IThe question nobody can answer
Every large organisation documents who reports to whom. Almost none can say, with confidence, who owns the order service, who decides the data retention policy, or who is the last person who understands the LDAP integration. The org chart answers a question about management. The question people actually ask each day is about ownership: who do I talk to about this?
The answer usually lives in someone's head, in a wiki page last edited two reorganisations ago, or in the memory of whoever handled the last incident. It is discovered at the worst moment: during an outage, in a steering meeting, or the week after someone leaves.
Two failures hide inside that. Orphans: things nobody owns, so nobody decides. And thin knowledge: things only one person understands, so one resignation turns a system into archaeology.
IIWhat organisations try
| Approach | What it answers | Why it fails |
|---|---|---|
| Org chart | Who manages whom | Says nothing about systems, decisions or knowledge |
| RACI matrix | Four roles for every task | Too many cells to keep true; rots within a quarter |
| Service catalogue | Who owns each system | Covers software, not initiatives or areas of judgement |
| Wiki of owners | Whatever someone wrote | No sign of whether it is still true |
| An index | Owner, builders and knowers of anything | Needs a light, regular habit of confirmation |
IIIThree relations and one rule
The index records only three relations between a person and a thing. A thing can be a system, an initiative, or an area of judgement such as security reviews.
- owns
- Decides about it, and answers for it. Exactly one person.
- builds
- Changes it, day to day. Any number of people.
- knows
- Can explain it without looking it up. Any number of people.
The rule: every thing has exactly one owner, and the owner is a person, not a team. Teams cannot be phoned, and committees cannot decide. 37signals has long argued for small groups with clear, individual responsibility, and for hiring people who can be trusted to run their own work. Drucker's test for any decision is blunter still: until a named person is responsible for carrying it out, nothing has been decided.
Knowing is deliberately separate from building. The person who wrote the integration five years ago may no longer change it, but is still the one to call. Recording that is how thin knowledge becomes visible before it becomes a loss.
IVFreshness without administration
An ownership record is only useful if it is believed, and it is only believed if it is recent. So every relation carries the date it was last confirmed. After ninety days it fades and gains a dagger: unconfirmed. Confirming takes one click, and people confirm their own entries; nobody maintains the index on everyone else's behalf.
A quarterly minute per person, spent saying “still true”, is cheaper than a single incident spent finding out who owns the thing that broke.
VReading the gaps
Printed in red, like the rubrics of old manuscripts:
- Nobody owns it. The most dangerous line in any index.
- Its owner is leaving, or has left. Found on the day the leaving date is entered, not after.
- Only one person holds it. One owner who is also its only builder and only knower: a single point of knowledge.
And in the margin, in grey: relations nobody has confirmed for three months, and people who own so much that they cannot own any of it well. Drucker's advice to executives was to concentrate: to do first things first, and one thing at a time. An owner of nine things is concentrating on none of them.
VIWhat it deliberately leaves out
Reporting lines, job levels and access rights belong to other systems. The index does not track tasks, dates or progress; other tools do that. It answers one question well, and that question is who.