How an EPM implementation exposes the misalignment that slide decks can hide for months.
A Florida retail chain with two brands, about 80 stores, and several sales channels had its executive committee approve a category-level profitability model. The vote was unanimous. All twelve members agreed. Nobody objected. The project sponsor presented it to the board as a textbook case of consensus.
Fourteen weeks in, during the build, the CFO walked into the project room while the team was loading fulfillment and last-mile costs into the electronics category. “This isn’t what we approved,” he said. The VP of Merchandising answered, “It’s exactly what we approved.” Both were right. They had signed the same memo and pictured different models. The CFO thought he had approved gross margin by category from the general ledger. The VP understood contribution margin, with cost-to-serve allocated by activity-based costing (ABC).
The company found the problem in week 14. It could have surfaced in week 2 at no extra cost. Recent Boston Consulting Group research published in Harvard Business Review gives this common mistake a name: “false alignment.” The study looked at why transformations fail. Among nearly 2,000 companies, over 70% failed to outperform their industry average after a downturn, whatever time frame was used. Technology and data have advanced enormously. Real executive alignment is as hard to get as ever.
In EPM projects, these failures usually start with misaligned definitions. The technology is rarely the culprit. The committee assumes a shared understanding that isn’t there. Three questions separate models that guide decisions from models that only reconcile figures.
- Why does a committee approve ‘better planning’ without realizing that each member has agreed to a different model?
- Why does the data model disclose this gap so late, and what is the cost of that delay?
- How can you settle the definitions early rather than waiting for the build phase to force the issue?
Identical terminology may carry divergent meanings for different stakeholders
False alignment is not about dishonesty; it is a cognitive bias. The BCG research connects it to the false consensus effect, described by Lee Ross and his Stanford team. We tend to overestimate how many people share our views. A Harvard decision scientist put it this way. If she loves vanilla ice cream, she will keep thinking most people do too. In the C-suite, an executive who backs an initiative assumes everyone else backs it for the same reasons.
In finance, this ‘vanilla’ has a technical name. It is the confusion between FP&A and BP&A. When the committee approves ‘planning,’ the CFO pictures the general ledger, IFRS, legal entities, and consolidation. The operating VP pictures capacity, cost-to-serve, and profitability by customer or category, built with ABC. When someone says ‘profitability model,’ one person imagines profit by entity and another thinks of contribution margin by managerial object. The words match. The meanings don’t.
The authors tell a story that could open any EPM kickoff. Two manufacturing executives both want to ‘improve the margin.’ One plans to raise prices. The other wants to cut unit costs. Until they get into the details, neither realizes they disagree, so neither gets to negotiate. The false consensus effect can keep that misunderstanding alive for months.
For the Florida retailer, those months were expensive. Our internal estimate put the master-data rework at about 9 person-weeks, with the launch slipping a full quarter. All because no one asked a simple question in week 2: Profitability of what? Approving ‘planning’ or ‘profitability’ without naming the managerial object only postpones the disagreement. It will come back, and later it costs more.
The data model provides objective clarity; the build phase ultimately enforces the decision the committee previously deferred
This is what sets EPM apart from other transformation programs. It is also why model-driven projects are so exposed to false alignment.
False alignment can hide in presentations for a long time, because PowerPoint tolerates ambiguity. A phrase like ‘profitability model’ can sit in a bullet point and nobody asks for a definition. A data model has no room for that. The dimensional model is a formal agreement written in system logic. Once you define dimensions, hierarchies, allocation drivers, and managerial objects, every open question has to be answered.
So the build becomes the decision forum that should have happened during design. The delay is expensive. It shows up as change requests, scope disputes, and implementation consultants mediating between executives who are only now discovering they expected different things. EPM is where false alignment comes to the surface, usually at the worst and most expensive moment.
BCG outlines three ways an execution team can fall apart when leaders never agreed, and each has a direct counterpart in a model project. Paralysis is the team that builds prototype after prototype and never goes live, because it cannot choose between competing visions. Hyperactivity produces a model loaded with every dimension any executive asked for, plus reports that exist to please one person and cannot be maintained. Tunnel vision builds the model for one narrow reading. You get a flawless GL consolidation that reconciles to the cent and answers no business question.
