The Onfido alternative that owns the risk narrative, not just the pass/fail.
Onfido — now Entrust IDV — is a serious identity verification vendor. So are Sumsub, Jumio and Veriff. All of them return a pass, a fail or a manual-review flag. In 2026, the differentiator is not which vendor you use for the check. It is whether the risk narrative your compliance team signs off is firm-owned.
1 — What Onfido (Entrust IDV) actually is
Onfido is an identity verification (IDV) and KYC platform: document verification, biometric check, database checks and increasingly integrated fraud signal. Following the Entrust acquisition it is marketed as Entrust IDV and positioned alongside PKI and credential lifecycle products, particularly for enterprise and government identity use cases.
For regulated financial firms, Onfido typically slots in at onboarding: capture ID, run document + selfie checks, return a decision, and hand the case to a human analyst if the check is ambiguous. The main competitive set is Sumsub, Jumio and Veriff; recent 2026 vendor guides describe headline per-check pricing across the category in a broad $0.80 – $3.80 range depending on tier, volume and geography, with Onfido typically at the higher end of that range for enterprise deployments.
The important thing to be honest about: all four vendors are technically credible. Choosing between them on capability alone rarely produces a durable advantage. Where firms produce durable advantage is in what happens above the check — the risk narrative, the case triage, the fraud signal integration and the audit trail.
2 — What Onfido (Entrust IDV) costs you
Onfido does not publish a public per-check price. Enterprise deals are negotiated per firm and priced per verification, per user or per volume tier, with materially different economics depending on region, document set and biometric options. Public 2026 vendor guides put category-wide per-check pricing in the $0.80 – $3.80 range with Onfido/Entrust IDV typically toward the higher end.
For a mid-sized regulated firm the loaded cost of an IDV vendor typically looks like this (indicative).
| Line item | Order-of-magnitude cost | Note |
|---|---|---|
| Onfido per-verification (blended enterprise) | Not publicly disclosed; category range $0.80 – $3.80 / check | Higher end for full biometric + document + database. |
| Ongoing monitoring / re-verification | Not publicly disclosed | Per-check on ongoing checks. |
| Analyst / L1 manual-review team | £200k – £1.5m+ / yr | The dominant cost line at scale; drops fast with better triage. |
| Fraud signal add-ons (device intel, velocity) | £50k – £300k / yr | Sift, Sardine, in-house device intelligence. |
| Case management + onboarding CRM wrap | £80k – £250k / yr | In-house or via case-management vendor. |
| Integration + change management | £100k – £400k one-off | Especially for regulated migrations. |
The per-check line is not the dominant cost at any real scale — the human review team is. Every ambiguous decision, every manual re-verification and every fraud escalation is a human analyst. That is the actual cost of KYC in 2026, and it is exactly what a firm-owned intelligence layer changes.
None of that spend is wasted if it produces a defensible onboarding posture. What is wasted is spending it and still owning nothing above the vendor's pass/fail.
3 — What Onfido (Entrust IDV) can't do for your firm
Onfido is a good IDV vendor. Five things it structurally does not do for a modern compliance function.
It returns a decision; it does not produce a firm risk narrative
Onfido returns "pass", "fail" or "consider". It does not know your firm's product-specific risk appetite, your regional concentration limits or the two facts that historically decide edge cases. That narrative has to live in a firm-owned layer.
Manual review is a queue, not an intelligence surface
The console is fine as a queue. It is not an AI-native cockpit that reads the document set, the selfie result, the device signal and the historical decisions on similar customers, and drafts the analyst's initial narrative.
Fraud signals live in silos across vendors
Onfido's fraud signal is one input. Device intelligence (Sift, Sardine), velocity, PEP hits and internal fraud history are separate. Correlating across all of them per customer is a firm-owned layer, not a vendor product.
The vendor's model is generic; your fraud pattern is not
Vendor models are trained across the full customer base of thousands of firms. Your firm-specific fraud pattern — for your product, your channel, your region — is not in that training data at meaningful weight. A firm-specific model trained on your own decisions is.
Every improvement is on the vendor's roadmap
New capability arrives when Entrust ships it. That is fine for the underlying document / biometric check. It is the wrong dependency for the risk-decision layer that determines your onboarding conversion, fraud loss and regulator posture.
4 — Why 'your own version' is viable in 2026
The right frame is not "replace Onfido." Onfido (or Sumsub, or Jumio, or Veriff) stays as the underlying check. The right frame is: build a firm-owned risk decision layer that treats every IDV vendor as an interchangeable data source and applies the firm's own model above the check.
