Cost per Issued Loan vs Cost per Lead: What to Optimise

The metric mismatch that costs lenders money

Your ad platform knows one thing: someone clicked, and then something happened on your site. Your credit operations team knows something else entirely: whether that person passed scoring, cleared KYC, accepted the offer, and had money disbursed.

If your campaigns are optimised on the first set of facts and your P&L is built on the second, you are running two different businesses. The gap between them is where marketing budget disappears.

Cost per lead (CPL) is the cost of a contact — a form fill, a phone number, a Meta lead form submission. Cost per issued loan (CPIL) is total media spend divided by loans actually disbursed from that spend. Between the two sit three filters that vary enormously by traffic source: application completion, credit approval, and issuance.

A campaign can have the lowest CPL in the account and the highest cost per issued loan. This happens routinely, and it is not a measurement glitch. It is the predictable result of buying attention from people who will not pass an affordability assessment.

The four stages, defined precisely

Before you measure anything, get the definitions agreed in writing with credit operations. Most reporting disputes between marketing and risk come from two teams using the same word for different things.

A workable set of definitions:

• Lead — an identifiable contact who has not submitted a complete credit application. A partial form, a lead-gen form on Meta, a callback request, a registered account with no application started.
• Application — a complete, submitted request for credit with the data your scoring engine needs, plus the required consents. The moment the decision engine receives it.
• Approved — a positive credit decision. Be explicit about whether this means an initial automated approval or a final approval after manual review and document verification. Many lenders need both: approved (auto) and approved (final).
• Issued — the contract is signed and funds are disbursed. This is the first event that generates revenue.

Add a fifth stage if your product has one: first repayment made. For short-term credit especially, an issued loan that never receives a payment is worse than no loan at all.

Each stage has its own drop-off causes, and they respond to different levers:

Marketing influences all three, but only indirectly for the middle one — by changing who arrives.

Where cheap leads lose money

The mechanism is simple: broad, cheap traffic shifts the composition of your applicant pool toward people your credit policy is designed to reject.

Common sources of cheap-but-worthless volume:

• Overly broad keywords and loose match types. Queries around urgent cash, debt consolidation under distress, or "money without checks" attract people who are already over-indebted. They are exactly the profile your affordability assessment under the EU Consumer Credit Directive is meant to screen out.
• Lead-generation forms with pre-filled fields. Meta's instant forms produce submissions at a low cost because they remove friction. Some of that friction was doing useful filtering work.
• Display, in-app and video placements with accidental or incentivised clicks. Volume arrives, intent does not.
• Geographic and age spillover. Non-resident clicks, under-age clicks, and clicks from regions where you do not lend. They cost the same as good ones.
• Ad copy that implies easy acceptance. Copy angles suggesting approval is automatic attract applicants who expect no checks. They do not convert and they generate complaints. Claims like "guaranteed approval" or "no credit check" are not permitted under Google's financial services policies and sit badly with consumer-credit rules in EU markets; use them only as an example of what to avoid.

None of this shows up in CPL. All of it shows up in CPIL.

A worked example (hypothetical figures)

Say you run two campaigns, each with €10,000 of monthly spend. These numbers are invented to show the arithmetic, not benchmarks.

On CPL, Campaign B looks 2.5× more efficient. On cost per issued loan it is roughly 1.7× more expensive. If your contribution margin per issued loan is, say, €90 in this example, Campaign A is profitable and Campaign B is burning money at scale — and the faster you scale it, the faster you burn.

Note also that the ranking flips at different points. On cost per application, B still looks better. The full picture only appears at the issuance line. This is why intermediate metrics mislead: each one is a partial view, and the partial views disagree.

Why cost per issued loan is still not the final answer

Optimising to issuance is a large improvement over optimising to leads. It is not the end state, and you should know the trade-off before you commit.

Issuance ignores credit performance. If one segment is approved and issued easily but defaults at a much higher rate, pushing bidding toward it will increase issued volume and reduce portfolio quality. Risk will notice before marketing does.

The practical compromise is to attach a value to the issued loan rather than counting it as 1. Options, in increasing order of sophistication:

1. Flat value per issued loan — simple, works when products are similar.
2. Value by product and principal — a €300 short-term loan and a €5,000 instalment loan are not the same conversion.
3. Expected contribution margin by segment — expected revenue minus expected credit losses, modelled by risk.

Even approach 2 is a meaningful step up from counting conversions. Value-based bidding in Google Ads and value optimisation in Meta can then allocate budget toward the loans worth having, not just the loans that happen.

The second trade-off is data latency and volume. Issuance happens days or weeks after the click. Smart bidding and Meta's delivery system need a reasonable flow of conversion signal to learn. If a campaign produces only a handful of issued loans per week, optimising directly on issuance will give you unstable bidding.

