From Blueprint to a Model That Runs: Closing the Implementation Gap
The most expensive cost model is the one that was beautifully designed and never ran. This piece is a composite scenario drawn from real field experience across cost-model implementations; the sector, the players and every figure are illustrative and represent no single client. It is about the quiet, common failure in management accounting: a pristine conceptual design, often signed off by a global advisory firm, that stalls somewhere between the slide deck and a live engine that answers questions on a Monday morning.
The implementation gap is the distance between a costing model that is designed and one that actually operates, refreshing on real data, producing numbers a controller will defend, and changing what managers decide. Most designs do not fail because the accounting is wrong; they fail in execution. Three forces do the damage: there is no implementer who holds both the management-accounting logic and the platform skill in the same head, so the design is handed across a translation gap; the platform was chosen for prestige and procurement comfort rather than for whether anyone can actually run it; and the scope outruns the data, promising a granularity the source systems cannot feed. Specialist implementation closes the gap by building the model as an operating asset from day one, a defensible cost architecture, a data pipeline that survives the next monthly close, and outputs wired to real decisions. The payoff is not a prettier model; it is the shift from arguing about the numbers to acting on them. With Cost and Profitability, the CostCtrl platform and AI-assisted mapping, the bridge that used to take many months of consulting effort can now be crossed in days.
A design is a promise; a running model is the payment
There is a particular kind of document that circulates inside large organisations: the cost-model blueprint. It is usually excellent. It has flow diagrams of cost pools and activities, a driver for every allocation, a governance chart, and an appendix that anticipates the awkward questions. It was produced by capable people, often a Big Four firm or a global advisory firm, and it survives its steering committee to warm applause. And then, in a surprising number of cases, nothing happens. Twelve or eighteen months later the design is still a design, the old spreadsheet allocations are still running the business, and the sponsor has quietly moved on.
This is the implementation gap, and it is not a rare accident. The long-running Standish Group CHAOS research has found for decades that only a minority of technology-enabled projects finish on time, on budget and with the intended result, with large and complex programmes faring worst. A costing model is exactly that kind of programme: part accounting, part data engineering, part change management. A blueprint is a promise about how the business will understand its own economics. A model that runs is the payment on that promise, and the two are separated by work that no slide can do for you.
How a strong design stalls, in one field-composite story
Picture a large, capital-intensive infrastructure operator around 2019, the details are a composite, drawn from several engagements and standing in for none. It commissioned a full activity-based costing design from a global advisory firm. The deliverable was genuinely good: a multi-layer cost architecture, several hundred activities, dozens of driver definitions, and a target of allocating shared network and support cost down to service and route level. The board approved it. Then the model had to be built on a heavyweight enterprise cost-management platform the group had licensed, chosen, as these things often are, because it was the safe, prestigious name that procurement and audit were comfortable signing.
That is where it stalled, and the reasons rhyme across many such projects. There was no single person who understood both the accounting intent and the platform. The advisory team knew the design but not the tool; the internal IT team knew systems but not why activity A should absorb cost pool B. The platform, powerful on paper, needed specialist configuration that nobody available could supply, so a change to one driver took weeks and a queue. And the scope had been drawn against an idealised data model: the granularity the design assumed simply was not present, or clean, in the source ledgers and operational feeds. Each of those three failures is survivable alone. Together they are how a board-approved model spends two years as a document.
The three gaps between the deck and the engine
The skills gap. A costing model lives at an unusual intersection. It needs someone who can defend why a cost is assigned the way it is, the management-accounting logic that a controller will be cross-examined on, and, in the very same decision, someone who can make the platform actually do it. In most programmes those are two different people on two different contracts, and the design crosses a translation gap every time it moves. Meaning leaks at that boundary. The implementer who holds both skills at once is rare, and their absence is the single most reliable predictor that a good design will not ship.
The platform gap. Enterprise software is frequently chosen for reasons that have nothing to do with whether anyone can run the model inside it: brand comfort, an existing licence, audit familiarity, a procurement scorecard. A heavyweight enterprise costing suite can be entirely capable and still be the wrong choice if every change needs a scarce specialist and a multi-week ticket. Usability is not a luxury in a cost model; it is what decides whether the model gets updated after the consultants leave, or freezes on the day they do. Our companion note on why models drift back to spreadsheets is really a story about this gap.