Every verification result, every fraud signal, every historical onboarding decision and every downstream customer outcome (chargeback, fraud loss, closure) streams into a firm-owned data spine. On top, a firm-specific model triages ambiguous decisions, drafts the analyst narrative, correlates fraud signals across vendors and — critically — is trained on the firm's own outcomes.
The analyst opens a case in a firm-owned cockpit with narrative, evidence and recommended action already drafted. Onboarding conversion improves. Manual-review cost drops. The IDV vendors become interchangeable data sources; the risk brain is firm-owned.
5 — Reference architecture
Firm-owned risk decision layer above Onfido and other IDV vendors
Onfido continues to run the underlying check. If the firm runs a secondary vendor (or wants optionality), that vendor plugs into the same spine.
Every verification, fraud signal, historical decision and downstream outcome (chargeback, fraud loss, account closure) streams into a firm-owned data spine. The intelligence layer applies the firm's own risk model, correlates signals across vendors, drafts the analyst narrative and ranks ambiguous cases.
The analyst works in a firm-owned onboarding cockpit with the vendor console still available. Every human decision feeds back into training data. The onboarding conversion, fraud loss and regulator posture become firm-owned outputs, not vendor-defined ones.
6 — Build vs. rent: 3-year TCO
A three-year illustrative comparison for a mid-sized fintech / regulated firm on Onfido.
| Dimension | Rent (Onfido) | Build (KJ Capital) |
|---|---|---|
| Onfido per-verification (enterprise) | Not publicly disclosed; category $0.80 – $3.80 / check | Unchanged |
| Manual review / L1 team | £200k – £1.5m+ / yr | Materially reduced per case; team retrained onto higher-value work |
| Fraud signal add-ons | £50k – £300k / yr | Unchanged as data sources; correlation moves into firm-owned layer |
| Case-management / wrapper tooling | £80k – £250k / yr | Rationalised into firm-owned cockpit |
| Intelligence layer (LLM infra + evals) | N/A | £200k – £400k / yr |
| Initial build (Yr 1 only) | N/A | £300k – £600k one-off KJ Capital build |
| 3-year total (illustrative) | Rising ops cost with volume growth | Case-cost breaks the escalator; owned proprietary layer |
The break-even is typically inside 12–18 months on realistic manual-review reduction and onboarding-conversion improvement. The strategic value is bigger: the firm owns the model that decides who gets onboarded.
7 — The three honest risks of building your own
"Our regulator expects to see a recognised IDV vendor."
How we solve it —They will continue to. Onfido remains the underlying check. The intelligence layer strengthens the audit trail — every decision, every AI proposal and every human sign-off is logged with provenance.
"We can't have AI making KYC decisions."
How we solve it —It doesn't. Every triage, every narrative and every decision is a proposal with explicit human sign-off on any non-clear case. The system exists to make analysts faster and more accurate — not to replace their judgment on regulated decisions.
"We don't have an in-house AI team."
How we solve it —You don't need one on day one. Build ships the first system in 6–12 weeks. Operator runs it in production while your team takes ownership.
8 — The CTA ladder
Onfido is a good IDV vendor. It is not, and does not need to be, your firm's risk-decision brain. That brain has to be firm-owned to compound and to give your onboarding a durable advantage over competitors on the same vendor stack.
The right sequence is Readiness → Audit → Blueprint. Two weeks with our team and you have a scoped plan you can put in front of your MLRO and your product lead.
Free Readiness Score
5 minutes. Honest score of whether your firm is in a position to build.
AI Diagnostic — £15k / 2 weeks
Founder-led. Map the specific systems where you'd get compounding return.
Blueprint conversation
30-minute call with Kasim to scope the Build.
Five questions we get asked most.
Do we need to leave Onfido?+
No. Onfido stays as the underlying identity check. The intelligence layer is additive and works above any IDV vendor.
How much does Onfido cost per check?+
Onfido does not publish per-check pricing. Public 2026 vendor guides put the category (Sumsub, Onfido, Jumio, Veriff) in a $0.80 – $3.80 per-check range, with Onfido / Entrust IDV typically at the higher end for enterprise.
How does this compare to Sumsub, Jumio or Veriff?+
Identical pattern. The intelligence layer is vendor-agnostic and works above any of them. Firms often keep two IDV vendors and let the intelligence layer route intelligently between them.
How much can this reduce manual review?+
Realistic firm-specific model results are typically a 30–60% reduction in cases requiring analyst time, with better onboarding conversion on genuine customers. Real numbers depend heavily on firm data quality and starting posture.
How long until the first system is live?+
The Diagnostic is 2 weeks. The first Build is 6–12 weeks and typically ships firm-specific case triage + analyst narrative on a defined customer segment, with Operator running it in production.