Counteroffers: Approving a Different Loan Instead of Declining | HL Hunt
Counteroffers: Approving a Different Loan Instead of Declining
An applicant fails a policy test and the standard output is a decline, which is worth nothing to anyone. Frequently they would have qualified for a smaller or differently structured loan: in the case below, three of five structures clear a ceiling the requested one failed. But there is a trap in the arithmetic. The only structure that clears the ceiling at the full amount is the longest term, and it costs the borrower $4,151 more in interest. The easiest counteroffer is usually the worst one.
What you'll learn
A decline is worth nothing to anyone
An applicant fails a policy test. The standard output is a decline, which produces no revenue, no relationship, and an applicant who needed something they did not get.
Frequently the applicant would have qualified for a different offer. Not a riskier one, a smaller or differently structured one, which is a distinction most decline logic does not make.
Take an applicant with monthly income of $3,400, existing obligations of $1,150, requesting $12,000 over 36 months at 26%. The policy ceiling is a 45% debt-to-income ratio.
| Offer | Monthly payment | Resulting DTI | Outcome |
|---|---|---|---|
| $12,000 / 36 months (requested) | $483.49 | 48.0% | Fail |
| $12,000 / 48 months | $404.61 | 45.7% | Fail |
| $12,000 / 60 months | $359.29 | 44.4% | Pass |
| $9,000 / 42 months | $328.53 | 43.5% | Pass |
| $7,500 / 36 months | $302.18 | 42.7% | Pass |
Three of five structures clear the ceiling. The applicant was not unapprovable; the requested structure was.
The counteroffer that passes and shouldn't
Notice which option clears the ceiling while giving the applicant the full amount: the 60-month term. Now price it.
| $12,000 at 26% | Monthly | Total paid | Total interest |
|---|---|---|---|
| 36 months | $483.49 | $17,406 | $5,406 |
| 48 months | $404.61 | $19,421 | $7,421 |
| 60 months | $359.29 | $21,557 | $9,557 |
This is the central problem with counteroffer design and it is almost never stated. A debt-to-income ceiling measures monthly payment against monthly income, so the cheapest way to pass it is to lower the payment, and the cheapest way to lower the payment is to extend the term. That reduces the measured risk and increases the total cost, and per our measurement analysis optimizing against a measure rather than the thing it stands for produces exactly this.
Per our pricing analysis, the honest framing is that term extension trades affordability for cost, and whether that trade is right depends on the borrower's situation rather than on whether it clears your policy. An automated counteroffer engine that defaults to term extension because it maximizes approvals is optimizing the wrong variable.
A defensible ordering
If term extension is the easiest counteroffer and often the worst, the ordering should be explicit rather than emergent.
- Reduce the amount first. $7,500 over 36 months clears the ceiling with the lowest total cost of any passing option. It gives the applicant less than they asked for and costs them least.
- Then adjust the rate if the file supports it. A better rate lowers the payment without extending exposure.
- Then consider security or a guarantor, per our cosigning analysis, recognizing that this transfers the risk rather than reducing it.
- Then extend the term, with the total cost difference disclosed alongside the payment.
- Then decline, with reasons that describe what would change the answer.
Item one is counterintuitive for a lender and usually correct. A smaller loan books less principal and generates less contribution per account, which is why the engine will not choose it unprompted. It is also the option a borrower is most likely to repay, and per our exposure analysis the one that leaves them room for the obligations you cannot see.
Item four should carry the comparison. If a 60-month offer is made, showing the applicant that it costs $4,151 more than a 36-month structure they cannot currently afford is both the honest presentation and the one most likely to produce a borrower who asks for less instead.
What counteroffering is worth
The commercial case does not depend on anything subtle.
- Value of a decline: $0
- Counteroffer of $7,500 at a 38% acceptance rate, contributing 9% of principal: expected $256 per declined applicant counter-offered
- Counteroffer of $9,000 on the same assumptions: expected $308
Per our review economics analysis, that is a large expected value for a computation that happens automatically and costs nothing per case. The reason it is often absent is not economics; it is that decline is the terminal state in most decision architectures and adding a branch after it requires someone to decide it should exist.
There is also a selection effect running the right way. Per our acceptance analysis, an applicant who accepts a reduced offer has revealed that they needed the money enough to take less, which is weak evidence of motivation to repay. An applicant who declines a counteroffer has cost you nothing.
Where a counteroffer becomes a different product
A counteroffer that changes the amount or the term is a variation on the same product. A counteroffer that moves the applicant to a materially different product, a secured facility, a much higher rate, or a different structure entirely, is a new offer and should be assessed, disclosed, and documented as one. Per our adverse action analysis, a declined application that results in a different offer still has notice implications, and the specifics vary by product and jurisdiction. This is a question for compliance before it is a question for product.
