STP - Straight through processing
What is straight through processing: STP explained
Straight through processing, abbreviated to STP, is the handling of a claim from notification to payment entirely by automated rules, with no manual intervention by a handler at any point. A claim either qualifies at intake and flows through to settlement on its own, or it falls out of the automated path into a manual queue.
For insurers
See the damage before you send anyone
Venta Capture, a product of VentaVid, sends the policyholder a link. They film the damage on their own phone, guided step by step, and the evidence lands with the claim.
The term comes from payments and securities settlement. In claims it is binary per file: there is no such thing as a mostly automated claim.
What does straight through processing mean in claims?
Straight through means no human touch, not no human oversight. Governance, sampling, audit and appeal all still exist around an STP path, and any of them can pull a file back.
Two neighbouring terms get used loosely and mean different things:
- Touchless: usually the same thing as STP, sometimes used more loosely to mean no customer contact rather than no handler contact.
- Automated decisioning: a rule or model makes one decision inside a claim that a person still owns. That is automation, not STP.
- Fallout or exception handling: what happens to the claims that fail eligibility. Designing this well matters more than the STP rate itself.
How does straight through processing work?
An STP path is a gate with a fixed set of conditions, checked in order.
- Eligibility: claim type, value band, policy status, cover in force, no excluded peril, claimant is the policyholder, no open related claim.
- Data completeness: every field the decision needs is present and validated, not merely supplied.
- Evidence sufficiency: the images, documents or readings the rules require exist, are legible, and cover the required angles or items.
- Risk screening: automated checks for duplicates, recent policy changes, claim history patterns and known fraud indicators. These route a file to a human. They do not accuse anyone.
- Deterministic settlement: the payable amount is computable from the data, typically a schedule, a network repair rate or a fixed benefit.
- Payment and closure: verified payee details, mandate in place, file closed with a full audit trail of which rules fired.
Fail any gate and the claim drops out. The design question is not how to automate every claim, it is which claims are genuinely deterministic and how quickly the rest get to a person.
Straight through processing example: two claims, one difference
Two windscreen claims arrive the same morning. Both are in cover, both are under the value threshold, both name the policyholder as the driver. The first includes clear images of the damaged glass, the vehicle identification number and the number plate in frame, so the rules validate the vehicle, price the job at the network rate, book the fitting slot and close the file in eleven minutes with no handler involved.
The second has two dark photographs taken through the interior of the car, with no plate visible. It fails evidence sufficiency, drops into the manual queue, and settles nine days later after two rounds of chasing better pictures. Identical claim, identical entitlement. The only difference was what arrived in the first submission.
How the STP rate is measured, and gamed
The rate is straight through claims divided by total claims, but the denominator is where the arguments live.
- Scope selection: quoting STP for one product line, or only for claims that entered through the digital channel, is the most common inflation.
- Counting partial automation: a claim a handler reviewed for ninety seconds is not straight through.
- Excluding fallout: dropping ineligible claim types out of the denominator rather than reporting them.
- Ignoring reopens: an automated settlement that gets reopened belongs out of the numerator.
Be sceptical of external benchmarks here. The STP rates in general circulation run from under 10 percent to over 80 percent, and almost all of them trace back to vendor marketing rather than to a published study with a stated sample and collection period. Your own rate, measured on a consistent denominator over time, is the only usable figure.
Why visual evidence is where STP usually breaks
Most claims that fail an STP gate do not fail on policy data. Cover, premium status and claim history sit in systems that answer instantly. They fail on the part that comes from the claimant, and in property, motor and equipment claims that part is almost always visual.
A photograph is either usable or it is not, and the claimant finds out which weeks later. Too dark, wrong angle, damage cropped out, serial number unreadable, no wide shot to establish context. Each of those is a fallout, and a fallout costs the full manual handling expense plus the elapsed days of the retake loop.
Which leads to the uncomfortable part of any automation business case. A claim only flows straight through when the evidence arrives complete on the first attempt, so the ceiling on your STP rate is set at intake, not in the rules engine. Invest heavily in decisioning while leaving evidence collection to whatever the customer thinks to send, and you get an excellent engine that most claims never reach.
That is why FNOL automation work usually starts with the submission itself. Guided capture exists to make first time completeness the normal outcome rather than the lucky one, guided photo capture covers how the prompting works in practice, and the retake request is the cost you are trying to design out.
What straight through processing is confused with
It is not automatic approval. A well built STP path declines claims as readily as it pays them, and a path that only ever pays is a control weakness. It is not artificial intelligence either: most production STP in claims runs on deterministic rules precisely because the decision has to be explainable to a regulator. And it is not a cycle time programme. STP removes handling from a minority of simple claims. Cycle time is mostly decided by the complex majority that will always need a person.
