The refresh was mostly deletion

This is the fifth version of this portfolio. It is the shortest.

The previous one went up in a hurry and listed what I had done. All of it was true. None of it was specific to me.

I rebuilt it expecting to design it. What replaced it is not a visual system. It is a set of things the page is not allowed to do, and the design is what is left.

One column

Every line on this site is 560px wide. A 1440px dashboard artboard rendered inside that column lands at 39%, which puts a 12px UI label at 4.7px. So there are two figure types and one rule: scale what is read as gestalt, never scale what is read as detail. An establishing shot survives being shrunk because the reader is taking in layout. A crop does not.

The rule killed the obvious options, all of them for the same reason. A lightbox as the primary treatment turns the in-page figure into a thumbnail, the worst scale of anything I tried: 39% on a laptop, 23% on a phone. Video of prototype interactions lands at that same 39% once you capture it full-screen, and it only ever proved the product moved.

Then I screen shared a case study in an interview. The figures were too small to talk over, and on a phone a dashboard is a gray rectangle with some blue in it. So the lightbox went back as an escape hatch, which costs nothing until someone wants it. That is the difference between a page you can read and a page you can show.

Decks wearing prose

The older drafts had a common ancestor, and it was a slide deck. I learned that shape at frog. An agency deck is meant to be talked over: the slide carries the claim and you supply the reasoning out loud. That is fine until you paste the slide into a page and expect it to argue on its own. I found that out walking people through the old site, rambling over all of it, which is the tell. A page that makes its own argument does not need you in the room.

They stated what I did rather than what I decided. "Created a reusable motion language, guidelines and agent documentation that improved cross-team consistency" is four activities and no judgment. The decision underneath it, that the brand already had a motion language and the job was to transcribe it rather than invent a second one, was nowhere on the page.

They came in threes. One draft had four bullet triads in a row: three goals, three objectives, three foundation tasks, three strategy shifts. Three is what a slide holds. It is almost never how many true things there are. They all ended in Key learnings, three of those as well, each general enough to append to any project anybody has ever worked on.

They hedged the outcome hardest of all. One read "informed features and capabilities later acquired by existing products", written so that nothing can be checked and nothing can be wrong. What actually happened is that the incubation wound down and the scenario planning engine went to other teams. Smaller, verifiable, and much better.

And they deleted me from my own work. "Was designed", "was leveraged", "was added", about fifteen times.

The test

Cutting the case study was not compression. It was finding the two moments that actually turn, then putting them where a reader hits them early. Analysts expected a result and got an editor. The legacy screen kept the queue in view and I had thrown it away. Everything that survived is attached to one of those two.

If a passage would still make sense in another designer's portfolio, it was never mine. That is the test that killed most of this, and it runs one scale down as well, sentence by sentence. What cannot move is the decision and the problem that forced it, which is the only part a reader cannot get somewhere else.

None of that happened in one pass. One case study has taken more commits than the other two together, and the last long stretch of them changed nothing about its length. It was the right size long before it said the right thing. I could not tell what I had decided until I had written it down wrong a few times.

What the rules are now

An exploration grid is padding unless each rejected option comes with the one line saying what it lost, because a rejection with a reason is a decision and without one it is process theater.

Showing the legacy product has to be earned. The test is whether a reader gets the problem in one glance with no narration: a 163-case queue where every row is an account number and a name passes, "it looks dated" does not.

Desktop-only is a decision, not a limitation. Nothing about the work needs a large screen. The artboards do, and that is what the figure rules above are for. The test is whether a case study reads end to end on a phone.

Placement is an argument, and that is the one I did not expect. A shot of the finished console sat at the top of a case study whose opening paragraph described the legacy product, and it read as the cause of the slowness. Moved below the verdict, it contradicted the sentence above it. Adjacency implies relation in both directions and no caption undoes it.

None of these rules is about taste. Each one is a thing I can check, which is why the sixth version of this portfolio will be shorter than this one.