Ways to handle this:

• Optimise bidding on a mid-funnel event with value (completed application, or auto-approval) while reporting and budgeting on cost per issued loan.
• Use approval as the bid signal and issuance as the budget allocation signal. Approval arrives faster and in greater numbers.
• Consolidate campaign structure. Fewer, larger campaigns accumulate signal faster than many small ones.

How to measure each step

This is the engineering part, and it is where most lenders stall.

1. Carry a click identifier all the way to the core system

When a visitor lands, capture and store the platform click identifier — GCLID (plus wbraid/gbraid for app and iOS-affected traffic) for Google, fbclid and the derived _fbc/_fbp cookies for Meta — along with the landing page, campaign parameters and a timestamp.

Write that identifier into the application record in your CRM or loan management system. It must survive the whole journey: lead → application → decision → disbursement. If the identifier is lost at the handover between landing page and core system, no amount of reporting work afterwards will recover the link.

2. Stamp every stage with a timestamp

You need lead_at, application_at, decision_at, issued_at, and ideally first_payment_at. These timestamps let you build maturation curves — how long after the click each stage typically occurs — which you need for the next point.

3. Report by click cohort, not by event date

Dividing this week's spend by this week's issued loans is wrong when issuance lags clicks. You will systematically understate recent performance and panic about campaigns that are fine.

Instead, group by click date: all clicks from week 12, and everything that happened to them since. Then track how the cohort matures. Once you know, for your own business, roughly what share of eventual issuances have landed by day 7, day 14 and day 30, you can forecast an immature cohort's final CPIL instead of waiting a month to make a decision.

4. Send downstream events back to the platforms

Reporting tells you what happened. Feeding events back changes what the platforms do.

• Google Ads: offline conversion imports keyed on GCLID, uploaded with the conversion time and a value. Check the current click-to-conversion time limit in Google's documentation before you design your upload schedule — if your issuance lag exceeds the window, those conversions are silently discarded. Enhanced conversions for leads, keyed on hashed email or phone, are the alternative where click IDs are unreliable.
• Meta: the Conversions API with the original fbc/fbp values and an event_time set to the real event. Deduplicate against any browser-side events for the same action.
• ChatGPT Ads: the ecosystem is newer and the available conversion mechanics are still changing. Hold it to the same standard — if you cannot tie a click to an issued loan, treat its reported performance as provisional and keep its budget sized accordingly.

Upload daily or several times a week. Batch uploads once a month are too slow for the bidding system to use well.

5. Respect consent, deliberately

Sending conversion data back to ad platforms is processing personal data for advertising. Under GDPR and the ePrivacy rules, that generally requires the user's advertising consent, captured properly and stored with the record.

Practical implications:

• Your consent record needs to travel with the click identifier, so uploads can exclude non-consenting users.
• Send the fact of a conversion, not the credit decision reasoning, scores, or any special-category data.
• Your pixel and tag behaviour before consent matters as much as the uploads.
• Confirm your setup with your DPO and compliance team, and read the current official texts. Nothing here is legal advice.

What to do this week

1. Write down the five stage definitions and get sign-off from risk and credit operations. One page. This alone resolves most reporting arguments.
2. Check whether the click identifier reaches your CRM. Pull 20 recent issued loans and see how many have a usable GCLID or fbc value. If the answer is low, that is your first engineering ticket.
3. Build one cohort table: by click week, by campaign — spend, leads, applications, approvals, issued loans, and cost at each stage. Even in a spreadsheet, it will change at least one budget decision.
4. Find your biggest CPL/CPIL inversion. Rank campaigns by CPL and by cost per issued loan. Where the rankings disagree most, you have either a budget to cut or a targeting problem to fix.
5. Audit ad copy and landing pages against the Google Ads financial services policy and your market's consumer credit rules — representative example where required, clear APR and total cost presentation, no claims implying approval without assessment. Verification requirements for financial services advertisers differ by country; check the current policy text for Latvia, Spain or wherever you run.

For the measurement plumbing itself, Ads Rehub — an internal tool operated by SIA Batwatex in Latvia — connects landing pages, Google Ads, Meta and ChatGPT Ads accounts with the lender's CRM, reports cost per application, per approved and per issued loan, and sends approved and issued loans back to Google and Meta as offline conversions for visitors who gave advertising consent. Its ad library checks copy against platform and consumer-credit rules and exports ads paused, so nothing goes live unreviewed.

Key takeaways

• Cost per lead measures the cheapness of attention; cost per issued loan measures the cost of revenue. Campaigns frequently rank in opposite order on the two metrics.
• Agree precise definitions of lead, application, approved and issued with credit operations before building any reporting — most disputes are definitional, not analytical.
• Carry the platform click identifier into the CRM and stamp every stage with a timestamp; without that link, downstream optimisation is impossible.
• Report on click cohorts rather than event dates, because approval and issuance lag the click and will distort same-period comparisons.
• Feed approvals and issued loans back as offline conversions with values, for consented users only, and verify the consent and data-handling design with your compliance team.