I have watched a specific failure play out enough times to be able to predict it: an organisation decides to take data stewardship seriously, launches a programme with real enthusiasm, fills in a spreadsheet of "data owners" and "stewards," holds some meetings — and then, quietly, over a few months, the whole thing dies. The spreadsheet goes stale, the meetings stop, and a year later there's a document nobody trusts listing owners who've changed roles, for data that's changed shape, maintained by no one. Stewardship that didn't stick. Having led data platforms through this, I want to write about why it dies and what actually makes it survive, because the difference isn't effort — the failed programmes often started with more of that than the successful ones.

Why stewardship programmes die

The failure is so consistent that it has to be structural, not a matter of the wrong people or insufficient will. And when I look at what kills these programmes, it's almost always the same three causes:

  • It's built as extra, virtuous work. The classic stewardship programme asks people to do additional work — maintain the catalogue, update the ownership records, attend the governance meetings — on top of their actual jobs, for no reward beyond having been good. And unrewarded extra work, however virtuous, loses to the day job every single time. Not because people are lazy, but because that's what unrewarded extra work does. The programme was designed to depend on sustained virtue, and virtue is not a sustainable fuel.
  • It's detached from where the work actually happens. The stewardship lives in a separate place — a governance spreadsheet, a catalogue tool nobody opens in their normal flow — disconnected from the tools and moments where people actually touch the data. So maintaining it requires remembering to go somewhere else and do a thing, and "remember to go update the governance thing" is exactly the instruction that quietly stops being followed.
  • Its value is invisible to the people doing the work. The steward maintaining the records rarely feels any benefit from it — the value (if any) accrues to some future person querying the catalogue. Work whose cost is borne by one person and whose benefit lands on someone else, later, is work that withers, because the person paying the cost has no felt reason to keep paying it.

Notice these aren't people problems or tooling problems. They're design problems — the programme was designed in a way that guaranteed its own decay.

What makes it stick instead

So the question isn't "how do we get people to care more" — that's trying to fix a design problem with exhortation, and it never works. The question is "how do we design stewardship so it survives without depending on sustained heroism." A few principles I've seen actually work:

  • Attach stewardship to the work, not alongside it. The durable move is to make the stewardship a byproduct of doing the normal work, rather than an extra task beside it. Ownership captured as data is created, in the tools people already use, rather than recorded separately later. Documentation that's part of building the thing, not a governance chore afterward. When the stewardship happens inside the flow of the actual work, it doesn't depend on anyone remembering to go and be virtuous somewhere else.
  • Make the right thing the easy thing. Wherever the governed path and the lazy path diverge, close the gap — so that the stewarded, documented, owned version is also the convenient one. If the easiest way to do the work is also the way that maintains the stewardship, it maintains itself. If stewardship requires swimming against the current of convenience, it drowns.
  • Make the value felt by the person paying the cost. Stewardship sticks far better when the steward benefits from it — when maintaining good metadata makes their own work easier, when the catalogue is something they actually use to find things, when ownership means their expertise is recognised rather than just their liability recorded. Align the cost and the benefit onto the same person and the work sustains itself; split them and it decays.
  • Start narrow and let it earn its expansion. The comprehensive programme — steward everything, catalogue everything — is a way to be busy for a quarter and finished with nothing. Start with the data that genuinely matters, make stewardship stick there, prove the value, and expand from a working core. Coverage is the enemy of stickiness; consequence is its friend.
A stewardship programme that relies on people doing extra, unrewarded, remembered work will die, every time, no matter how much they care at launch. One that's a byproduct of the normal flow, easy to do right, and beneficial to the person doing it, survives — because it stopped depending on virtue and started depending on design.

The scale dimension

The "at scale" part of the title matters, because scale is exactly what exposes a fragile stewardship design. A stewardship programme that depends on heroism might limp along for a small, dedicated team. Scale it to a large organisation with hundreds of data assets and dozens of teams, and the heroism-dependent design collapses immediately — there simply isn't enough sustained virtue to go around, and the gap between the pristine launch and the stale reality becomes a chasm.

So designing for stickiness isn't a nicety at scale; it's the only thing that works at all. At scale, the only stewardship that survives is stewardship that's been engineered to not require sustained effort from anyone — built into the flow, made convenient, made beneficial, kept narrow enough to be real. Everything else is a spreadsheet with a death date, and the date arrives faster the bigger you get.

The honest takeaway

If you're about to launch a data-stewardship programme, or watching one you launched quietly die, the lesson is the same and it's not about trying harder. Stewardship fails when it's designed as virtuous extra work detached from the flow and unrewarded to the doer, and it succeeds when it's designed as a low-effort byproduct of the normal work that benefits the person doing it. The determinant is design, not dedication — and the good news in that is that a dying programme usually isn't a sign your people don't care. It's a sign the programme was built to depend on caring, and caring was never going to be enough. Rebuild it to depend on design instead, and it'll stick — quietly, without anyone having to be a hero about it, which is exactly how you know it's working.