Lesson16 min

How many services, and at what grain

The granularity test

The right grain is determined by one question: can a single, sensible impact tolerance be set for this service?

"Retail banking services" — too broad. Withdrawing cash and taking out a mortgage do not carry the same urgency.
"PIN validation at the payment terminal" — too fine. That is a technical component, not a perceived service.
"Make a card payment" — right grain. One tolerance, an identifiable client harm, a mappable chain.

For each candidate, ask:

Does an external client consume it directly? If not, it is a supporting service, not an important one.
Does its disruption cause harm — not mere inconvenience?
Can one tolerance be set that holds across all its use cases?
Is there an identifiable chain of resources supporting it?
Can an owner be designated who can answer for it?

Five yes answers: it is an important service. A single no warrants revisiting the phrasing.

Entity size	Number of services	Note
Small single-line entity	5 to 10	A short list signals clarity
Mid-sized entity	10 to 20	The most common range
Multi-business group	20 to 40	Often segmented by business line
Beyond 50	—	A signal the grain is too fine

A list of 120 services is an application catalogue in disguise. It will be neither mapped, nor tested, nor reviewed.

Three families almost systematically escape the first inventory:

Services delivered to internal clients who are themselves regulated — for example a group entity depending on a shared platform;
Seasonal or episodic services — an annual campaign, a filing deadline;
Exit services — account closure, data retrieval, termination. Their disruption is rarely visible but immediately litigious.

An inventory containing none of these three cases deserves a second pass.

Key takeaways