FP&A & Finance Analytics

For many finance teams, the real work only starts after the month-end close. Figures are exported from the ERP and accounting, merged and commented on by hand. That takes time, and in the end someone asks why the report differs from the ERP. Planning meanwhile runs in files that circulate by email, and the forecast is updated once or twice a year and is often out of date by the time it arrives.

We build finance analytics solutions on one shared data foundation. Actuals from the ERP, plan values and forecasts sit in the same model, metrics are defined once, and finance gains time for analysis instead of compiling numbers. Internationally this field is called FP&A, financial planning and analysis. General reporting automation and forecasting models are only touched on here. This page is about the finance side: chart of accounts, cost accounting and the planning process.

A finance data model for P&L, contribution margins and cost centers

The foundation is a clean model. Postings from the general ledger are mapped to report lines through an assignment of the chart of accounts, so the P&L structure is defined once, centrally, and not rebuilt in every report. If the layout changes, it is adjusted in one place.

Contribution margins need a clear definition for each level. Which costs count as variable, what is assigned to products, customers or regions, and what stays as fixed cost at a higher level? We settle this together with finance, before the first dashboard exists. Cost centers are modeled with their hierarchy, and row-level security makes sure that those responsible see only their own area.

Plan-vs-actual comparisons require versions: budget, forecasts and prior year are held as separate plan states that can be set side by side. From a variance, a drill-down leads through account and cost center to the individual posting. That only works if plan and actual use the same framework of accounts, cost centers and periods. If they are two separate worlds, the comparison fails at the mapping.

Numbers that can be checked

We plan traceability in from the start. Totals in the report are checked automatically against balances from the ERP, so differences show up immediately. Every report shows which data state and plan version it is based on. Entries in planning are logged with person and time, and who may view, change or approve figures is set technically. What requirements from internal audit, tax law and external audit apply is for the responsible people to clarify. We make sure the reports are technically traceable.

The close can also be sped up. Most of the time goes into collecting data, reconciling and writing commentary. Automated loads, plausibility checks before the cut-off date and reports on preliminary figures shorten exactly those steps.

Planning, forecasting and drivers

A rolling forecast extends the planning horizon on an ongoing basis, for example so that the next twelve months are always planned. The outlook stays current without a full budget being built each time. How often it is updated depends on the business: monthly, quarterly or as needed.

In driver-based planning, you do not estimate results but their causes: sales volume, price, headcount, utilization, exchange rate. The P&L follows by calculation. Assumptions become visible and open to discussion, and scenarios such as “sales minus ten percent” can be run quickly. Not every line needs drivers. Where they cannot sensibly be determined, direct planning is the better choice, and the finer the planning, the greater the effort. The right level of detail is the one that actually changes decisions.

Planning is also a process with many participants: cost center owners enter values, finance consolidates, management signs off. That calls for versions, deadlines and approvals in the system rather than in emails. Controllers like working in Excel, and solutions that ignore that get bypassed. It is better to offer a connection that allows Excel as the interface while the data stays central.

ERP and planning tool

Actuals come from the ERP or accounting system. From SAP, the general ledger, cost center and profitability accounting and master data can be loaded, and the effort mostly lies in understanding the structures. In small and mid-sized companies DATEV is often the leading accounting system, with data available through interfaces or exports depending on the setup. Microsoft Dynamics 365 with Finance and Business Central brings its own interfaces and fits well in Microsoft environments.

For planning and reporting there are several routes. For reports and plan-vs-actual comparisons, Power BI is usually enough. Entering values directly in a report is possible in Fabric through translytical task flows, which write the entries to a database via data functions, and third-party planning extensions exist as well. With Planning in Fabric IQ, Microsoft has introduced a planning feature of its own, in which budgets, forecasts and scenarios are built directly on Power BI semantic models.

In SAP environments, SAP Analytics Cloud is the obvious candidate, combining reporting and planning in one product and including automatic forecasts. Specialized planning platforms offer multidimensional models, workflows and scenarios. The choice depends on the complexity of the planning, the number of planners, how close to Excel it should be, the ERP connection and the licensing model. For manageable planning needs, a well-built model with an input option is often enough.

Where AI helps in finance

AI can take work off finance but does not replace judgment. Statistical and machine learning methods provide a starting value for the forecast, which finance reviews and adjusts. The technical side of those models is a topic of its own. For variances, AI can draft a commentary from the numbers, for instance which accounts and cost centers drive them. The background that the data does not know, such as a one-off effect or a contract change, is added by finance. Unusual postings, duplicates or outliers in cost centers can be flagged, as a hint and not a verdict.

Calculation happens in the data model and not in the language model, because totals and variances have to be reproducible. The AI explains and summarizes, a person reviews and approves, and the source of every statement stays visible. A fluent but wrong commentary does more harm than a missing one. Financial data is also sensitive. Where it is processed for AI features, we clarify that with data protection and legal counsel.