What ten failed redesigns taught us

For six years we have kept an internal document we show to nobody: the list of projects that ended badly. Ten entries so far. Each one describes what failed, the moment we could have seen it coming, and what we changed afterwards. It is by far the most useful document in the studio.

Here is what it contains, with the names taken out.

The client was not the decision-maker

Four cases out of ten. We worked for weeks with a convinced marketing director, only to discover at the sign-off meeting that a quiet founder held a veto and an opposite opinion. The work was not bad. It was addressed to the wrong person.

What we do now: the first meeting does not start until we have a written list of the people who can say no. If that list is withheld, we do not sign.

The brief described a symptom

"Our site looks dated." In three of our failures, that sentence covered a commercial problem design could not treat: an unreadable offer, a price in the wrong place, a market that had moved. We delivered a new site on an untouched problem, and the client quite reasonably found that nothing had changed.

What we do now: we ask what should measurably happen six months after delivery. If the answer is not something we produce, we say so before quoting.

Sign-off happened on mockups that were too beautiful

A polished presentation page, with perfect copy and three commissioned photographs, proves nothing about how a system holds up. Two of our projects were approved on those images and collapsed on contact with real content: nine-word headlines, low-resolution assets, product pages with no photograph at all.

What we do now: every presentation contains at least one degraded case. The longest headline on the current site, a page with no image, a twenty-row table. It is less seductive. It is the only way to approve something true.

Nobody owned the brand after us

The most frequent case, and the most painful, because it only becomes visible a year later. We deliver a coherent system, it is applied correctly for three months, then a brochure is rushed out, a salesperson digs up an old template, and the whole thing comes apart without anyone having decided to.

What we do now: the rollout phase names a responsible person on the client side, with identified working time. No name, no rollout. We have lost two engagements on this point and have not regretted it.

We accepted a schedule we knew was wrong

Twice. A launch fixed by a trade show, a perfectly real constraint, and our agreement given in the knowledge that the approval cycles would not fit the deadline. Both projects shipped on time and badly, which is the worst combination: the client pays full price and gets work we do not show.

What we do now: we propose a reduced scope that holds rather than a full scope that does not. Ship the brand and the marketing site for the show, move the sales tools out by two months. That proposal has been accepted every single time we have made it.

What we did not learn

Two failures remain without a clear lesson. In one case the company changed chief executive mid-engagement and the new leader wanted their own brand, which is understandable. In the other, we simply never clicked, without either side being at any identifiable fault.

It felt important to leave them in the list with no conclusion. A document where every failure dutifully produces its lesson is a document that has been tidied, and it eventually stops teaching anyone anything.