An operating model is a picture of a steady state. The organisation that has to arrive at that steady state is not in it yet, and almost all of the work, the cost and the risk sit in the distance between the two. That distance is rarely drawn, never costed, and it is where reorganisations quietly fail.
A global payments acquirer and processor had a design for how its consolidated digital estate would be run once the pieces were in place. Eleven roles, a task split across nine domains, a headcount impact of net one or two extra people after new, modified and removed roles were netted off. That is a defensible number and a cheap one, which is the kind of number a business case is built on.
Then a second pack was written for the period before any of it applies. The window runs from the launch date to a point three to four months later, when the model would be approved and populated. Inside it: eleven activities, three governance groups to create from nothing, and a site migration now phased across roughly four months rather than completed at launch.
The challenge
The pack gives three reasons for the window. Momentum on the site had to be maintained rather than paused while the organisation caught up. The operating model itself was presented but not finally approved, and the roles it names were not filled. And the migration of the remaining properties had moved from a single event to a phased one running across the same period.
The supply side is where it gets hard. The advisory engagement covering digital leadership, programme management, technical leadership and experience design was due to finish on the launch date. The build agency's main work finished on the same day, with a clean-up increment already written into its contract, which the agency wanted to run for two weeks from the day after. The hosting arrangement was contracted for twelve months, which made it the only commitment on the board that outlasted the launch by design. Internally, some roles were in place, with a stated mix of gaps and upskilling still required, and several people were due to come off the wider programme over the following weeks.
On the day after launch the organisation has the fewest people it will have at any point in the story, and the largest number of open decisions.
The approach
The work broke the window into six steps: hand over the site and its support, stand up governance and the methods that go with it, agree the scope and timing of the next development increment, confirm what third parties would still do, define the immediate tasks and how to resource them, and take the whole thing to the executive team for sign-off on the future model.
Underneath sat the activity list and two grids. Every activity got a row, and every row got three columns: the specific tasks, the resources needed, and the options for supplying them. Twelve rows in total. That is the artefact worth copying: it forces the awkward question one line at a time instead of once, in aggregate, at the end.
Six of the twelve rows carry an explicit resource figure. One person to set up and run the three governance groups, with programme office support. One to map, document and train people on the content workflow and the prioritisation process. One for day-to-day technical ownership of the site and the vendor relationships underneath it. One analytics resource to build the reporting and teach people to use it. Around half a person spread across three areas of standards and policy. And one to two days a week over two months to finish the organisation design itself, plus a project management role to deliver it. The five whole and half figures add to roughly four and a half people for the length of the window. The design line sits on top of that, and the project management role written beside it carries no figure at all.
The other six rows carry no number. Content publishing is described as individuals trained on the platform following a defined process. The next release increment is described as user experience work, business analysis, project management, creative and build. Agency support is described as an experience-led role plus administrative and project support. Those are the rows that overrun, because a row with no number is not a plan, it is a hope with a heading.
Against all of it sits an end-state design costed at one or two additional people. The two figures are not in contradiction, since the transitional demand is temporary and the end-state number is permanent. They are in contradiction in the only way that matters, which is that the end-state number had been costed and put in front of people, while the transitional demand arrives at the back of a working session pack as budget and people implications still needing executive approval.
The governance detail is the sharpest thing in the pack. Three groups were to be created inside the window: a pan-group governance forum, a content and editorial group, and a prioritisation forum. A later slide describes two main forums to establish, extending an existing steering group, with subject-specific groups underneath on named initial topics. Whether that is three bodies or two plus satellites was still unresolved, in a pack whose whole purpose was to resource the standing up of them.
And the terms of reference for the main forum were assigned to the advisory team whose engagement was ending on the launch date. The group that is leaving writes the rules for the group that is staying. Every transition contains a line like that one, usually in a resourcing annex rather than on the summary page.
The outcome
What was delivered here is a working session pack: objectives, a current-state summary, an activity list, two resourcing grids with options against each line, and a closing summary stating that some internal capacity was available immediately, that governance set-up, requirements and prioritisation, programme management and policy work were not covered, and that there were budget and people implications requiring executive approval. The summary also flags that continued support from the departing specialists needed confirming quickly, before the individuals who held the knowledge were assigned elsewhere.
No structure was implemented in this phase. The operating model had been presented and final approval was not confirmed. The launch date had not yet arrived. Every figure above is a proposal, an option or an estimate written before the fact.
Three things would be done differently today, and only one is about tooling.
The first is the six unnumbered rows. A rough number that is wrong is more useful than no number, because it can be argued with. Publishing an estimate against every line, with its confidence stated, converts a resourcing grid into something an executive can approve or refuse.
The second is that the handover would be instrumented from the first day rather than described. The content workflow that a full-time line was to spend the window mapping and documenting is a process leaving traces in a content system, a helpdesk queue and a deployment log. Process mining over those events is what our Platform work puts in first, because it produces the map with a source attached, refreshed continuously, and it keeps producing it after the person who wrote the document has moved on. The same applies to the analytics build: reporting that carries lineage back to the events it came from survives a change of owner, and reporting that lives in one person's spreadsheet does not.
The third is a caution rather than a method. The early and careful language-model pilots now being proposed in content operations tend to be offered at exactly this moment, as cover for a publishing or support gap. Sometimes they help. But a pilot cannot chair a governance forum, cannot hold a vendor relationship and cannot decide what ships next. Our Consult engagements start by separating the work that is genuinely automatable from the work that is simply unstaffed, because the second kind does not respond to technology at all.
The end state was costed to the person. The period that had to produce it was costed on half its lines and funded on none of them. That is the usual order, and it is the wrong way round.