Product at the end of its life cycle? Try the Scream Test

One of my teams was responsible for maintaining over 3,000 applications, utilities and audience facing services. We didn’t know who owned or cared about the majority of these. This mysterious majority had been commissioned in the mists of time by people who’d moved on. 3,000 was too many and we needed to cut. Our only recourse was the “Scream Test”.

Scream test: early surprise shutdown of a service you are retiring to see if anyone screams

Simple but brutal. Turn off the service and wait for the shouting.

I’d started a new job and found my team struggling to push ahead with new development and also support the legacy. We needed to cut our maintenance overhead. 3,000 was too many. Way too many. And we didn’t even know if they were active. It was only through organisational habit that we kept them alive.

So we turned off 1,000 of the 3,000 as a first attempt at tidy up.


Nobody screamed. Nobody noticed. Nobody cared. Those applications, utilities and audience facing services were no longer relevant to the organisation and my team had been diligently keeping alive software that should have died years before. They stayed turned off.

After that we continued the approach. Eventually we got to the point where people did scream, or even better, complained when we even suggested we’d turn off their precious service. So we got more selective about what we tried the scream test on and what we turned off.

The great thing was that by the time people started screaming the maintenance load was markedly reduced. And that meant we’d also significantly increased our capacity to develop new stuff.