Driver-Based Forecasting Multi-Entity Groups Can Defend to the Board
The board asked which 12 percent was real and nobody could answer. Driver based forecasting multi entity groups can defend starts where the roll up breaks.
Executive summary
- A single growth rate applied across entities with different business models produces a precise group number that no operator can explain.
- Driver-based forecasting in a multi-entity group needs two or three causal drivers per entity, chosen because an operator controls them.
- The causal chain usually breaks at the group tab, where heads, stores, and backlog get FX-translated and summed into one revenue line.
- Every driver assumption needs a named operator who signs it off, so a miss traces back to an assumption and its owner instead of the model.
- Carry entity ranges and FX effects into the board pack, or the group forecast shows a false single point and a blended growth rate again.
A Riyadh-based group builds its FY27 budget across four entities. They are a stable facilities-services business in Dubai, a two-year-old retail rollout in Riyadh, a project-based contracting arm in Jeddah, and a shared logistics unit. FP&A applies one assumption to all four, 12% revenue growth, AED and SAR alike. The consolidated forecast lands at SAR 184.35M. On the page, it carries all the precision of driver-based forecasting across a multi-entity group, and the board approves it.
Three months in, retail is running 40% behind plan and facilities services is 6% ahead. At the Q1 review, a director asks which of the four 12% figures was ever real. Nobody in the room can name the headcount added, the stores opened, or the contracts renewed behind any of them. The group had a number precise to two decimal places and no operational reasoning a controller could defend.
The usual response is to refine the rate. FP&A splits it into 9% for services and 18% for retail, and the problem survives in smaller pieces. Each entity needs the two or three operational drivers that move its revenue and cost lines. Those drivers then have to reach the board pack with the causal link intact, and most groups lose that link at the roll-up.
One growth rate, four business models
Facilities services grow when it adds billable headcount and renews contracts at better rates. Retail grows when it opens stores and those stores mature toward steady-state footfall. Contracting grows when a signed backlog converts into billed work. The logistics unit grows only when its sister entities do, because its top line is an internal allocation.
The flat rate survives because it is fast and ties to the target the board already expects. Nobody asks what sits underneath until the entities diverge. This article assumes the group already has the shared structure a roll-up forecast needs before drivers even enter the picture, and starts where the drivers come in.
The easiest driver to pull is usually last year's revenue
Groups that move off a flat rate often label the new model driver-based, then fill the driver column with whatever the system exports cleanly. Last year's revenue by entity is the common choice. The model gains a tab called Drivers and still behaves like a trend line, because none of its inputs cause the outcome.
Driver-based forecasting across entities needs a stricter test for each candidate. Can an operator in that entity control the driver or commit to a value for it? Can you explain in one sentence how a change in it moves revenue or cost? Billable headcount and utilization pass for facilities services. Store count, footfall per mature store, and basket size pass for retail.
Correlation is the quieter trap. In a young retail rollout, revenue may track marketing spend for two years, but store openings drive both. A driver chosen for correlation holds until the cause shifts, then fails without warning. Two or three causal drivers per entity is usually enough, and a longer list tends to model noise.
Multi-entity driver-based forecasting snaps at the SUM formula
Assume the group gets entity drivers right. Facilities services forecasts billable heads times utilization times rate, in AED. Retail forecasts stores times footfall times basket, and contracting forecasts backlog burn by project, both in SAR. Each model is sound in its own units and currency.
Then FP&A builds the group tab. It links each entity's revenue line, translates at the budget rate, and sums. Heads, stores, and project milestones become one SAR revenue line, often before anyone has checked each driver against what its operator can commit to. The board pack shows group revenue up 11.8%, and the board reads one blended growth rate again, the number the exercise set out to replace.
Currency hides the problem for a while. AED and SAR both peg to the dollar, so translation between them barely moves the number. That changes when the group adds an entity in Egypt. A large move in the pound can outweigh every operational driver in that entity's model. A summed group line gives the board no way to separate the two.
Ranges disappear at the same step. Retail might carry five to nine store openings and contracting six to ten new awards. FP&A sums the midpoints, and the group forecast arrives as one number with no band around it. The gap widens when entity risks move together, as when retail and contracting both slow with the same government spending cycle.
Nobody signed the store-opening assumption
In the Riyadh group, FP&A wrote eight new stores in H2 into the retail model. The head of retail saw the figure on a budget review slide and did not object. That silence became the assumption. By September three stores had opened, the variance reached the board as a forecast miss, and the discussion turned to the model.
The miss belonged to an assumption with no owner. FP&A can design the driver structure and test it for causality, but it cannot commit to store openings, hiring plans, or contract renewals. The operator who controls each driver has to sign off its value for each forecast version. When that operator leaves mid-cycle, the replacement restates the assumption before FP&A rebuilds the roll-up.
Ownership also sharpens the conversation after the forecast. Each variance lands in the variance-driver categories that show up after the forecast, not before it, with a name beside it. The head of retail explains five openings against eight, and FP&A explains the model.
A framework for driver-based forecasting across a multi-entity group
Run each entity through six steps before it enters the group tab.
- Name the business model. Services, retail, projects, or internal allocation. The type tells you where causal drivers live.
- Pick two or three causal drivers. Apply the control-and-explain test and reject any driver whose only merit is a clean export.
- Name one owner per driver. Choose the operator who can commit to the value, and record their sign-off each forecast version.
- Fix the unit and currency. Keep heads, stores, contracts, AED per billable hour, and SAR per basket in local terms until the last step.
- Set the roll-up rule. Translate at a stated rate, report FX effect apart from driver effect, and keep entity and driver lines under the group total in the board pack.
- Carry the range. Hold low, base, and high values per driver, roll them up per entity, state which ranges move together, and show the group as a band.
Applied to the Riyadh group, the first pass looks like this.
The logistics unit has no external driver, so it inherits volumes and ranges from the entities it serves and eliminates on consolidation. A fifth entity added mid-year, say an operation in Cairo, goes through the same six steps. If it has no drivers yet, label it trend-based so the board knows which part of the number has nothing underneath it.
When the roll-up becomes a reconciliation job
The chain runs from entity data to drivers, local-currency models, FX translation, roll-up, and board pack. A disciplined team can rebuild it by hand each cycle at two or three entities with stable drivers. Past that, one late change breaks it.
The head of retail revises openings after FP&A builds the group tab. FP&A then spends two days tracing which driver version fed the number the CFO already sent. A director asks to see retail at five openings, and the team has to flex one entity without disturbing currency and timing assumptions in the other three.
Kudwa does not pick drivers or set assumptions. That judgment stays with FP&A and the entity operators. Kudwa's analysis and forecasting layer sits on top of the existing entity models.
With multi-entity consolidation, it keeps driver-level detail intact through FX translation and roll-up. The board pack then shows the group number with entity and driver lines still attached. The same structure supports measuring whether the driver-based forecast actually performed once the period closes.
The next time a director asks what moved, the CFO can answer in stores, heads, and contracts, with a name beside each.
See the drivers behind the group number
If your group forecast loses its entity drivers at the roll-up, see how Kudwa carries them through FX translation into the board pack on the Analysis & forecasting page. Book a demo to walk through it with your own entity structure.
Latest from Kudwa
Deferred Tax in Multi-Entity Consolidation: The DTA No UAE Entity Return Shows
Every entity return reconciles, yet the group is short a DTA. See where deferred tax in multi entity consolidation hides in eliminations, and how to measure it.
Segment Reporting vs Entity Reporting: Building a Business-Line View That Survives the Close
The analyst who splits the trading entity was out, so distribution vanished from the board pack. Segment reporting vs entity reporting needs a written map.
