CEL Solicitors · Client evidence portal
Getting evidence to the right place without the wrong friction
People who’d just lost money to fraud were getting stuck uploading evidence. I redesigned CEL’s portal to fix that.
At a glance
The problem
More than half of new cases needed at least one evidence chase, costing case handlers an estimated working week every month. Most of the causes traced back to the portal itself.
What I did
Audited the old portal against Nielsen’s heuristics, then redesigned it mobile-first: separate sign-in, a guided evidence picker, card-based upload and a confirmation that proves it worked.
Where it landed
In internal testing, 5 of 6 participants completed every task unaided on the new prototype, up from 2 of 6 on the old portal, and average time to submit fell from about 4 minutes to under 2. It was in development when I left CEL, with a target of at least halving the share of new cases that need an evidence chase.
Where this sits
CEL Solicitors is a UK consumer law firm that handles fraud and scam recovery cases on a no-win, no-fee basis. Once a client signs up, they need to submit evidence through a web portal: bank statements, chat transcripts, scam correspondence, photo ID. A case can’t move forward until this evidence is in, so a confusing upload step doesn’t just frustrate someone, it holds up their claim.
The existing portal was a single-page form. Enter your case reference and postcode, drag files into an upload area, pick a category from a dropdown, hit submit. It technically worked, but clients kept struggling to upload evidence correctly, and case handlers were the ones hearing about it. That feedback loop, clients reporting friction to the case handlers who then flagged it to me, is where the problem definition came from. The portal assumed people knew how to create a zip file, could categorise their own documents through a generic dropdown, and would understand jargon like “CEL Number”. None of that was reasonable for the audience.
Over half
of new cases needed at least one evidence chase
1 week
of case handler time a month spent chasing evidence (estimated from a two-week log)
No. 1
reason case handlers had to contact clients
What was falling apart
Case handlers told me chasing evidence was the most common reason they contacted clients. To size it, I asked three of them to log every evidence chase on new cases for two weeks. More than half of new cases needed at least one follow-up about evidence, many of them more than once. Scaled up from those two weeks, that came to roughly a working week of handler time every month spent asking clients to re-send things.
The reasons broke down clearly, most common first:
- Wrong category
- Required document missing
- Upload failed or nothing attached
- Wrong file type or unreadable
- Other
Most chases traced back to things the interface could prevent: unclear categories, a submit button that accepted nothing, and no file constraints. Missing documents were a different problem. Clients often didn’t know what to send, and the portal only partly solves that.
I reviewed the existing portal against Nielsen’s heuristics and cross-referenced the findings with the client complaints case handlers had been relaying. It hadn’t been formally reviewed since it was built, and grounding the audit in the actual user, someone dealing with a fraud case, likely anxious, often on mobile, made the failures harder to dismiss. Seven issues mapped directly to usability failures, spanning error handling, categorisation and trust.
The authentication problems surface immediately. The error banner fires before the user has done anything, overlapping the page heading and undermining confidence before a single field has been touched. The “CEL Number” label offers no format hint or example, so people either guess or call support. All three validation errors appear at once on an empty form, which is disorienting and reads as the portal blaming the user for something they haven’t done yet.
The upload experience has its own problems. The upload area shows no file type or size constraints, so people only discover the limits after a failed attempt. The evidence type dropdown expects clients to map their own documents onto CEL’s internal categories with no guidance. “Add More Evidence” is ambiguous: it isn’t clear whether it adds another file to the same category or starts a new one. And the submit button stays active with nothing attached, leaving the door open to empty submissions.
Making the case
I was the sole designer and owned the work end to end, from the heuristic review of the old portal through to the final designs handed over for development.
The old portal had no analytics, only basic upload logs, so the case had to come from the people absorbing the cost. I took the handlers’ two-week log to the fraud team leader as hours lost each week rather than as a usability complaint. The pushback was development capacity, not the problem itself.
So I scoped the redesign to work within the existing backend, with no new file processing, which turned a “not this quarter” into a yes. Anything needing new backend features had to wait. Once it was approved, the case handlers who’d logged the calls became my first reviewers.
The rebuilt flow, end to end
Before the individual design decisions, here’s the flow I designed, from first landing through to confirmation.
How I fixed it
The redesign started from the problems, not the interface. Each decision answered something specific that was failing: a support call generated, a file sent to the wrong category, a client left wondering whether their evidence had even arrived. The goal was a portal a fraud victim could use without having to think too hard, on whatever device they happened to have in their hand.
Two more things shaped every decision. This user arrives already primed to distrust: they’ve just been scammed, and now a website is asking for bank statements and ID. For them, confusion and suspicion are the same feeling, so reducing confusion was really about lowering the sense of risk at each step. And the design is mobile-first because the data said so: the portal’s upload logs showed most evidence arriving from phones. Every layout, touch target and line of instruction was designed and tested at mobile widths first, then adapted up.
