A side project put me in charge of scheduling recurring work for physical spaces, and it quietly fixed how I think about service level objectives.
The setup: recurring house and office cleans, mostly every two weeks. The customer is not buying a spotless room. They are buying a guarantee that the room never gets past a certain point.
That distinction took me embarrassingly long to see, and it is the same one I had been fumbling at work for years.
Perfect is the wrong target
Early on we scoped every recurring visit like a full reset. Same work as a deep clean, every time. It was expensive, it ran long, and customers did not value it more. They rated it the same as the cheaper version.
What they noticed was variance. A visit that ran short. A week that got skipped. The bathroom being fine four times and then not fine once.
Availability targets work the same way. Nobody experiences 99.97 percent. They experience the one outage that landed during their demo. The number is a proxy for a much simpler question: can I stop thinking about this?
The threshold is the product
Once we scoped recurring visits as keep it above this line rather than make it perfect, three things fell out:
- The work became predictable. Same rooms, same order, same duration. Predictable work can be staffed, priced and measured. Heroic work cannot.
- We could tell when we were failing. A defined line means a missed line is visible. Without one, quality is just vibes and whoever complains loudest sets the standard.
- We stopped over-delivering on things nobody checked. That capacity went into the two or three things every single customer looked at first.
That last one is the error budget idea wearing different clothes. If you are never missing your target, you are not being reliable, you are being wasteful. You have bought reliability nobody asked for with money and time that had somewhere better to go.
Maintenance is not remediation
The other thing that transferred cleanly: the gap between a maintained space and a neglected one is not linear. Skip two visits and the work does not double, it quadruples, because things bond, set and stain. Recovery costs far more than prevention, and the curve gets steeper the longer you wait.
Every engineer already knows this about dependencies, migrations and test suites. We just do not act on it, because prevention never produces a visible win and remediation always does.
When we finally wrote the standard down, it went into the scope of work for recurring cleaning in Toronto as a defined threshold with a defined cadence, which turned out to be the part clients compared vendors on.
The takeaway
An SLO is not a quality bar. It is a promise about consistency, and consistency is what people are actually buying. Aim at a line you can hold every time rather than a peak you can hit occasionally. The peak impresses nobody who has to live with the average.
Top comments (0)