Outcome-based mechanics: pricing a claims operation on results, not seats.

Every insurer I talk to would rather pay for a claim that closes faster than for a seat. So would I.

The obstacle is not appetite on either side. Under a per-seat licence my revenue tracks your headcount. Under a time-and-materials contract, where I bill for hours worked, it tracks how long the work takes. In both, your operation getting better does not change what I earn.

Everyone can see that, and it is still the model most insurance software is sold under, because the alternative is genuinely harder to write down.

It gets written down well or badly in two places.

What makes an outcome measurable

An outcome two parties can price has a start event, an end event, and a stated rule for the cases in between.

Cycle time from the first report of a loss to settlement is measurable. Improved claims performance is not, and an agreement written on the second version becomes an argument the first time the number matters.

The edge cases are what a definition is actually for, and they are not exotic. A claim reopened after settlement. A claim withdrawn by the claimant. A claim held pending a police report. A claim where the vehicle owner is unreachable for three weeks. Each needs a stated treatment agreed in advance, because each will happen, and whoever raises it first afterwards looks like they are changing the rules to suit themselves, even when they are not.

There is a subtler edge case I would now put in every agreement, having got it wrong before. Different parties in the same claim are on different clocks, and correctly so. In our motor work the insurer's steps run against working hours.

In Vietnam, it runs from 08:00 to 17:00 on weekdays excluding public holidays. The service centre's steps run around the clock every day, because a workshop's obligation does not pause at five. Measure both against one calendar and you produce a number one side can always dispute. Write both calendars into the agreement.

Where the number has to come from

If I am paid on a number, that number cannot come from a system only I can see. It has to derive from the same audit trail you can inspect: each event, its timestamp, its actor, in sequence.

This is the practical reason I think the audit model and the commercial model are one piece of engineering rather than two. A trail built to satisfy a regulator already has the properties a fee calculation needs. It is written as the work happens, it is complete, and it is not mine to revise.

It also means the accountability model has to be specific before the commercial model can be. We break a motor claim into nine steps. Each names the party responsible, the event that starts its clock, the event that stops it, and the time it is allowed its allotment. Insurers get 1, 2 or 4 hours to accept a claim notification depending on region. Quote approval gets more time the larger the repair, from hours to days. Compensation payment runs five working days.

You do not have to adopt those numbers, and they are configuration rather than architecture. The point is that a fee cannot attach to a cycle time until somebody has written down what the cycle is made of and who owns each part.

In Vietnam there is a useful starting floor, because the regulator has already done some of this. Every vehicle must carry third-party liability cover, and Decree 67/2023 puts that cover on a clock: the insurer must tell the claimant how to file within one hour of being notified, assess damage within 24 hours, and pay an advance against injury or death within three working days.

Where a commercial model and a regulatory obligation are measured on the same trail, the compliance work and the commercial work stop being two exercises.

The real problem: Attribution

Here is the objection I would raise if I were you.

Of those nine steps, more sit with the insurer than with anyone else. That is not a negotiating position, it is where the decisions are. But it means a fee attached to end-to-end cycle time can penalise me when your quote approvals are slow, and reward me when they are fast. In neither case have I done anything.

So end-to-end cycle time is the wrong thing to price on its own, however clean it looks on a term sheet. The versions that survive are narrower. Pay on the steps the platform can actually move. Or pay on a composite, with counterparty-caused delay carved out and the carve-out defined in advance. Whoever proposes end-to-end and leaves attribution unsaid has not finished the work.

What reconciliation has to produce

Reconciliation should run on an agreed period and produce the claim-level detail behind the figure, not the figure alone. Your finance team should be able to take the period's total apart and put it back together without asking for a spreadsheet. If they have to ask, the number is mine rather than shared.

There is one more thing to settle in advance, and it is uncomfortable enough that it usually goes unmentioned. Whoever runs the operation during the baseline period influences the price that gets struck against it, and has an incentive to look slower than later. Agree who sets the baseline window, how long it runs, and what independent evidence fixes it before it starts.

What the model would ask of each of us

If I priced on results, I would carry the risk of the operation performing. A bad quarter in your claims operation would be a bad quarter in my revenue, and I could not recover a difficult month by billing more hours. I would rather have my incentive on that side of the number, and I want to describe it in the conditional, because we have not signed one of these yet and I do not want to claim credit for a risk I am not currently carrying.

What it would ask of you is smaller in the contract and larger in the building. Audit rights, and the willingness to have your own steps measured. That second one is what stalls these conversations, and understandably.

Your acceptance touchpoint is the one that will breach most often, and somebody has to carry that number into a management meeting where it has never been visible before. It helps to agree in advance that the first quarter's figures are for calibration and not for penalty.

Why I would not sign one today

We have not signed an outcome-based agreement, and there are two standing rules behind that.

On the claims side, we do not finalise an outcome-based service-level model before pilot data exists, because a target set from a projection is a number I would be asking you to trust rather than check. On the distribution side, we do not finalise commercial terms before audit rights over counter-party policy sales and the reconciliation infrastructure are both in place.

I should also be straight about what that means commercially, because it is the part that could be read as having it both ways. It means the near-term work is priced conventionally while the measurement gets built. If you want the aligned model, the thing to negotiate now is not the rate. It is the trigger: what has to be true, by when, for the conventional agreement to convert, and what the conversion formula is. Put that in the first contract or it will not happen in the second.

The sequence I believe in is unglamorous. Build the measurement. Run it long enough to have a baseline that is observed rather than assumed. Then price against it, with attribution written down.

Someone who prices before the baseline exists is taking more risk than I am, not less, and I would rather say that than imply otherwise. The reason I still would not do it is that they are pricing an estimate, and when the estimate is wrong the argument lands in the first period where the number disappoints.

I would rather have the harder conversation with you now than later