Separated authentication from upload
The new flow starts with a dedicated sign-in screen: CEL reference number and postcode, with placeholder examples showing the expected format (e.g. FRA123456, M1 1AA). The format examples matter more than they look: they remove the fear of getting it wrong on the very first step, before any trust has been built. Trust badges for encrypted transfer, GDPR, SRA regulation and no third-party sharing run along the footer, in view while someone decides whether to hand over their details.
Replaced the dropdown with a guided evidence type picker
The portal now asks “What are you uploading?” in a modal with clearly labelled options, each with a plain-language subtitle. An “Other document” option means no one gets stuck. The picker does the categorisation work for the user instead of expecting them to understand CEL’s internal taxonomy.
Card-based upload with clear hierarchy
Each evidence type becomes its own card with its own upload zone. File constraints are shown directly in each upload area. “Add more evidence” and “Upload Evidence” are visually separated: the first sits inside the card container, the second outside and muted until at least one file is attached, which prevents empty submissions.
The upload landing screen
The landing screen sets expectations before anyone uploads. A short “How to upload” block explains the three steps in plain language, so first-time users aren’t guessing at the pattern. The “Add new evidence” button opens the type picker, and the main “Upload Evidence” button stays disabled until at least one file is attached, so the primary action never pretends to be ready before it is. For an anxious user, that removes a specific fear: the “did I just submit nothing?” moment that sends people straight to the phone.
Redesigned the confirmation screen
This screen carries more trust weight than any other in the flow. A defrauded person needs proof it worked before they’ll believe it did, so the new thank-you page shows the case reference number, a submission timestamp, a summary of what was uploaded by category, and a “What happens next” section with three concrete steps. An explicit “You can safely close this window” line gives people permission to leave without second-guessing whether it worked.
Designed where it matters most
Most clients upload their evidence on a phone, usually at home and often while they’re still dealing with the stress of the scam. So these are the screens I designed first, and they’re shown here at the size people actually use them. Every tap target, label and line of instruction had to work on a small screen before anything else.
The three key screens at phone scale: upload state with files added, evidence picker, confirmation.
Internal testing
I tested the old portal and the new prototype with 6 volunteers from across the business (none from the legal team), on their own phones, using the same four tasks:
- Sign in with a given reference and postcode
- Upload a bank statement showing the scam payments
- Upload two types of evidence in one submission
- Confirm your evidence was received
| Measure | Old portal | New prototype |
|---|---|---|
| Completing every task unaided | 2 of 6 | 5 of 6 |
| Average time to submit two types of evidence | About 4 minutes | Under 2 minutes |
| Failed to sign in at least once | Several, some more than once | 1 of 6 (a postcode typo) |
| Choosing the wrong category | Common | Rare, only between the two bank statement options |
| Tried to submit with no files | 2 of 6 | 0 of 6 |
| Sure their evidence had arrived | Most unsure | 5 of 6 |
Sign-in was where the old portal lost people. Most participants didn’t know what a “CEL Number” looked like, and several failed more than once before getting in. With format examples in place, the only error was a typo in a postcode.
The biggest shift was confidence. On the old confirmation screen, most participants weren’t sure anything had happened. On the new one, 5 of 6 were sure their evidence had arrived: the reference number, timestamp and summary of uploads did the reassuring for them.
The only person who needed help on the new prototype got stuck choosing between the two bank statement options, and others hesitated there before picking correctly. That’s why sharpening those labels is first on my refinement list.
These were internal participants, not real clients, so treat this as directional. Live data will be the real test.
Where I left it
The redesign was complete and tested internally, and in development when I left CEL in October 2026. The target I set was to at least halve the share of new cases that need an evidence chase within three months of launch, and to see “wrong category” drop out of the top reasons, since that’s the clearest test of the evidence type picker.
What I recommended for launch
Internal volunteers aren’t the real audience, so I handed over three recommendations for launch.
- Collapse the instruction block for returning users. It earns its space on a first visit but eats a lot of vertical room on mobile.
- Sharpen the two bank statement labels. “Scam transactions only” versus “full 12-month history” would draw the line more clearly than the current pair.
- Track completion from launch. The old portal tracked nothing, so completion rate and time to submit need measuring from day one.
The real test is simpler than any metric: someone who’s just been scammed getting their evidence in, on their phone, without having to ring anyone to do it.
That call was the original symptom, and a case can’t move until the evidence lands. If the redesign holds, the call doesn’t happen and the claim progresses. That’s the bar it has to clear.