The last scenario is the most dangerous, because it looks like success. The model passes validation and matches the general ledger. It guides no decision. I repeat this in every steering committee: validation shows the numbers add up, but governance shows you can rely on the model. They are different tests. A model built on false alignment can reconcile perfectly and still guide nothing.
The verdict before the build
The report also has good news. False alignment can be prevented, and the steps to real agreement fit neatly into the early stages of any implementation. The authors highlight Alexander Lacik at Pandora. His case matters for finance because what he fixed was a problem of definitions.
When Lacik became CEO in 2019, the Danish jewelry company had just been through a hard decade. In a single day in August 2011, it lost 65% of its market value, and four CEOs came and went over the next few years. The transformation plan was already in place, including a $400 million cost-reduction program. Lacik found 46 active priorities on his leadership team’s list. Instead of asking for more alignment, he gathered the team for two days with one rule. Nobody leaves until the list is down to 12 priorities with full commitment. He encouraged open disagreement, went through each priority, and had the team vote on what stayed.
The agreement rested on one specific metric. Growth was measured only from end-customer demand, in stores open more than 12 months, including online sales. Anyone in FP&A or BP&A will recognize that as the precise definition of a managerial object. EPM projects tend to postpone this kind of definition until it gets expensive. Lacik settled it before execution started.
Bring this into your Analyze and Define stage. Your version of Pandora’s off-site is settling the FP&A versus BP&A differences during design, before the build. Start with a facilitated workshop where every key stakeholder says openly what the model is for and how they will use it. Use structured facilitation methods such as decision matrices or the RAPID framework to clarify who has input, who owns the outcome, and how decisions will be made. To make it concrete, ask each function to write down its position on the costing method (ABC or full absorption, which driver, which object). Individually and in writing. Writing alone curbs groupthink and brings silent disagreements into the open. Next, meet one-on-one with each owner of a managerial object to pin down boundaries and clear up what is still ambiguous. Close with a group workshop that reviews every position and negotiates one shared definition. Then get a formal, signed agreement on the dimensional model. Treat it as seriously as signing a check. Sign off the master data definition before the build, and the data model becomes a working tool instead of a source of conflict.
Three essential concepts and a practical action for immediate implementation
Three takeaways. First, in EPM, false alignment is a definition risk, and it lives in the gap between FP&A and BP&A. Second, the dimensional model is what brings disagreements to light, and if you wait for the build, that always happens late. Third, real agreement is cheap at the start or expensive at the end. There is no middle ground.
If you have an EPM project underway or about to begin, here is what you can do this week. Bring together the owners of your three core definitions (profitability, planning, and budget). Ask each person to write down, on a single page and by themselves, which managerial object they believe the model will produce. Do not let them discuss it first. Then compare their answers. If they differ, you have just identified, in an hour and at no cost, the same issue that took the Florida retailer nine weeks to uncover. Write down the agreed definition and have everyone sign it.
If the group still can’t agree after the first round, don’t force a temporary compromise. Escalate the dispute to a neutral executive sponsor or governance panel for a facilitated decision. Boil the disagreement down to its points and use a structured decision model to weigh the impact of each option. Real consensus may take more workshops, or even a formal vote. Don’t proceed until everyone holds the same clear definition. Planning for this keeps the process transparent and stops the project from stalling on unsettled differences.
Over the next 12 to 24 months, as models move inside ERP systems and AI agents take over parts of planning, the temptation to ‘configure first, define later’ will grow, because the tools seem to handle ambiguity. They don’t. The more automated the model, the sooner and more expensively false alignment will bite, since no human consultant is left to bridge conflicting definitions. Settling definitions before the build stops being good practice and becomes the last line of defense. In regulated sectors like banking, where supervisors such as the CNBV, SIB, and SBP require traceability for the managerial numbers behind decisions, a signed agreement on definitions goes beyond efficiency. It is becoming a governance requirement.
Alignment also has to survive changes in the business and the technology, so governance can’t stop at the first sign-off. Set a review cycle, quarterly or semiannual, where the ownership group revisits the core definitions behind your data models and formally reaffirms or updates them. The governance committee should watch for shifts in strategy, reporting needs, or legal requirements that touch those definitions. Document each review. That leaves a traceable record and stops silent drift. Name a definition owner, or a cross-functional panel responsible for definition stewardship, so shared understanding and accountability last. Treat definition governance as an ongoing discipline and the model will keep pace as the organization grows.
