There is a particular kind of document that circulates in large organisations: beautifully formatted, forty pages deep, describing a three-year technology transformation in confident quarterly blocks. It names platforms, draws target-state diagrams and ends with a benefits curve that sweeps up and to the right. A surprising number of these documents share one quiet trait: nobody involved in writing them has ever shipped and run a production system.
It shows. Not in the diagrams, which are usually fine, but in the sequencing. Plans written by people who have operated systems put the riskiest assumption first, because they have been burned by discovering the hard problem in month eighteen. Plans written at altitude sequence by theme: foundations, then capability, then value, with the hard problem quietly parked in a later phase, where it waits patiently to detonate the whole timeline.
The riskiest assumption goes first. Everything else is scheduling.
Three tells of an operator-written plan
First, it pilots before it commits. Somewhere in the first quarter there is a deliberately small, deliberately real test of the thing the plan most depends on: the data migration nobody has scoped, the integration the vendor swears is standard, the team that has never run what it is about to own. The pilot is designed to fail cheaply if it is going to fail.
Second, it prices operations alongside delivery. Every new platform in the plan carries its run cost, its patching burden, its on-call load and its exit cost, because the authors have personally been on the wrong end of a system that was cheap to build and ruinous to keep. Total cost of ownership is not a slide, it is a habit.
Third, it names owners. Not teams, not workstreams: people. A milestone without a name against it is a wish. Operators know this because they have watched unowned milestones slide, quarter after quarter, while the steering committee nodded.
Why this matters more now
Technology decisions are getting harder to reverse. Cloud commitments, data platforms and the systems now arriving with machine learning inside them all reward organisations that sequence carefully and punish the ones that discover constraints late. The cost of a plan written without operating scars has gone up, and it was never cheap.
What to do about it
- Ask who on the planning team has run, in production, something like what is being proposed.
- Find the riskiest assumption and check where the plan tests it. If the answer is "phase three", renegotiate.
- Require a run-cost line for every new platform, owned by the person who will pay it.
- Put a name, not a team, against every milestone in the first year.
We write roadmaps this way because we build and operate our own platforms, and the plans we hand to clients are ones we would be willing to execute ourselves. That sentence is either true of your advisers or it is not, and it is worth finding out which.