.svg)
Forecasting per entity and cost center for SME CFOs: complete guide (2026)
Model decentralized at entity level, consolidate centrally at group level, and steer at cost-center level where it matters. Practical guide for SME groups.
.png)
Multi-entity forecasting: per-entity forecasts with own drivers + per-cost-center where relevant + central consolidation. Multi-level: consolidated, per entity, per cost center.
Forecasting per entity and cost center for SME CFOs: complete guide (2026)
Why you model decentralized, consolidate centrally, and steer at cost-center level where it counts.
TL;DR
For SME groups with multiple entities, forecasting at consolidated level alone is insufficient. The right architecture: per-entity forecasts with own drivers (manufacturing ≠ sales ≠ services), per-cost-center forecasts where relevant (healthcare, retail, services with hundreds of cost centers), and central consolidation with IC elimination, NCI, and multi-currency. Multi-level: consolidated, per entity, per cost center. Finstack supports this natively plus 2-way Excel/Sheets sync for model flexibility, from EUR 39/month per entity.
Why forecasting per entity matters
A consolidated group forecast tells you the total. It says nothing about where in the group that total comes from or where it goes wrong when reality diverges. For an SME group with multiple entities or business units, that’s a serious gap: operational decisions are typically made at entity level, not at group level. The holding director steers on group numbers; the operating-company manager steers on their own entity. Without per-entity forecasts, there’s no common language for the discussion between those two levels.
The practical pain points without per-entity forecasting: if group revenue is 8% below expectations, you don’t know which entity is the cause. If entity C structurally outperforms the group forecast, you don’t know whether that comes from internal efficiency or favorable market conditions you should be investing in. In board conversations about portfolio strategy, you have no per-entity projection to work with. In investor reporting, you’re forced to improvise about per-entity developments.
Per-entity forecasts solve this by connecting operational context to strategic steering. Each entity gets its own drivers, assumptions, and scenarios — built by whoever sits closest to it (entity controller or business-unit director). Holding level gets the consolidated view that aggregates across entities and eliminates IC flows. Both levels share the same underlying data, with different aggregation levels for different decisions.
For the broader context of forecasting in SMEs — including the four forecast types (budget, rolling forecast, latest estimate, and 13-week cash flow) — see the main guide on forecasting for SME CFOs. Per-entity forecasting works best in combination with driver-based modeling — each entity with its own drivers within a consistent structure — and for the consolidation part see the main guide on consolidation for SME CFOs.
Decentralized modeling, central consolidation
The role split in multi-entity forecasting follows one logical principle: let the modeling work sit as close as possible to operational reality, and the aggregation as close as possible to the strategic decision. Concretely:
Decentralized: the entity controller builds their own forecast
The controller or business-unit leader of each entity builds and maintains their own forecast with their own drivers and assumptions. This person sits close to the operational reality of their entity — knows the market, the pipeline, the operational challenges, the personnel changes — and therefore builds a more accurate forecast than someone at holding level could. Monthly update cycle per entity, fixed role division, own assumptions documentation.
Centralized: the holding controller consolidates
The holding controller or group FP&A consolidates per-entity forecasts into a consolidated group view, including IC elimination, currency conversion where relevant, and NCI handling for minority interests. The central work is therefore not the modeling itself but the consolidating, validating, and strategic interpreting. Without automated consolidation, this becomes manual Excel aggregation of separate entity workbooks — workable with 2 entities, unworkable beyond 5.
The practical pitfall: without tooling, this role split often collapses into one person at holding level trying to build all forecasts. That doesn’t scale and misses context. Good tooling (like Finstack) makes the decentralized/central architecture feasible without breaking version control or making consolidation manual work.
In practice it works as follows: the entity controller builds their own model in Excel or Sheets, Finstack delivers actuals automatically via the ERP connection, and the consolidated view at holding level emerges through aggregation with IC elimination. The holding controller doesn’t need to “understand what entity C has in their model” — only whether the group movements make sense and which entity is responsible for variances. That’s the right division of labor: detail per entity, coherence at group level.
Drivers differ fundamentally per entity
One of the reasons that one holding model for all entities rarely works: drivers differ fundamentally. A manufacturing entity and a sales entity within the same group need totally different forecast architectures. Examples from practice:
Forcing one forecast structure on all entities loses accuracy: a manufacturing entity with sales FTE as a primary driver says nothing useful; a services entity with sales volume says nothing useful. Let each entity define its own driver set within a consistent overarching structure (same P&L categories, same currency handling, same reporting frequency) and consolidate the outcomes.
