The inventory always surprises

Nobody has ever handed us an accurate list of what runs. What two weeks of walking the floor finds, and why the surprise rate never drops.

Pixel drawing of a clipboard checklist with two items ticked and one not

Inventory first. We have run the two-week walk on estates from forty apps to four thousand, and the surprise rate has never once been zero. Plan on it: whatever list you have, production disagrees.

The surprises sort into types. Write these down before you start and check them off as they appear.

  1. The running thing nobody listed. Average found per hundred servers: six. Scheduled tasks, services installed by a vendor in 2016, the "temporary" sync job in year nine.
  2. The listed thing not running. Decommissioned in spirit, billed in fact. We found one shop paying maintenance on an app whose server had been powered off since 2022. Nobody had told procurement.
  3. The owner who left. Every estate has systems whose last knowledgeable human has a farewell-cake photo in the intranet archive.
  4. The dependency that runs the company. A spreadsheet, an Access database, a shared mailbox with rules. Not on any diagram. Processes payroll adjustments.

The wider market is carrying the same blindness at bigger scale. SAP's own deadline makes the point: ECC maintenance ends December 2027, and industry surveys still put over 60% of ECC customers not yet migrated as the window closes. You cannot sequence what you have not counted. Most of those organizations do not have a migration problem yet. They have an inventory problem wearing a deadline.

Our method is not clever, which is why it works. Pull actuals, not intentions: network flows, process lists, login events, billing exports. Interview the people who do the work, not the people who own the org box; ask "what breaks if we turn this off Friday" and watch which direction they look. Every entry gets four fields: what it is, who answers for it, what business function dies without it, when a human last touched it on purpose. Resist extra fields. Forty-column inventories die in the spreadsheet stage.

Two weeks is enough for the first honest pass. Not for perfection; for honesty. The goal is a list the CFO and the on-call engineer both nod at, dated, with the unknowns marked UNKNOWN in capital letters instead of quietly omitted. An inventory that admits its gaps gets maintained. One that fakes completeness gets ignored by March.

Nobody has ever regretted knowing what actually runs in production. Plenty have regretted the discount version. The roadmap comes next, and it is only as good as the count it stands on.