Designing the engine
- Compute the full passing set, not the first passing option. Three of five structures passed above, and the engine should know all three before choosing.
- Rank by borrower total cost, not by principal booked. This is the single decision that determines whether the engine serves the applicant or the volume target.
- Cap the term extension available as a matter of policy, so the cheapest passing route cannot be the longest.
- Present the comparison. Payment and total cost for the counteroffer and for the requested structure, side by side.
- Keep the counteroffer open for a stated period, per our latency analysis, so an applicant who wants to reduce what they are borrowing elsewhere first can come back.
- Log the counteroffer, the acceptance, and the outcome, because this is the only way to learn which counteroffers perform.
- Monitor performance of counter-offered accounts separately, per our monitoring analysis, since they are a different population from directly approved ones.
- Report acceptance and performance by group, because a counteroffer process is a discretionary layer on top of a decision and per our exception analysis that is where uneven outcomes accumulate unmeasured.
Item two is the whole design. An engine ranking passing structures by principal will offer the 60-month full-amount loan every time, because it books the most. An engine ranking by borrower total cost will offer $7,500 over 36 months. Both are compliant, both pass the policy, and they are different businesses.
Item seven matters for a reason that is easy to miss. Counter-offered accounts were declined once, which means they sit on the wrong side of a cutoff your model drew, and per our threshold analysis their performance is the best available evidence about whether that cutoff is in the right place.
Counteroffers ranked on the borrower's cost
HL Hunt AI Underwriting computes every structure that clears policy rather than the first one, ranks them by total cost to the borrower rather than by principal booked, caps available term extension, and tracks counter-offered accounts as a separate population so the cutoff they failed can be tested against their actual performance.
What to measure
| Metric | Why |
|---|---|
| Share of declines where a passing structure existed | The size of the opportunity, and most lenders have never computed it |
| Counteroffer acceptance rate by type | Reduced amount and extended term are accepted at different rates and perform differently |
| Performance of counter-offered accounts against directly approved ones | Tests whether the original cutoff was right |
| Average total cost of accepted counteroffers | Whether the engine is ranking on the borrower's cost or on yours |
| Term distribution of counteroffers | A distribution clustered at the maximum term is the engine optimizing the ceiling |
| Acceptance and performance by group | A discretionary layer on a decision needs the same fair treatment monitoring as the decision |
The first row is the diagnostic to run before building anything. Take last quarter's declines, recompute each one against a set of alternative structures, and count how many had at least one that cleared policy. That number is the whole business case, it requires no new system, and for most lenders who have never looked it is considerably larger than expected.
The fifth row is the check on your own engine. If counteroffer terms cluster at your maximum, the engine found the easiest way to pass the ceiling and took it every time, which is the failure this article is about.
Make the decline a branch, not a terminal state
Get started with HL Hunt AI Underwriting to recompute last quarter\'s declines against alternative structures and size the opportunity before building anything, then run counteroffers with logged acceptance, separate performance monitoring, and term distributions you can actually see.
Frequently asked questions
Often. In the worked case three of five alternative structures cleared a 45% DTI ceiling that the requested structure failed at 48.0%.
It reduces measured risk and increases total cost. On $12,000 at 26%, stretching from 36 to 60 months costs the borrower $4,151 more in interest.
Reduce the amount first, then rate, then security, then term extension with the cost difference disclosed, then decline. The easiest option is usually the worst.
A decline is worth $0. A $7,500 counteroffer at a 38% acceptance rate and 9% contribution is worth about $256 per declined applicant counter-offered.
Key takeaways
- A decline usually reflects the requested structure rather than the applicant: three of five alternatives cleared the ceiling in the worked case.
- A debt-to-income ceiling measures payment against income, so the cheapest way to pass it is to extend the term.
- Stretching $12,000 at 26% from 36 to 60 months costs the borrower $4,151 more while reducing measured risk.
- Reduce the amount before extending the term: $7,500 over 36 months cleared the ceiling at the lowest total cost.
- Rank passing structures by borrower total cost rather than principal booked; that single choice defines the business.
- Cap available term extension in policy, or the engine will cluster counteroffers at the maximum term.
- A decline is worth $0 against roughly $256 expected per declined applicant counter-offered at $7,500.
- Monitor counter-offered accounts separately, since they failed a cutoff and are the best evidence about whether it is correctly placed.
This guide is educational and does not constitute legal, compliance, or model risk advice. All figures are stylized illustrations using assumed incomes, obligations, rates, terms, acceptance rates, and contribution margins, and do not describe any particular lender or applicant. Adverse action notice obligations where an application results in a different offer, affordability assessment requirements, disclosure rules, and fair lending expectations for discretionary or counteroffer layers vary by product and jurisdiction. Consult your compliance, legal, and model risk functions before implementing any counteroffer process.