Eleven markets · one integration

Eleven markets. Forty-one identity sources. One endpoint.

Idverra verifies identity documents, runs liveness and screens against sanctions and PEP lists across African markets — behind a single API that does not change shape when you enter a new country.

Get API keys See the pipeline

Sandbox access is immediate and unmetered. Production requires a signed data processing agreement and takes about a day.

POST /v2/verifications 200 OK

// response
{
  "id": "ver_8Fq2xK4nTb",
  "country": "NG",
  "decision": "approved",
  "sources": [
    { "type": "bvn",      "match": 0.991 },
    { "type": "nin",      "match": 0.984 },
    { "type": "liveness", "match": 0.973 }
  ],
  "screening": {
    "sanctions": false,
    "pep":       false,
    "adverse_media": false
  },
  "audit_ref": "aud_20260912_004417"
}

p50 1.9s p95 4.2s uptime 30d 99.97%

Illustrative response · sandbox payload

Why this is hard

Eleven countries is not one integration repeated eleven times.

Nigeria verifies against BVN and NIN, through two agencies with different uptime and different tolerance for load. Ghana has the Ghana Card. Kenya has Huduma and the older national ID register. South Africa has Home Affairs. None of them agree on field names, image standards, error semantics or what a timeout means.

A fintech expanding across three markets does not write one verification service. It writes three, then maintains them separately while each upstream source degrades on its own schedule — usually at month end, when volume peaks.

Idverra absorbs that. One request shape, one response shape, one decision model. When a government source is slow we fail over to a secondary and tell you which one answered, rather than returning an error your support team has to interpret.

What you would otherwise integrate
Market	Primary sources	Fallback
NigeriaNG	BVN · NIN · Passport · DL	Voter register
GhanaGH	Ghana Card · Passport	SSNIT
KenyaKE	National ID · Huduma · KRA PIN	Passport
South AfricaZA	Home Affairs ID · Passport	Credit bureau
UgandaUG	National ID · Passport	—
SenegalSN	CNI · Passport	ECOWAS card

The pipeline

What happens in the two seconds after you call it.

Every stage is logged with its own timing, its own result and its own source. When a regulator asks why a customer was approved, the answer is a record rather than a reconstruction.

Capture

Document image and selfie arrive from your app or our drop-in SDK.

~ 40 ms

Extract

OCR and MRZ parsing pull the fields, with a tamper check on the document itself.

~ 310 ms

Match

Extracted identity is checked against the government source for that market.

~ 900 ms

Liveness

Passive liveness plus face match against the document portrait and source photo.

~ 420 ms

Screen

Sanctions, PEP and adverse media, with fuzzy matching tuned for local name forms.

~ 180 ms

Decide & log

A decision against your thresholds, written to an immutable audit record.

~ 60 ms

The API

Three calls, and none of them change when you add a country.

Request cURL

curl -X POST https://api.idverra.com/v2/verifications \
  -H "Authorization: Bearer $IDV_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "country": "NG",
    "sources": ["bvn", "nin"],
    "liveness": true,
    "screening": ["sanctions", "pep"],
    "person": {
      "first_name": "Adaeze",
      "last_name": "Okonkwo",
      "date_of_birth": "1994-03-17",
      "identifier": "22XXXXXXXXX"
    },
    "webhook": "https://yourapp.com/hooks/idv"
  }'

Endpoints
Method	Path	Purpose
POST	/v2/verifications	Run a verification
GET	/v2/verifications/:id	Fetch result and stage timings
POST	/v2/screenings	Screen without verifying
GET	/v2/audit/:ref	Retrieve the audit record
POST	/v2/reviews/:id/decide	Resolve a manual review
GET	/v2/coverage	Live source availability

SDKs Node · Python · Go · PHP · Kotlin

Coverage

Sources, latency and liveness, per market.

Median response time measured over the trailing thirty days, including upstream time. Where a government source is unavailable, the fallback column is what answers instead.

Market	Sources	Median	Liveness	Status
Nigeria	6	1.8s	Passive + active	Operational
Ghana	4	2.1s	Passive + active	Operational
Kenya	5	2.4s	Passive + active	Operational
South Africa	4	2.9s	Passive	Degraded
Uganda	3	3.1s	Passive	Operational
Tanzania	3	3.4s	Passive	Operational
Senegal	3	2.7s	Passive	Operational
Egypt	3	3.8s	Passive	Beta

Figures shown are illustrative of the format published on the live status page, which reports real measured values per market and per source.

Screening

Fuzzy matching that understands how names actually work here.

Generic screening engines flag every Ibrahim Mohammed against every sanctioned Ibrahim Mohamed, then leave a two-person compliance team to clear hundreds of hits a week. The cost of a bad matcher is not a missed criminal — it is an alert queue nobody can keep up with.

Idverra scores name similarity against transliteration variants, patronymic ordering, and the spelling conventions of the markets it operates in, then shows you why a hit scored what it did. Every alert carries the list, the entry, the score and the reason.

Cleared alerts stay attached to the customer record, so the next screening run does not surface a decision your team already made.