What does need to stay consistent across entities: the overarching structure. Same P&L categories (revenue, COGS, personnel cost, marketing, sales, tech, G&A, etc.), same reporting frequency (monthly rolling), same currency handling (per-entity functional currency, consolidated in reporting currency). Within that structure, each entity is free to choose its own drivers and assumptions that fit its business model. That isn’t a contradiction but exactly the essence of decentralized modeling: freedom in detail, discipline in structure.
For the broader explanation of driver-based forecasting: see the complete guide on driver-based forecasting for SME CFOs.
Cost-center forecasting: when you need an extra layer
For some SMEs, per-entity forecasting isn’t deep enough. Within an entity, a serious cost-center structure may exist that itself deserves forecasting: per department, per location, per project, per team. Sectors where this is standard:
- Healthcare — hospitals, clinics, and care groups with tens to hundreds of departments, each with its own cost structure, personnel mix, and productivity norm
- Retail — chains with multiple locations where each location has its own revenue and cost pattern (rent, salary level, local competition)
- Professional services — consultancy or law firms with multiple practices or teams, each with its own utilization rate and hourly rate level
- Construction and project organizations — per project an own P&L with start/end date and project-specific margin
- Education and non-profit — per program, course, or grant with its own allocation
For these sectors, cost-center forecasting concretely means: budgets and projections at the level where they are operationally managed. A hospital with 80 departments wants to budget and forecast per department; a retailer with 40 locations wants to see per location how expected revenue and margin develop relative to budget and prior year. Without the cost-center level, you get aggregated entity numbers that are too coarse for operational steering.
Finstack supports cost-center forecasting natively: insight per cost center, re-mapping of transactions to the P&L at cost-center level (rather than only at GL level), grouping of cost centers across entities (e.g., all marketing cost centers together regardless of which entity they sit in), and forecasting at cost-center level with own drivers and assumptions. For groups with hundreds of cost centers, this is the difference between “we’ve spent EUR 300k too much somewhere in the group” and “cost center X in entity B is EUR 40k over budget due to specific Y reason”.
Consolidation: from entity forecasts to group view
Consolidating per-entity forecasts into a consolidated group forecast requires three capabilities: aggregation of entity numbers, IC elimination between entities, and handling of currency and NCI where relevant. None of these is trivial in separate Excel workbooks per entity with manual aggregation at holding level.
The Finstack approach to forecast consolidation works in parallel to actuals consolidation: same three elimination methods (full GLA, per IC relationship, per transaction), same entity grouping with toggle, same NCI handling. A forecast in which entity A projects sales to entity B is automatically eliminated in the consolidated view. A forecast in foreign currency is converted to reporting currency with the IFRS-compliant FX methodology you choose (fixed historical, period-end, period-average).
The practical benefit: the holding controller spends time not on manual consolidation but on strategic interpretation. “The difference between our consolidated forecast and last month’s comes from three changes: entity A has higher new-business pipeline, entity B is delaying production expansion, and FX shift on the UK entity adds 2%.” That kind of analysis is only possible when the mechanical consolidation is automated.
An often-underestimated requirement: consistency of consolidation methodology between actuals and forecast. If you consolidate actuals with method A (e.g., per IC relationship) and your forecast with method B (e.g., full GLA), the numbers will structurally drift apart and variance analysis becomes unreliable. Finstack’s shared consolidation engine for actuals and forecasts prevents this error: same methodology, same mapping, same NCI percentages.
Practical consequence of this shared engine: you can deliver variance reports between consolidated actuals and consolidated forecast without dispute — the numbers are structurally comparable because they’ve passed through the same consolidation steps. For PE portcos and groups with external reporting obligations, that’s the difference between reports that get debated during the board meeting and reports that get accepted as a factual baseline.
Toggle, NCI, and what-if scenarios at entity level
One important capability for multi-entity SME groups that’s rarely mentioned: the ability to turn entities on and off in scenarios. Finstack’s entity grouping with toggle supports this natively. Applications:
- Acquisition scenario — add a new entity to the consolidation set, with pro-forma forecast for the first 12-18 months post-acquisition. Compare consolidated group with and without the acquisition to see the impact on cash flow, margin, and EBITDA.
- Divestment scenario — take an entity out and see what that does to the consolidated group. Useful for portfolio discussions or in preparation for a sale.
- Sub-group views — toggle a subset of entities for specific board reporting or segment analysis (all Dutch entities, all B2C entities, all holdings).
- Minority interest (NCI) — for ownership stakes below 100%, Finstack supports ownership percentages so consolidated numbers reflect the right ownership ratios.
The effect: forecasting becomes an instrument to support portfolio decisions, not just to project the existing. For SME groups actively making portfolio choices (acquisitions, joint ventures, divestments), this is the difference between forecast-as-reporting and forecast-as-strategic-instrument.
