FP&A in Anaplan can become very large, very quickly.

This is especially true when we look to retain access to prior forecasts and/or large libraries of prior year data. So how can we do it without adversely impacting model performance?

Through archiving?

This is the process of isolating data which is no longer relevant to the live modelling period and removing it from the dynamic calculation or live data flow.

We find that our clients work live on only a narrow band of relevant time periods during any single forecast or budget round. However, they often need to be able to view and compare live forecast with historical data in order to make more informed decisions. This historical data sits outside of the narrow forecast or budget window and for modelling purposes is inherently irrelevant as it does not need to participate in the same dynamic process relevant to the live forecast data.

Therefore, it is a prime target for being archived.

Archiving involves flattening the target data by using a combination list to code all relevant dimensions into a list item code. Consolidated data is then imported against the combination list which can then be referenced in a target module by using aggregation functions to map the data in a target line item.

The benefits of using an archiving process for this 'irrelevant' data is that we can effectively store significant volumes of not only historical actuals, forecasts and budgets but also scenarios for use in what if analysis. This data can be accessed when required and 'pulled' into reports to be assessed against the live forecast. It removes 'sparsity' and allows the model design to utilise 'sparsity dollars' for design structures which will enhance user experience in areas such as reporting allowing final users to have a more natural experience.

The key here is 'relevance' - is the dimensionality relevant to the current forecast or budget process. Will the data move in response to system or user generated assumptions. If not, then we would recommend that an archive process is designed and the data removed from the live calculation flow.