Screening result · ver_8Fq2xK4nTb
List	Closest entry	Score	Result
OFAC SDN	No entry above threshold	0.41	Clear
UN Consolidated	No entry above threshold	0.38	Clear
EU Sanctions	No entry above threshold	0.35	Clear
PEP · domestic	Okonkwo, A. — state official	0.72	Review
Adverse media	No entry above threshold	0.29	Clear

Lists refreshed every 4 hours Retained 7 years

Audit trail

The thing you will actually be examined on.

Nobody has ever failed an inspection because their verification was slow. They fail because they cannot evidence, eighteen months later, what was checked and who decided.

2026-09-12 04:44:17.201createdverification ver_8Fq2xK4nTb · country=NG · key=live_ba91

2026-09-12 04:44:17.244capturedocument=nin_slip · selfie=present · tamper_score=0.02

2026-09-12 04:44:17.558extractmrz=parsed · fields=11/11 · confidence=0.988

2026-09-12 04:44:18.462matchsource=bvn · upstream=primary · match=0.991

2026-09-12 04:44:18.981matchsource=nin · upstream=primary · match=0.984

2026-09-12 04:44:19.402livenesspassive=pass · face_match=0.973

2026-09-12 04:44:19.585screensanctions=clear · pep=review(0.72) · media=clear

2026-09-12 04:44:19.646decidethreshold=0.95 · outcome=approved_with_review · ref=aud_20260912_004417

2026-09-12 09:12:03.118reviewpep cleared by user_44 (compliance) · note="state official, different DOB"

Immutable and addressable

Every record is write-once and retrievable by reference through the API for seven years, including the exact list versions used at the time of screening.

Reviewer attribution

Manual decisions carry the reviewer, the timestamp, and the note they wrote — not just a status change with no author.

Data residency

Personal data for each market is stored in-region where local law requires it, and never leaves the region for processing.

Export on demand

Examiner-ready exports for any date range, as structured data or as a signed PDF bundle, generated without engineering involvement.

Thresholds

The only decision that really costs you money.

Set the match threshold too low and bad actors get through. Set it too high and your team drowns in manual review while good customers abandon sign-up. Move it here and watch both sides move.

Sent to manual review each month12,655

Reviewer hours, at 4 minutes each844 hrs

Full-time reviewers required5.3

Verification spend at your tier$2,800

Distribution modelled on a typical consumer wallet onboarding mix. Your real curve depends on document quality, source coverage and fraud pressure in your segment — we fit it against your first ten thousand verifications during onboarding, at no charge.

Pricing

Per verification, falling with volume. Screening included.

Tier	Monthly volume	Per verification	Screening	Support
Sandbox	Unlimited	Free	Included	Docs & community
Starter	to 10,000	$0.09	Included	Email, 1 business day
Growth	10,001 – 100,000	$0.07	Included	Shared Slack channel
Scale	100,001 – 500,000	$0.05	Included	Named engineer, 4 hr SLA
Enterprise	500,000+	Negotiated	Included	SLA, dedicated region

No charge for failures

A verification that cannot complete because an upstream government source is unavailable is not billed, and does not count toward your tier.

Re-runs are free for 24 hours

A customer retrying a failed capture inside a day is the same verification, billed once — not three attempts billed three times.

Screening is not a separate line

Sanctions, PEP and adverse media are part of the per-verification price, including ongoing monitoring of customers you have already onboarded.

No annual commitment

Tiers apply automatically at the volume you actually do. There is no minimum spend and no contract term on anything below Enterprise.

Questions

Asked by compliance, and by the engineer who has to build it.

What happens when a government source goes down?

We fail over to the secondary source for that market and mark which source answered in the response, so your decision logic can treat a fallback match differently if you want it to. If every source for a market is unavailable, the verification returns a pending status with a retry window rather than a hard error — and it is not billed.

Will a regulator accept your audit record as evidence?

The audit record captures what was checked, against which source, at what time, with which list version, and who made any manual decision. That is the evidentiary standard examiners work to. We provide exports in the formats examiners in each market have asked for, but the regulated entity remains you — we are a processor, and the obligation to satisfy your regulator stays with your licence.

Where is personal data stored?

In-region, per market, where local law requires it — Nigerian data stays in Nigeria, Kenyan data in Kenya. Processing happens in the same region as storage. Document images can be set to be discarded after a configurable retention period while the verification record and its hashes are kept for the full seven years.

How do you handle false positives on name screening?

By scoring rather than flagging, and by showing the reason for the score. You set the threshold at which a hit becomes a review item, and cleared alerts are remembered against that customer so they do not resurface on the next run. During onboarding we tune the matcher against your first ten thousand verifications, which is usually where a generic engine's alert volume drops by an order of magnitude.

Can we run it on our own infrastructure?

There is a dedicated deployment on Enterprise, where the processing region and the data stores are yours alone. A fully on-premise install is not offered, because the screening lists and the upstream source connections require a maintained path out to systems we operate.

How long does integration actually take?

Sandbox to a working verification is typically under an hour with one of the SDKs. Production readiness is usually a week of real work — mostly deciding your thresholds, wiring the review queue into your own back office, and agreeing retention with your DPO. Adding a second market after that is a configuration change, not a project.