A concrete example from SME practice: a group of four operating companies is considering the acquisition of a fifth entity with EUR 8M revenue. With Finstack’s toggle, you can run the consolidated 24-month forecast with and without the acquisition, using explicit pro-forma assumptions for synergies and integration costs. The board sees three scenarios side by side: current portfolio (base), portfolio + acquisition (upside), portfolio + acquisition with conservative synergies (base case). Without that toggle, that kind of M&A discussion happens in separate Excel models with numbers that are outdated within a week.
For NCI specifically: the Finstack model supports non-controlling interest at the right levels — minority stakes are presented correctly in the consolidated P&L and balance sheet, and the EBITDA share attributable to the parent company is explicitly distinct from the full consolidated number.
Scenarios per entity, consolidated at group level
Per-entity forecasting makes scenario planning richer. Instead of “group base case vs. group downside,” you can now run “entity A in upside, entity B in base, entity C in downside.” For SME groups with diverse business units, this produces more realistic scenarios — reality is rarely that all entities move in the same direction at the same time.
A concrete setup for an SME group of 5 entities: 15 base scenarios (3 per entity) plus 1-2 strategic group-level scenarios on top. Finstack’s consolidated scenario views let you see different combinations: “what if all entities are in their downside?”, “what if three are in base and two in downside?”, “what if the largest entity is in upside and the rest in base?”. That kind of question is practically impossible in separate Excel models per entity without an integrating layer.
The reporting output to the board becomes richer with this setup. You can now visualize: “our group EBITDA comes 60% from entity A, 25% from entity B, 15% from entity C. In the downside scenario, we see risk mostly in entity B due to specific market dynamics; entity A and C stay stable in base.” That’s a management-level discussion that’s operationally workable, instead of one aggregated group number where you have to figure out which entity makes the difference.
A second benefit: scenario iteration becomes scalable. Board asks for a specific what-if calculation? In separate Excel models that’s 1-2 working days of manual work per scenario. With consolidated scenario toggles in Finstack, it’s half an hour. For SME groups serious about preparing strategic discussions with numerical underpinning, that’s the difference between 2 scenarios per quarter and 8 scenarios per quarter.
For the broader explanation of scenarios — communication, sensitivity analysis, strategic scenarios — see the complete guide on scenario planning for SME CFOs.
What your tool needs to do for per-entity + per-cost-center forecasting
Five requirements for SME groups with multiple entities or serious cost-center structure:
1. Unlimited forecasts per entity with own drivers. Each entity must be able to maintain its own drivers, assumptions, and scenarios without one holding model being imposed. Finstack supports this as core functionality.
2. Central consolidation with IC elimination, NCI, and FX. Per-entity forecasts must automatically consolidate to group level with three elimination methods, ownership percentages for minority interest, and multi-currency with IFRS-compliant FX. Finstack has this built in.
3. Entity grouping with toggle. For what-if scenarios around acquisitions, divestments, and sub-group views: entities must be groupable, ungroupable, and toggleable. Critical for portfolio discussions.
4. Cost-center functionality at forecasting level. Not just cost-center reporting but also forecasting at cost-center level with own drivers and assumptions. Plus grouping of cost centers across entities. Finstack supports this; Speedbooks, Visionplanner, and Liquid do not, or only very limitedly.
5. Spreadsheet flexibility for the modeling itself. Because per-entity drivers differ fundamentally (manufacturing vs. sales vs. services), no pre-built tool UI can offer the specific modeling each entity needs. A tool that locks forecast architecture is by definition too narrow — you’ll fall back to spreadsheets anyway. Finstack is designed around 2-way Excel and Google Sheets sync: models live in the spreadsheet (infinite flexibility per entity), Finstack delivers actuals automatically and consolidates to group view. Multi-level: consolidated, per entity, or per cost center. From EUR 39/month per entity, 14-day free trial.
Start with one entity and validate the cycle for one quarter. Then add the remaining entities within days — Finstack’s self-service onboarding makes this feasible without consultancy. For cost-center functionality: tackle the entity with the most cost-center complexity first (healthcare/retail/services) and build the cost-center mapping there. Start with the 14-day free Finstack trial.
Explore Finstack.
Free trial for 14 days.
Access to all features. No credit card required.
.png)
Three common mistakes with multi-entity forecasting
Forcing one holding model on all entities
Drivers differ fundamentally between entities — a manufacturing entity with sales FTE as primary driver gives no meaningful forecast. What works: let each entity define its own drivers within a consistent overarching structure (same P&L categories, same reporting frequency) and consolidate the outcomes. Ownership stays with the entity controller; consolidation and strategic interpretation at holding level. It’s not a tooling choice but an organizational choice that’s supported by tooling. For SME groups new to this architecture: start with the two largest entities and then extend to the rest.
