Three of the five things I'm proudest of aren't on this page, and that's worth explaining before you scroll.

Two years under an NDA — A payments product for a bank. I can describe the problem, the constraint and what I got wrong, and I can't show a pixel of it. Ask in an interview and you get the useful half
One thing that got killed — Six months of work, cancelled a fortnight before launch for reasons that had nothing to do with it. It taught me more than anything on this page and there is nothing to look at
Work that was the team's, not mine — I won't put a shared piece up with only my name under it. Where I contributed to something good it's listed with what I actually did and whose it was
Old work I've outgrown — Deleted rather than kept with an apologetic date on it. A portfolio isn't an archive and nobody is owed a complete one
The messy middle of everything below — The version history, the routes not taken, the thing that tested badly. I'll walk anybody through it, and it's the part I'd rather be judged on than the final screen

What I did, in one line — Not "led the redesign" where I made three screens. The specific verb and the specific scope, so nobody has to work out later that it was smaller than it read
Who else was on it — By role, and by name where they're happy to be named. Design is almost never one person, and a page implying otherwise is lying quietly rather than loudly
Who the client actually was — Named, or described honestly as "a mid-size retailer" where the contract says I can't. Never "a global brand" when it was that brand's smallest market
Whether it shipped — Some of the best work never does. It's still worth showing, marked as what it is rather than left ambiguous
What I'd change now — On every piece. It is the difference between somebody who has grown and somebody with a tidy grid
No invented numbers — Nothing like "increased conversion 40%" unless it was measured properly and I can say who measured it and how. Most portfolio metrics are guesses set in a serif

The piece — A checkout rebuild for a mid-size retailer. Twelve weeks, three of us, and the only one on this page I can talk about in full
What I did — The research and the interaction design. The visual design was somebody else's and it was better than mine would have been
What happened — It shipped, and the drop-off moved. I don't have a defensible number because the release went out alongside a pricing change, and separating the two afterwards would be arithmetic dressed as evidence
What I'd change now — I designed the happy path for far too long. The error states got two days at the end of it and that showed within a week of launch
What it taught me — Start at the worst state and work backwards. It makes the first week of every project uglier and every launch considerably calmer
Replace this with your own — One piece, honestly self-critiqued, is worth six with a case-study template laid over them

A team with at least one other designer — I've been the only one twice. It is survivable, and it is not how anybody does their best work for very long
Somewhere research already happens — Rather than somewhere I'd be arguing for it monthly. That argument is winnable and it takes about two years, which is most of a good stint
Problems with constraints on them — Regulated, technical, awkward, boring to outsiders. Blank-page brand work is somebody else's idea of a good time
What I've stopped wanting — Seniority for its own sake. I stepped back off a management track once and it has looked like the right call twice over since
Notice period — One month, worked properly, with the handover written before the last week rather than during it
Salary — Happy to give a band before either of us spends an hour on a call. Ask and you get a number, not a question back

No unpaid design test — A paid half-day is fine and I'll suggest it before you have to ask. "Redesign our onboarding as an exercise" is a brief with the invoice removed
No spec work and no pitches — Including competitions. Somebody gets paid for those and it isn't the four teams who lost
No design in front of a panel — It measures how well somebody performs while being watched. I'll walk through a real project in as much detail as anybody wants instead, including the parts that went wrong
No portfolio review with no role behind it — Where you're building a pipeline, say so — I'll probably still come, just not under the impression there's a job at the end of it
No take-home over three hours — And if the real number is six, tell me it's six. Everybody knows, and pretending otherwise only penalises the people who take the instruction literally
No asking what I earn now — It's a number about my last employer rather than about this job, and it's how a low salary follows somebody around for a decade