Identity infrastructure · YC W22

Turning six dashboards back into one platform.

Dojah’s identity APIs worked. The product around them had grown six separate dashboards, and users got lost before their first verification. I found the one problem hiding behind six complaints, and rebuilt the platform on a single information architecture.

Role
Product design, IA strategy
Timeline
2021–24
Product
Identity verification, YC W22
Dojah’s overview after the rebuild, with six identity products grouped by the job and a short list of tasks to finish before going live
The overview after the rebuild. Six identity products grouped by the job, with a short list surfacing the compliance and go-live tasks that used to get skipped.
Amara Nwosuvs_9f2a41
Individual KYC · Nigeria
In review
Document check100%

Nigerian passport · MRZ valid, not expired

Liveness99%

Active liveness · three gestures captured

Face match98%

Selfie matches the document photo

AML screening1 hit

One possible PEP match on the national list

Decision

Held until the AML match is cleared

Identity cleared. One AML hit holds the decision for a person.
−50%
onboarding time
−60%
integration time
+91%
organic traffic
−40%
customer lookup time

Context

Dojah is Africa’s first end-to-end identity verification platform. I led the redesign of its core web application between 2021 and 2024, working with engineers, PMs, and data scientists to restructure the dashboard, surface compliance, and cut integration time.

The product served two audiences in one screen. Developers came to integrate identity APIs and cared about keys, environments, and docs. Compliance and operations users came to verify customers, run lookups, and manage AML risk, and cared about workflows, not endpoints. The dashboard treated both the same, and both struggled.

The challenge

Every product had grown its own surface. KYC, KYB, AML, document checks, widgets, and lookups each had a place, but none of them shared a shell. New users landed and could not tell where their job began.

Developers could not find their API keys. Operations users skipped critical compliance tasks because they were buried under product screens. Everyone confused Sandbox and Live, hit the wrong endpoint, then opened a support ticket.

In the support queue

I read the tickets and expected six problems. Where are my keys. Why did my call fail. Where is the compliance step. Am I in Sandbox or Live. How do I look up a customer. Why is this task blocking my launch.

They were not six problems. They were one problem asked six ways.

The reframe

It was never six problems. It was one architecture problem.

The information architecture was built for the org chart, not the job. Engineering features lived under Apps. Compliance lived under Settings. Sandbox and Live were a single dropdown most users never saw. Once I saw the IA as the root cause, every fix lined up against the same north star: every feature should live where the job actually starts.

The rebuild

Three moves, each aimed at one way the old architecture got in the way.

Move 01

From the org chart to the job

The old dashboard was grouped the way the platform was built. Engineering features sat under Apps, compliance sat under Settings, and Sandbox and Live were one dropdown most people never noticed. I regrouped every surface around the job a person came to do, so keys, verifications, and AML each live where the work starts. A Quick Actions list surfaces the compliance and go-live tasks that used to get skipped. A few advanced features moved one click deeper. In return, the tickets asking where is X went quiet.

Before and after: six scattered product dashboards next to one grouped information architecture
The shift, side by side. Six product surfaces with no shared shell, versus one architecture grouped by the job, with keys and compliance surfaced and one clear environment switch.

Move 02

From a verification you could not follow to one you can

On the operations side, a single check was a black box. I rebuilt it as a pipeline: document, liveness, face match, AML screen, decision. Each step carries its own status, so a reviewer sees exactly where a case sits and what is holding it. Clean identity and liveness no longer wait behind an AML hit. The one thing that needs a human is the only thing a human has to touch.

A single identity verification shown as a pipeline of document, liveness, face match and AML steps with status states
A single verification as a pipeline. Document, liveness, and face match cleared. One AML hit holds the decision for a human, and nothing else waits behind it.

Move 03

From a cold start to a first verification in minutes

Developers used to hunt for their keys and guess which environment they were in. The new get-started surfaces the sandbox key, a copyable first call, and one persistent switch between Sandbox and Live. Integration time fell by more than half. The Sandbox and Live mix-up, once the single highest-volume support ticket, nearly disappeared.

A developer get-started screen with a setup checklist, sandbox API key, a first API call, and a Sandbox and Live switch
The developer’s first ten minutes. Sandbox key surfaced, a copyable first call, a persistent Sandbox and Live switch, and a test verification that returns before you leave the page.

What changed

Onboarding and compliance time dropped by more than half. Developer integration time fell 60%. Easy Lookup, a six-step flow I cut to two, took 40% off the most-used task on the dashboard. Support tickets across the IA and API-visibility categories dropped sharply.

“Nobody files a ticket about information architecture. They file six tickets that all have the same cause.”

Adoption was smooth because the change was legible: a one-time welcome screen highlighted the new layout, paired with a targeted email from marketing. The same rebrand launch lifted organic traffic 91%.

Reflection

Dojah taught me that navigation is a polite word for a broken information architecture. Users do not complain about IA. They complain about features being hard to find, tasks being skipped, environments being confused. Six different complaints, one root cause.

The hardest part was not the design. It was convincing engineering and product leadership that a redesign of the menu was a redesign of the product’s mental model, and worth four months of focus.