The process graveyard audit: how stale is your documentation?

Every company has a process graveyard. It is the SharePoint folder with the org's "official" procedures, the Confluence space nobody has touched since the reorg, the Visio file that only opens on one laptop. About 80% of business processes are never documented at all, and most of the rest are documented once and never again.

The graveyard is invisible day to day because work still gets done. People ask the person who knows. The cost only becomes visible at the worst moments: an audit, a resignation, a new hire's third month of guessing, an automation project that stalls because nobody can describe the process it should run.

Here is the audit. Answer honestly; nobody is watching.

The five questions

1. The open-rate test

Pick your operation's single most important procedure. When did anyone last open the document that describes it? If your wiki or document system shows view history, look it up. Months is common. Years is not rare. A document nobody reads is not documentation; it is liability theater.

2. The new-joiner test

Ask your most recent hire how they actually learned the job. If the answer is "shadowing" or a name rather than a place, your real documentation is a person. Onboarding by shadowing is the most expensive training program ever invented, and it compounds: each generation learns a slightly degraded copy of the last one's habits.

3. The version test

Find the current version of any SOP. If the filename contains FINAL, v7, or a date in parentheses, you have already failed this one. The question that matters is not "does a current version exist" but "can everyone, without asking, tell which one it is". If the answer involves asking someone, the version control system is that person's memory.

4. The exception test

Open any process document and look for the day things go wrong: the dangerous goods shipment, the failed payment, the missed handoff. Most documentation describes the happy path only, because drawing ten variant diagrams is miserable and maintaining them is worse. But exceptions are where the money and the risk live. A process description that only covers the easy case documents the part nobody needed help with.

5. The leaver test

Name the one person whose resignation would hurt operations most. Now ask: what fraction of what they know is written anywhere? One retirement is all it takes for twenty years of process knowledge to walk out the door. If your honest answer is "under half", their notice period is your documentation deadline, and notice periods are short.

Scoring

• 0 to 1 failed: rare and genuinely impressive. Your problem is keeping it true, not creating it. Skip to governance: review cycles, sign-off, and ownership.
• 2 to 3 failed: the normal zone. The knowledge exists but lives in people and stale files. You do not need a six-month mapping project; you need to resurrect what is already written and put owners on it.
• 4 to 5 failed: the graveyard is running the company. The good news: you are exactly who modern tooling is built for, and the distance from here to "audit-ready" is shorter than it has ever been.

Why it got this way

It is not laziness. The tools made this outcome. Diagram tools produce pictures that are stale the day they are exported. Document tools store prose nobody can query. Neither connects the description of the process to the running of it, so the description and reality drift apart from day one, and everyone rationally stops trusting the description.

The way out: resurrect, do not rewrite

The instinct after a bad audit score is to commission the big rewrite: workshops, consultants, a wall of sticky notes. Six months later you have a new graveyard. The better move is to start from what already exists. Your old SOPs, wikis, and Visio files are wrong in places, but they are 70% right, and extracting and correcting beats authoring from blank.

That is the approach FLOW takes: drop the old documents in, AI pulls out the steps, owners, and decisions, and a human approves the result. From there, review cycles and sign-off keep it true, and the pilot is designed around exactly this: bring one SOP, leave with it living.

But the audit above is useful whatever tooling you choose. Run it quarterly. The first time is a diagnosis; the trend is the real metric.