Manual Excel aggregation of separate entity workbooks
Workable with 2 entities; breaks down beyond 3. Version control breaks (which entity has the latest workbook?), IC elimination becomes error-prone, and consolidation takes days per cycle. What works: invest in central consolidation via a tool — the holding controller should spend time on strategic interpretation, not on manually summing workbooks. Plus: a shared consolidation engine for actuals and forecast prevents methodological inconsistency. Version control gets solved by central storage, and error-prone manual work shifts to structured validation.
Skipping cost-center forecasting in sectors where it’s structural
Healthcare, retail, professional services, and project organizations have serious cost-center structure. Forecasting only at entity level in these sectors misses the operational steering that actually matters. A hospital CFO doesn’t benefit from 80 department budgets if projections only happen at holding level. What works: build cost-center forecasting where operational reality demands it, with own drivers per cost center and grouping across entities for cross-entity insight. The department manager needs quarterly budget input, not aggregated holding numbers.
Frequently asked questions
Can't find your question? Let us know
Why is forecasting per entity important for an SME group?
Operational decisions are made at entity level, not at group level. A consolidated forecast tells you the total — without per-entity detail you don’t know which entity is causing the variance. For SME groups with multiple business units, per-entity forecasting with central consolidation is standard: each entity has its own drivers, holding level gets the consolidated view.
What is the difference between decentralized and centralized forecasting?
Decentralized: each entity builds its own forecast with its own drivers — the entity controller sits close to operational reality and builds more accurately than a holding controller can. Centralized: one model for the whole group — workable with 1-2 entities, loses detail beyond that. The practical combination: decentralized modeling + central consolidation via a tool.
How do forecast drivers differ between entities in the same group?
Often fundamentally. Manufacturing entity: sales volume, unit cost, machine utilization. Sales entity: pipeline conversion, deal size, sales cycle. Services entity: billable FTE, utilization rate, hourly rate. Geographic entity: currency and market dynamics. Forcing the same forecast model loses accuracy — let each entity define its own drivers within a consistent structure. Driver ownership stays with the entity controller; consolidation at group level.
What is cost-center forecasting and when do you need it?
Cost-center forecasting projects costs (and sometimes revenue) at cost-center level within an entity — per department, location, project, or team. Essential in healthcare (hospital departments), retail (locations), professional services (teams). The difference between “we’re missing EUR 300k somewhere in the group” and “cost center X in entity B is EUR 40k over budget because of Y reason”.
How do I consolidate per-entity forecasts into a group forecast?
The tool must do three things: maintain per-entity forecasts independently, automatic IC elimination between entities, and a consolidated view that aggregates current forecasts. Finstack supports this natively with three elimination methods, entity grouping with toggle for sub-group views, and consolidation that moves with every per-entity update. Without central consolidation: manual Excel aggregation, unworkable beyond 5+ entities.
Can I turn entities on and off in a scenario (for acquisition or divestment)?
Yes, with Finstack’s entity grouping with toggle. You can group, ungroup, and individually toggle entities — useful for what-if scenarios around acquisitions, divestments, or structural changes. For minority interest: Finstack supports ownership percentages and non-controlling interest (NCI) so consolidated numbers reflect the right ownership ratios.
Does multi-entity forecasting also work with multi-currency entities?
Yes. Finstack supports multi-currency multi-entity with IFRS-compliant FX methodology: fixed historical, period-end, or period-average rate configurable per scenario. Per-entity forecasts in foreign currency are automatically converted to reporting currency. For SME groups with NL/DE/UK/US entities, this is essential to stress-test currency shocks in scenarios.
What tool fits best for forecasting per entity and per cost center?
Finstack is the standard choice for SME groups with 1-30 entities: unlimited forecasts with scenarios per entity and per cost center. Native ERP connections at transaction level, entity grouping with toggle, cost-center functionality, plus 2-way Excel/Sheets sync. Multi-level: consolidated, per entity, or per cost center. From EUR 39/month per entity, 14-day free trial.
CFO turned Founder - Finstack
Sources and provenance
- finstack.io — Consolidation — per-entity consolidation, IC elimination, NCI
- finstack.io — Reporting & insights — per-entity and per-cost-center reporting
- finstack.io — Spreadsheet sync — 2-way Excel and Google Sheets integration
- finstack.io — Integrations — native ERP connections per entity
- Finstack Help Center — Sources & connections — ERP connections and sync frequency (3 hours)
- finstack.io — Pricing — from EUR 39/month per entity
Last reviewed: 19 June 2026 · Next review: September 2026
Consolidated reporting and forecasting.
