Anyone who uses IBM Planning Analytics knows the pattern: until now, forecasting was done either with custom-built rules in TM1, with external tools – or not at all, because the effort was too high. IBM has changed that.
With version 2.1.24 of Planning Analytics for Excel (PAfE), or v3.1.11+ in the newer architecture, Univariate Forecasting is integrated directly into Universal Reports. Xavier Osorio, Senior Product Manager for Planning Analytics at IBM, calls it “one of the most significant enhancements to PAfE to date.” And for good reason: previously, forecasting was available exclusively in Planning Analytics Workspace. Now users can create forecasts directly in Excel – and write the results back to TM1 without any detours.
This article explains what changes in concrete terms, which controlling tasks the feature is suited for – and where it reaches its limits.
What does Univariate Forecasting mean – and what doesn’t it?
IBM Planning Analytics for Excel (Definition): PAfE is the Excel add-in for IBM Planning Analytics and TM1. It gives you direct access to TM1 cubes from within the Excel environment. As of version 2.1.24, Univariate Forecasting is embedded directly in Universal Reports – without switching to an external tool.
Univariate Forecasting analyzes a single time series and derives future values from it. The input is always one variable, for example monthly revenue over the last three years. External factors such as marketing spend, raw material prices or seasonal campaigns are not taken into account.
That sounds like a limitation. In controlling practice, it is often exactly the right approach. Many planning figures can be estimated reliably from their own history: overhead costs, personnel expenses, infrastructure costs, stable customer revenue.
Multivariate Forecasting: IBM has already announced that multivariate forecasting will follow in a later release. The first version is deliberately focused – quick to deploy, without complex configuration.
What exactly has changed in PAfE v2.1.24?
The key difference from the previous situation: forecasting in IBM Planning Analytics was available exclusively in Planning Analytics Workspace. If you used PAfE as your primary working environment, you had to rely on workarounds: custom TM1 rules, external statistics tools, or switching to a different interface.
That changes now. The new feature is embedded directly in Universal Reports. Specifically, this means:
- You continue to work in your usual Excel environment with PAfE.
- The forecasting feature reads the time series from your existing TM1 cube.
- The forecast is output as an additional column or row directly in the Universal Report.
- The results can be written back to TM1 directly, without leaving Excel.
- No switching to PA Workspace, no external workflow, no manual model reconciliation.
Prerequisite: PAfE v2.1.24 or higher (classic architecture) or PAfE v3.1.11+ (new architecture). You can find the version in the PAfE menu under Help » About.
| Criteria | Previously (before v2.1.24) | NEW: PAfE Univariate Forecasting |
|---|---|---|
| Availability | Only in PA Workspace | Directly in PAfE / Universal Reports |
| Setup | TM1 rules or external tool | No setup, ready to use immediately |
| Model selection | Manual / expert decision | Automatic by IBM |
| Writing results | Manual export / workaround | Directly back to TM1 |
| Setup time | Hours to days | Minutes |
| License | Possibly additional (PA Workspace) | Included in PAfE |
Are you currently planning a new budget round or want to improve your forecast quality?
We will show you how the feature fits into your existing TM1 structure – concretely and without detours.
Which controlling tasks is the feature suited for?
The feature delivers the most value where time series are stable and not heavily dependent on external shocks. Specific use cases:
1. Revenue forecast for stable product lines or existing customers – monthly extrapolation from historical actuals, directly in the forecast report.
2. Cost trend planning – energy costs, rent, IT licenses: variables that can be predicted from their own dynamics.
3. FTE and headcount planning – estimate staffing trends across quarters when no major restructuring is planned.
4. Initial budget draft – instead of starting from the prior-year view, the feature provides a data-based starting point.
5. Variance analysis – compare forecast vs. actual over the reporting period, directly in the same Universal Report.
How do you start the forecast in PAfE? Step by step
The feature is integrated directly into Universal Reports. In practice, the process takes just a few minutes – as IBM demonstrates in its official video documentation:
1. Select the forecasting month: In the Universal Report, switch to the month from which the forecast should start.
2. Select a row: Highlight the row you want to forecast. Multiple rows are possible; one is enough to get started.
3. Start the forecast: Right-click a cell in the highlighted row → IBM Planning Analytics menu → “Launch Forecast.”
4. Define the time period (Panel 2): Select the time dimension of your cube. If the report contains non-time dimensions (e.g., Version), remove them. Then specify which historical periods the model analyzes – and which future periods the forecast is written to.
5. Check the forecast context (Panel 3): The Universal Report passes on the row and slicer context automatically. In most cases, you can leave these settings unchanged.
