Cutting your SOP library by 60–70% — without losing control
Ask a team how many SOPs they have and the honest answer is often "too many to know." Libraries grow by accretion — a new document for every incident, every audit finding, every well-meaning improvement — until nobody can say which version is current, or whether two documents contradict each other.
A library that big stops being a control. It becomes a risk.
Smaller, if done right, is safer
It sounds counter-intuitive that removing documents improves control, but it's consistently true. In the rationalisation work I've led, a 60–70% reduction in SOP volume is normal — and the result is a library that's more accurate, more consistent and far easier to keep current.
The reduction isn't deletion for its own sake. It comes from:
- Mapping first — anchoring every SOP to a process step in the architecture, so you can see overlaps, gaps and orphans.
- Merging duplicates — collapsing near-identical documents into one authoritative version.
- Retiring the dead — removing SOPs for processes that no longer exist or have been superseded.
- Rewriting in plain language — so the people who actually do the work can follow them.
Control comes from ownership, not volume
The thing that makes a rationalised library hold its shape is governance: named ownership for every document, a review cadence, versioning, approvals and a change log with a full audit trail. That's what turns "we tidied up the SOPs once" into a continuous improvement pipeline that stays tidy.
It's also what makes the estate demonstrable — to an auditor, to a regulator under Consumer Duty, and increasingly to the AI tools you'll want to point at it.
A practical first move
Don't try to rationalise everything at once. Take one Level 3 process, map it, and rationalise the SOPs that hang off it. You'll prove the method, free up the team's time, and produce a template you can roll out — with the confidence that nothing important was lost along the way.
Fewer documents, each one trusted and owned. That's the goal — and it's entirely achievable.