a lender. Does the same model apply?
Yes, with the commission you earn on a placed loan as the value basis. The funnel is the same: application, documents, approval by a lender, funded.
Should the value be the fee, or the interest over the loan's life?
The fee or contribution your system records when the loan funds. Interest over the life of the loan depends on repayment and is not forecast, so the values do not claim it.
Our funding rate changed when we tightened criteria. Does old history mislead?
It can. Applications from before the change still show which kinds of business fund, but at rates that no longer hold. Fit on the period that reflects today's criteria, and refit when they change again; the report says when fresh deals have drifted far enough.
Check whether your lead data is ready
Upload a CRM export or connect HubSpot. In about five minutes you see which of your fields relate to value, what new leads would be estimated at, and whether your history is enough to send. The file is read in your browser; you are asked for a name and work email before the full report.