6. Set the target version (Panel 4 – Advanced Settings): Select the version of the cube to which the forecast values should be written. The range of options depends on your IBM user role: admins see all settings, analysts a reduced selection.
7. Preview and write: “Preview” shows a graphical representation of the forecast. “Forecast” writes the values to TM1 – the report updates automatically.
When does the feature reach its limits?
Univariate models are not a cure-all. There are situations in which they systematically fail:
- External shocks: When revenue depends on advertising pressure, price increases or market changes, the model lacks that information.
- Discontinuities: Acquisitions, product discontinuations, restructurings – the time series breaks, and the model cannot anticipate the break.
- Short histories: If you have fewer than 18 to 24 data points, you give the algorithm too little to learn from. The results become unreliable.
- Highly seasonal businesses with promotion effects: When promotional periods create patterns that the model does not know from the past, it underestimates peaks.
In these cases, multivariate modeling – with Watson Studio or specialized forecasting platforms – remains the better route. These are selective exceptions, though, not the rule. IBM itself plans to deliver multivariate models in a later release.
Why the right partner makes the difference
The feature is technically easy to activate. The real work comes before that: are your cubes structured so that forecasting delivers meaningful results? Are the time series right in terms of granularity and length? Does the forecast fit your planning logic? And how do you prevent automatically generated forecasts from flowing into planning without a plausibility check?
BI2run has been supporting IBM Planning Analytics and TM1 for years – from setup and optimization to integrating new features like this one. We will tell you honestly whether the update makes sense for you and what it concretely delivers. If the check shows that you have no need to act, we will tell you that, too.
Glossary
| Term | Definition |
|---|---|
| IBM Planning Analytics | IBM’s business intelligence and planning platform, built on the TM1 database engine. Used for integrated enterprise planning, forecasting and reporting. |
| PAfE (Planning Analytics for Excel) | Excel add-in that serves as the front end to IBM Planning Analytics / TM1. Enables direct access to TM1 cubes from the Excel environment. |
| Universal Reports | Reporting framework within PAfE that enables flexible, rule-based reports with a live database connection. As of v2.1.24, the basis for Univariate Forecasting. |
| Univariate Forecasting | Time series forecasting based on a single variable. The model learns from the historical trend of that one variable and extrapolates into the future. |
| Extended Forecasting Engine | Advanced forecasting engine introduced with PAfE v2.1.24 / v3.1.11+. Whether the univariate feature is based on it is still an open question in the IBM community. |
| TM1 | Database engine behind IBM Planning Analytics. Known for in-memory processing and TurboIntegrator processes. Also marketed as IBM TM1. |
| PA Workspace | Web-based interface of IBM Planning Analytics. Until PAfE v2.1.24, it was the only way to use forecasting in IBM PA. |
Frequently asked questions
What is the difference between univariate and multivariate forecasting?
Univariate forecasting uses only a time series’ own history as its basis. Multivariate models draw on several variables at the same time – for example, revenue as a function of marketing budget and market share. Multivariate models are more powerful, but require more effort in data maintenance and modeling. IBM plans to deliver multivariate forecasting in a later PAfE release.
Which PAfE version do I need for the new feature?
The classic PAfE architecture requires version 2.1.24. If you use the new architecture, you need PAfE v3.1.11 or higher. You can check your version in PAfE under Help » About.
Can I save the forecast results directly in TM1?
Yes – that is one of the key new features. You can write the forecast results back to TM1 directly, without leaving Excel. However, you should test the interaction with cell security, security overlays and element locks beforehand, because IBM has not yet fully documented these points.
Can I configure the forecasting model myself?
Not in the current version. IBM automatically selects the algorithm with the best goodness of fit for your data. That is sufficient for most controlling tasks and saves configuration effort. If you need full control over the model parameters, stay with Watson Studio or specialized tools.
Do I need additional licenses for the feature?
No. The Univariate Forecasting feature is part of PAfE from v2.1.24 and requires no additional IBM license. The prerequisite is an active IBM Planning Analytics license with PAfE access.
How long does it take to create the first forecast in PAfE?
In practice, a few minutes – provided your Universal Report is built on a TM1 cube with enough historical data. The process: select a row, click “Launch Forecast,” define the time period and target version, check the preview, write the values. If you are working with the feature for the first time, plan 15–30 minutes to get to know the panels and understand the cube structure.
Why was forecasting limited to PA Workspace until now?
IBM has expanded forecasting step by step. The initial availability in PA Workspace made it possible to stabilize the feature in a controlled environment. With v2.1.24, the step into the more widely used PAfE environment now follows, which IBM itself considers a significant enhancement.
