The replacement I designed kept 74% of the revenue we were losing

ESP was collectionHQ's selection module, the thing that told librarians what to buy. We were losing it because the data that fed it was owned by our parent company, Baker & Taylor.

Baker & Taylor went into liquidation. As a result the data stopped, the platform ESP lived inside went with it, and every subscriber was heading for the exit. We couldn't just wait it out, because renewals were coming up and there were only two honest things we could offer a customer. We could give them a refund, or we could give them something else to buy.

ESP was sold as a subscription service, and Baker & Taylor staff ran it on the customer's behalf. So the interface behind it was made for their own trained super-users and nobody else. It was over a decade old and pretty slap dash, and it had never needed to be easy because the people in it were in it every day.

That meant this was never going to be a rebuild. The librarians who'd now be sitting in front of it had never seen it before. So whatever I made had to do a specialist's job in the hands of someone doing it for the first time.

I designed the study, picked and recruited the participants, and made sense of what came back. I was away for part of the fieldwork, so I coached a colleague to run the calls. I sat in on the early sessions so I could tell him what he'd done well and what to do better next time. I moderated the second study myself.

We spoke to eleven US public library systems: Boston, Frisco, Gail Borden, Hinsdale, New Orleans, New York, Phoenix, Saint Charles, San Francisco, San Jose and Scottsdale. Every conversation ran for an hour or longer, and what they described was a workflow that had quietly fallen apart.

Library software had settled on a look. It was all blues, greens and the occasional red, and very little of it looked like it had been built this decade. Nobody was going to mistake a new entrant for a safe pair of hands anyway, so I stopped trying to look like one.

The one colour I couldn't use was our own. cHQacademic is our sister product and it already owns teal, so sharing it would've made cHQselect look like a variant of something else right when it most needed to look new. I wrote that down as a rule: no teal, no orange, no circles. That ruled out collectionHQ's own visual language as well as everyone else's.

I wrote the positioning and the voice as well as doing the visuals. Then I took four logo routes to the director team, gave them my reasoning for each one, and told them the decision was theirs. They went with the stacked books, and I built the design system out from there.

We launched cHQselect at PLA 2026 in Minneapolis, which is the US public library conference. I designed and wrote the launch page as a sales tool more than a brochure, so the team could book demos straight off the stand. What happened next was we generated just over £2m in sales pipeline that week.

At launch, the only way to get titles into a cart was to upload a list of ISBNs. Librarians told us plainly that if they couldn't search for titles, and see the data they'd need to judge what to buy, they wouldn't use cHQselect for selection at all. So discovery was the next big piece of work, and it lets them search right inside the tool.

I ran ideation workshops with the wider team to get the concepts out, and then I built a clickable prototype of the whole journey. It had four screens. There was a search screen with two ways in, an AI prompt and a manual search. There was an AI screen that asked follow-up questions to narrow things down. Then there were the results, and a single title view.

Then I put the prototype in front of ten librarians from seven libraries across the UK and the US, all of them cHQselect customers. I ran every session myself, about thirty minutes each, and I told them to be as brutal as they liked.

The research had validated a transparent title score. It answered the problem selectors kept describing, which is that if a number hides its workings, people just go and check it somewhere else.

I couldn't build it in time. There were far too many technical constraints, and if we'd persevered it would've delayed a launch the business couldn't afford to delay. So a few weeks after launch we shipped Key Insights instead, which gives you little bits of information about an author, a subject or a title in a clean, digestible way. It isn't as good as a score, and I'm not going to pretend it is. But it gave people part of what the research had validated on the date that mattered, and we're actively working on scoring now.

I'd make the same trade again, because a smaller thing on the date the business needed it was worth more than the right thing late.