Captive Loss-Run Clinic
Start with one difficult loss run.
A short working session on a file you already have. We look at how it is structured, where it would fight an ingestion and reserving workflow, and what standardizing it would actually involve.
What the clinic is
Loss runs arrive in whatever shape the TPA produces. Column names differ between carriers and between years from the same carrier. Valuation dates are sometimes a column, sometimes only in a file name, and sometimes nowhere at all. Claim identifiers change format mid-book. None of this is unusual, and all of it has to be resolved before a triangle means anything.
The clinic is a review of that resolution work on one real example. You bring a synthetic, redacted, or representative file; we go through it together and show where the obstacles are.
The clinic is a product-discovery and workflow-review session. It is not an actuarial opinion, insurance recommendation, regulatory filing, or audit.
What a review can usually identify
These are the categories that come up most often. Which of them apply depends entirely on the file.
- Structure. Whether the claim table can be located reliably, and whether headers, totals rows, and banner rows can be told apart from claim rows.
- Field meaning. Which columns carry claim identity, date of loss, paid and incurred amounts, and the valuation date — and which are ambiguous enough to need confirming rather than guessing.
- Valuation chronology. Whether the file supports a development history at all, or represents a single valuation that cannot show development on its own.
- Consistency across files. Whether the same claim can be followed across several valuations, and what breaks that continuity.
- Data quality. Amounts that will not parse, duplicated claim rows within one valuation, and cumulative figures that move in unexpected directions.
A decrease in cumulative paid is a good example of why review matters: it is often a recovery and entirely legitimate, so it should be surfaced for a person to judge rather than quietly smoothed away.
What the workflow does with it
If the file is a fit for the pilot, the same steps run in the customer application rather than by hand:
- ingest the loss-run sources;
- confirm what each column means, rather than inferring it;
- validate and normalize, surfacing what could not be read instead of dropping it;
- analyze the development history the data actually supports;
- review before anything is published;
- keep an auditable trail from source file to result.
Outputs stay subject to review. The workflow is built to make an actuary's judgement faster to apply and easier to evidence — not to replace it.
What to bring
One loss run that has given you trouble. Synthetic, redacted, or representative data is fine and is what we prefer for a first conversation.
Please do not send claim data by email. If a file is useful to the discussion we will agree a controlled way to share it first.
Apply to the Clinic
Send us one loss run and we will tell you what we can read from it — what is there, what would need confirming, and what would get in the way of an analysis. It is a short, bounded look at your own data, and it does not produce reserves or any figure you could file.
We read every application by hand, so it takes a little time. If we go ahead, you will get a sign-in link at the work address you apply with.
No application origin is configured yet, so the button above may not appear. Until it does, email hello@priorexperience.io with your name, company, work email and what you would like to find out.