Getting to multiple daily releases

A short time ago I helped a company, and part of our achievement was to go from fortnightly releases to twice-daily releases within three months. This has raised eyebrows with some people when I mention it. So I thought I’d note down two factors which helped make it happen…

1. People who believe

There was definite skepticism when I proposed that we could be releasing our web software every day. It was large and complex and inevitably consisted of much legacy code. But we talked through the concerns and eventually enough of the team thought that it might be possible, and started to act as if it was.

A particular example was the objection that regression testing usually took two full days. But we agreed that with smaller, clearer changes there should be much less to regression test. It was particularly important that the QA Manager supported this. I agreed that quality should not decline — indeed should increase. With her on-board much else followed.

2. Lack of technology

This team had very little automation in the release process, and that was actually a good thing. I’ve said before that less technology is better, and this is a case in point. There was very little release process locked down, so it was relatively easy to change things.

I’ve worked with teams with heavy automation, and for them it would have been much more difficult to make that change, because so many automated processes would have had to have been reconfigured and rewritten.

The future

The first time we put out two releases in a day was within three months of starting our new approach. Within another three months twice-daily releases was a regular occurrence and the team started talking about what more they’d need to optimise to make thrice-daily releases the norm. The fact that they started asking this without prompting was fantastic to see, and was a sign of great confidence and trust in each other. I’m looking forward to seeing even more wonderful things from them in future.