Steve Messer

The Workshop

Weeknotes S03E09

Light-bulb moments, chats with good people, reading research. All part of this week in my life as a civil servant. Tell you what, I could do with a reset.

Three Things That Happened

We did more things than this


For what may be the last time during this quarter, I got a few folks on the Explore Personalisation mission together in a room to think about personal data as part of service design and features. We spent about 90 minutes saying ‘Hmm…’ while moving some Post-its around, then drew a diagram on the wall (which is covered in whiteboard paint). It was a funnel with four levels and some words and arrows either side. That was a bit of a light-bulb moment.

It was a great moment because, to be honest, working on this nebulous mission has pushed me to the limits of my craft – and at times I’ve doubted I’m even any good at it. Trying to move a team of highly intelligent people through a sensitive design problem is Rather Hard, especially when they’re nabbed to work on other, more pressing issues in government.

So hopefully the product of our hard-thinking is valuable and answers a question correctly for the higher-ups. We would have preferred to focus instead on being present for the important task at hand, not sliced up, wafer-thin, and draped across several sandwiches.


Another large portion of the week was spent beginning to craft the narrative to sell our work. That means I poured some words into a slide deck and reviewed the history of GDS, while also tearing apart some poor research (Internet of Things for your benefits history, anyone?) which informs decision-makers. Better polls are available, so the narrative will include those.


We’re getting mighty close to rolling the Elasticsearch upgrade out to Production, but first we’re checking whether the upgrade will affect the ordering of search results – which it will, though not by much. 80 per cent of the top 1,000 searches return the same link in first place; 60 per cent have the same links in first and second place. The next question to answer is ‘How many searches return the same links in the first set of ten results?’ Because as long as a link doesn’t move more than four or five places, users will likely still be able to find it easily.

Since we’ve spent a while digging in to how weightings and similarity scores work, we’re rather well placed to work on those next quarter – which we’ll have to anyway since indexing changes completely between v5 and v6.


What I Learned

That, hooray, I’m not an imposter! Neither are you, probably. We’re all making it up as we go along. And so here’s the best life advice I’ve ever received, for you to remember and act on.

What I Could Do Better

Make my monthly objectives part of my weeknotes. I said before that I’d do that and I haven’t. Definitely worse at making lists recently. Need to return to making lists. Self-organisation? Yes, I could be better a self-organisation.