The data gap. Ambition in a costing design is measured in granularity, allocate everything, to the lowest possible object, with the finest drivers. But a model can only be as detailed as the data that feeds it every month, cleanly and on time. When scope outruns the source systems, the model either cannot be populated at all or is populated with heroic manual effort that collapses at the second monthly close. Gartner's research on analytics has repeatedly found that a large share of built assets never reach real, sustained business use, and a costing model that cannot be fed is the accounting version of exactly that.
Building the model as an operating asset, not a document
Specialist implementation is not "the design, but faster". It is a different discipline, and it starts from a different question: not "is this architecture elegant?" but "will this still run, unattended, at the third close?" That reframing changes every choice. Cost pools are shaped around drivers the data can actually supply, not the drivers a whiteboard would prefer. The allocation logic is built so a controller can trace any number back to its source and defend it, the same traceability that activity-based management depends on to change behaviour rather than just report it. And the whole thing is wired to a repeatable data pipeline, so the monthly refresh is a routine, not a project.
The specialist also does the unglamorous triage that decides success: which 20% of the design carries 80% of the decision value, so the model can go live useful and grow, rather than waiting years to go live complete. Kaplan and Anderson made exactly this argument when they introduced time-driven activity-based costing, that the older, fully elaborated ABC models collapsed under their own maintenance weight, and that a leaner, time-equation approach was what let a model actually stay alive in a real company. Specialist implementation is the operational expression of that lesson: build the model to run, not to impress, and let the running earn the right to more detail.
What changes when the model leaves the slide
The clearest way to see the value of closing the gap is to watch what happens to decisions. The table below contrasts how a management question is handled when the cost model lives in a deck versus when it runs as a live engine. The examples are illustrative, representative of the pattern rather than any one client's results.
| Decision | Blueprint in PowerPoint | Model that runs (illustrative) |
|---|---|---|
| "Which services actually lose money?" | A one-off study, months old, argued over because no one can reproduce the numbers | Refreshed every close; a loss-making service is visible the month it turns, with the cost trail attached |
| Pricing a new contract | Gut feel plus a blended average rate, because the model cannot be run on the new scenario | The specific cost-to-serve is modelled in minutes, so the bid reflects real economics not an average |
| Cutting cost under pressure | Across-the-board percentage cuts, because there is no defensible view of where cost concentrates | Targeted at the activities and idle capacity the model exposes, sized against a shadow price not a hunch |
| Reviewing unused capacity | Invisible; fixed cost is fully absorbed into output and the waste hides inside the rates | Cost of unused capacity is broken out explicitly, turning a hidden loss into a managed number |
| Board and investor questions | Answered slowly, defensively, with caveats about the model's age and assumptions | Answered in the meeting, from a live model, with drill-down when someone pushes on a figure |
| Trust in the numbers | The debate is about whether the model is right | The debate is about what to do, because the model has stopped being the argument |
That last row is the real prize. The point of closing the implementation gap is not analytical vanity; it is moving the entire organisation's energy from litigating the numbers to acting on them. That is also the foundation for anything downstream, honest profitability KPIs, an economic value added view that charges for the capital each unit consumes, or a portfolio-level profitability diagnostic across holdings. None of those can stand on a model that only exists as a slide.
The traps that keep the gap open
Treating go-live as the finish line. A model that runs once is not a running model. If there is no owner of the monthly refresh and no simple way to change a driver, the asset decays the day the implementers leave. PMI's Pulse of the Profession research has long linked poor outcomes to weak ownership and change discipline, and a cost model is unusually exposed to both.
Confusing the platform's power with the project's success. The most capable enterprise suite in the world adds nothing if no one on the team can operate it between engagements. The right question is never "how powerful is the tool?" but "who changes a driver next quarter, and how long does it take them?"
Designing for a data reality that does not exist. Granularity the source systems cannot feed is not ambition; it is a stall waiting to happen. Scope the model to the data you actually have each month, and let cleaner data earn more detail later.
Big-bang scope. A design that insists on allocating everything before it produces anything will usually produce nothing. The models that survive go live narrow and useful, then widen.
