Good enough SLOs

When it comes to defining SLOs, I’m aiming for good enough.

Good enough to meet the requirements of the consumers.

No lower.

But no higher.

There’s a (potentially significant) cost to investing in better reliability, and that cost is only worth paying if it provides significant value to the consumer.

For example, if my consumer would like to have their data product updated daily, but must have data at least once a week, the SLO I will promise is once a week.

I’ll still run it daily, but now when something goes wrong at the weekend I know it’s fine to fix it on Monday and not need to have my team on-call for it, because providing on-call support is expensive.

I’ll also make different trade-offs when building it, with less investment in resiliency, which means I can deliver it faster and move onto the next product, delivering more value for my business.

So, good enough SLOs for the consumer, and greater value for the business.