Do you store data and serve it from a cache?

Some sources are copies we keep in step with the register, and every field says so: live_from_registry or cached_from_registry. Verification only uses authoritative sources. Prefill may use a faster, non-authoritative copy, and the authoritative flag on each datapoint tells you when.

Can you guarantee an SLA per register?

No one can promise uptime for a government register. What we can do is show it: the status page tracks each register's availability separately from ours.

What if a register is down during Prefill?

Continue with the fields you have, flag the file and verify later. The status page tells you when it is back.

How do we know when a country's coverage changes?

Check the coverage matrix. It is built from the same public catalog your code can read at GET /v2/catalog, which also marks every datapoint and document fetched by hand.

What about your own outages?

They are on the status page, with uptime and incident history, separate from register incidents.

For product and engineering teams

What happens when a register is slow, down, or out of date

Topograph pulls company data from official registers. Each field shows its source and whether it came live from the register or from a copy, and each datapoint says whether it is authoritative.

Status page Reliability guide API reference Coverage matrix

Two modes

Prefill and Verification don't behave the same

Same API, different promise. Prefill is for speed; Verification is for evidence. Use Prefill to fill a form. Use Verification for anything that goes into a case file.

Aspect	Prefill	Verification
Use it for	Filling a signup form	Checks that gate your decision
Source	The fastest source for each datapoint: the register or a copy. Each field says which	Authoritative sources only: the official register, or a copy kept in step with it
Time limit	10 seconds per datapoint. Datapoints that finish in time come back as normal. Slower ones return `onboarding_timeout`	No fixed deadline. Seconds for most registers, minutes for slow ones, business days for records fetched by hand
If the register is down	Fails straight away, so your customer isn't kept waiting	We retry for up to an hour, then it fails
Speed	Typically under 5 seconds per datapoint	Depends on the register

Failures

When something goes wrong

Every failure says whose side it is on: the register, us, coverage, or your request.

Billing: failed datapoints and documents are never charged.

What happened	Whose side	What you get	What to do
The register doesn't publish this datapoint in this country	Register, by law	`datapoint_not_supported`, a final answer, not an empty field	Nothing. Get it from the company, or skip it in that country
The register holds no record for this company	Register	An empty list, and the datapoint succeeds	Nothing. We looked; there is nothing there
The company isn't in the register	Register	`resource_not_found`	Check the identifier or run a search
The register is slow	Register	`in_progress` on the datapoint, then a webhook when it lands	Wait for the webhook. It is not an error
The register is down	Register	`source_unavailable`. Prefill: usually final. Verification: after up to an hour of retries	Prefill: move on and verify later. Verification: resend after the incident
The register changed its format	Register	`source_unavailable` or `processing_failed` while we adapt	Check the status page. We aim to fix it within 24 business hours
No fast source for this datapoint here	Coverage	`fast_source_unavailable`	Use Verification
No authoritative source in this country	Coverage	`authoritative_source_unavailable`	Use Prefill and treat the result as non-authoritative
Fetched by hand (beneficial owners in China, Croatia, Cyprus and 4 more)	Expected	`in_progress`, typically for 1 to 3 business days	Nothing. Tell your analyst it is normal
Rate limit	Yours	`429` with a `Retry-After` header	Slow down, and resend once the limit resets
Something broke on our side	Ours	`service_unavailable` on the datapoint, or `500` on the request	Resend after `retryAfterSeconds`. If it persists, contact support

Method

How we measure

What every response tells you, what we publish for every country, and a live status page.

Response time

How long the register takes to answer each datapoint, measured on successful production requests and refreshed regularly. Shown per datapoint and mode on each country pricing page. Records fetched by hand show a delivery window in business days.

Authority

Every datapoint carries an authoritative flag: true when it comes from an authoritative official source, false when Prefill used a faster, non-authoritative one.

Freshness

Every field says how it was obtained: live_from_registry, or cached_from_registry for a copy we keep. Copies follow the register's own publication cycle, from daily to monthly.

Status page. Our availability and each register's, shown separately, so you know whose incident it is before you open a ticket.

status.topograph.co

FAQ

Questions teams ask

Run your hardest countries before you commit

Send us the countries that worry you. We will run them through both modes and return production timings, failures and results.

Read the reliability guide