No thread from number to decision. If a manager cannot trace an output back to a cost they recognise, they will not trust it, and untrusted numbers do not change behaviour. Adoption is a design constraint, not an afterthought.
Closing the gap in days, not quarters
What used to make the implementation gap so wide was that all three failures were expensive to fix by hand. Finding one person with both the accounting and the platform skill was rare; configuring a heavyweight suite was slow; and mapping messy source data to a clean cost model was months of manual reconciliation. Each of those is now a different problem. Cost and Profitability brings the combined discipline, management-accounting logic and hands-on implementation in the same team, so the design never crosses a translation gap. CostCtrl replaces the prestige-heavy, specialist-only platform with one built to be operated by the finance team itself, where changing a driver is a task, not a ticket. And AI-assisted mapping collapses the data gap, proposing how source accounts and operational feeds map into cost pools and drivers, so the pipeline that once took a quarter of reconciliation is stood up and validated in days.
The claim is deliberately modest and specific: not that the accounting becomes trivial, a defensible cost model still takes judgement, but that the distance between a sound design and a model that runs has genuinely collapsed. The blueprint no longer has to sit in a drawer waiting for a scarce implementer, a friendlier platform and a data miracle. That is what it means to close the implementation gap: the model that was designed is, within days, the model that runs.
Common questions about the cost-model implementation gap
- Why do so many well-designed cost models never get implemented?
- Because design and implementation are different disciplines that rarely sit in the same hands. A model stalls when no one holds both the management-accounting logic and the platform skill, so the design leaks meaning every time it is handed across teams; when the software was chosen for prestige rather than usability, so every change needs a scarce specialist; and when the scope assumes data granularity the source systems cannot actually supply each month. Broad project research, such as the Standish CHAOS studies, has long shown that large, complex programmes fail in execution far more often than in concept, and a costing model is exactly that kind of programme.
- What is the difference between a cost model that is designed and one that runs?
- A designed model is a blueprint, architecture, drivers, governance, that describes how costing should work. A running model refreshes on real data every close, produces numbers a controller can trace and defend, and feeds actual decisions. The gap between them is operational, not conceptual: a data pipeline that survives the next month-end, a platform someone can actually update, and outputs wired to pricing, cost and capacity choices. The design is a promise; the running model is the payment.
- Does buying a powerful enterprise costing platform guarantee success?
- No, and often the opposite. A heavyweight enterprise suite can be entirely capable and still be the wrong choice if only a scarce specialist can operate it and every change takes weeks. Platforms chosen for brand comfort or procurement familiarity frequently freeze the model the day the consultants leave, because the finance team cannot maintain it. The decisive question is not how powerful the tool is, but who updates a driver next quarter and how long it takes them.
- How does specialist implementation actually close the gap?
- By building the model as an operating asset from day one rather than as a document. That means shaping cost pools around drivers the data can really supply, making every allocation traceable so a controller can defend it, standing up a repeatable monthly pipeline, and going live with the high-value 20% of the design rather than waiting for the complete 100%. It is the operational form of the argument Kaplan and Anderson made for time-driven activity-based costing: build a model lean enough to stay alive in a real company, and let its use earn more detail.
- Can AI and modern tooling really shorten a cost-model implementation to days?
- For the parts that used to consume the schedule, yes. The three classic delays, finding a dual-skilled implementer, configuring an unfriendly platform, and manually mapping messy source data, are each addressed by a combined implementation team, a platform built to be run by finance itself, and AI-assisted mapping of source accounts to cost pools and drivers. What does not shrink is the judgement in a defensible cost architecture. So the honest claim is narrower than "instant costing": the distance from a sound design to a live model has collapsed from quarters to days, not the accounting itself.
References
Standish Group, CHAOS Report (project success and failure rates, with large and complex programmes faring worst). · Project Management Institute, Pulse of the Profession (the link between project outcomes, ownership and change discipline). · Gartner research on analytics and data-and-analytics adoption (the persistent gap between assets built and assets used). · Kaplan, R. S. & Anderson, S. R. Time-Driven Activity-Based Costing (why elaborate cost models collapse under maintenance, and the leaner model that survives). · Kaplan, R. S. & Cooper, R. Cost & Effect (designing cost systems that inform decisions rather than merely report).