ID Scan & Age Verification

I led the strategy and design of a range of features that got us from a 36% industry average to over 90% success rate, with great feedback from couriers, clients and end customers.

Lead Product Designer, 2025

Summary

By early 2025, Stuart’s age-verification rate was at 36%. That’s typical for digital logistics, but far below physical retail, and it was putting key client relationships at risk.
I led the design work to fix this, working across Product, Engineering, Operations and Data Protection.
We took a two-phase approach: first we fixed the clarity and support around existing flows, then we introduced a privacy-first ID scan. Verification went from 36% to over 90%, nearly 3× the industry average, months ahead of target.
The bigger shift was organisational. Stuart had spent years treating this as a fraud problem. This work showed it was a clarity problem, and that changed how other compliance work was approached.

What I did

I was the design lead across the whole effort. That meant:

reframing the problem away from courier fraud and towards clarity and experience
helping teams work through concerns across risk, privacy, operations and engineering
running research: doing deliveries myself, reviewing audits, surveys, interviews, a diary study
finding and testing a privacy-safe, on-device OCR approach to ID scanning
designing the end-to-end flow across the app, support and operations

Context

Age-restricted deliveries (alcohol, knives) legally require ID checks. Digital logistics typically manages 30–40% verification rates. Stuart was at 36%.

This was causing problems:

Enterprise grocery clients were escalating
A significant percentage of revenue was at risk
Couriers were frustrated by awkward customer interactions

A cross-team working group was put together to fix it. I joined as the design lead.

Starting with what couriers actually experience

Internally, people assumed couriers were committing fraud. I wasn’t sure about that.

On a trip to the London office, I went out and did deliveries with age-restricted items.

The moments where couriers need to ask for ID are difficult. Customers can be confused, irritated, or angry. Couriers had very little support to handle that well.

One customer became angry when I followed the product’s instructions. The rules weren’t working in practice.

This helped me make the argument that we needed clarity and better support for couriers, not more suspicion or stricter enforcement. The goal wasn’t to tighten rules — it was to make the right thing easier to do.

That was a big shift internally, and unproven in digital logistics, but it became the idea we built everything around.

What the audit notes showed

I reviewed past audit comments in detail. Couriers were often doing everything right — being polite, asking questions — except checking the physical ID:

He then asked for my date of birth, put it into his phone, and said have a nice day.

He said hello and asked my DOB. He then put it into his phone and gave me the parcel after taking a photo. He then left thanking me.

The intent was there. The flow just didn’t support the final step.

This helped move the conversation away from fraud and towards the actual obstacles people were facing.

Working out what to do

There was a lot of urgency and people were jumping to solutions before we had shared understanding of the constraints — fraud risk, privacy, technical feasibility, operational impact.

I mapped out all the possible directions into a comparable set of options with clear criteria (impact, effort, privacy risk), and facilitated a discussion about what to do now versus later.

That gave us a way to talk about trade-offs and make decisions.

A two-phase strategy

Based on the research and constraints, I proposed two phases:

Improve clarity and support for couriers across the app, operations and support. Make it practically and socially easier to ask for ID at the right moments.
Explore whether privacy-safe ID scanning was viable — technically, emotionally, and from a data-protection standpoint.

We aligned on this and moved straight into execution.

Phase 1: from 36% to 60%+

I redesigned the flows around a simple principle used in UK physical retail: if someone looks under 25, check their ID.

The main changes:

Dedicated, clear ID-check screens
Couriers had to actively confirm whether they checked ID
Consistent visual and written guidance
Alignment across Ops and Support so messages weren’t contradictory
Removing the requirement for ID checks in obviously unnecessary cases

This made asking for ID less awkward, easier to explain, and meant compliance happened naturally rather than being forced.

Internal teams worried this would increase fraud. I pushed to test it anyway. Compliance jumped to 60%+ within weeks.

Phase 2: ID scanning to 90%+

As Phase 1 data came in, we looked at ID scanning again.

People across the team were uncomfortable with it. Scanning IDs felt invasive. Customers wouldn’t trust it.

While I was exploring possibilities with an LLM, it suggested something we hadn’t considered: on-device OCR that reads only the date of birth. No images captured, stored, or sent.

I checked with a mobile engineer and it was feasible. I proposed a spike and within two days we had a rough demo. It wasn’t reliable yet, but it proved we could build a privacy-safe scan at much lower cost than expected.

Quick conversations with customers showed they were far more willing to present ID when they knew nothing was being captured or sent.

We got behind building a first version, improved accuracy across ID types and languages, and soon had something ready to test.

Diary study and iteration

We ran a diary study with Research to capture what was working and what wasn’t in real deliveries. That led to:

A flashlight toggle for low-light conditions
Manual confirmation when the algorithm flagged someone as under 18
Blocking screenshots and screen recording at system level
Watermarking to protect customer data in edge cases
A calm fallback path for customers who refused scanning

Courier feedback

85% of couriers were satisfied with the new experience.

It makes it easier to convince the customers to present their ID when needed.

If I scan, I’m covered, it protects me.

It was really quick. I just pointed the camera at the ID and it was done. No issues, scanning completed.

Many said it reduced guesswork and made awkward moments easier. A few were concerned about very privacy-conscious customers, which we addressed through the manual fallback.

Results

This piece of work was a huge success, way beyond our stretch target:

90%+ ID verification rate
~3× the digital industry average
Months ahead of schedule
No negative customer feedback
Stable delivery times

Beyond the numbers

This project changed how teams at Stuart think about compliance:

From assuming fraud to looking at what people actually do
From enforcing rules to making the right thing easier
Product, Ops and Data Protection started working from the same model

The approach influenced other compliance work. It turned out that making the experience clearer worked better than making the rules stricter.