What did we discover are the key reasons Anaplan for FP&A projects fail?

We have carried out quite a few Anaplan health checks, reviews and assessments since launching in December 2021. Where the initial launch was less than stellar models shared the following common traits;

➡Unstructured model architecture - data, inputs, calculations and outcomes lacked a coherent structure which made these models difficult to audit and parse modelling logic.
➡Poorly designed data model - lists and hierarchies were overextended and too long. Attempting to satisfy all with a single structure actually caused these models to become too large and cumbersome.
➡Lack of model hygiene - naming and indexing standards were poor and the models simply looked a mess. On all occasions this felt like the developer either lacked motivation, was just lazy or simply didn't appreciate the impact of poor standards.
➡ Overly complex modelling - simple modelling challenges had been overengineered, utilising exotic formulas and building elaborate staging processes. Mostly due to lack of experience or challenge on the part of the consultant, calculation steps contained irrelevant or redundant elements and were often squeezed into as little steps as possible under the assumption that this is more efficient.

While some models will never recover from such poor implementation we have found that others, with some targeted remedial action can ensure the rocket ship doesn't come crashing back down to the ground.

✅ Parallel hierarchies - we have found that rebuilding key model structures which are too long and breaking them up into multiple shallower parallel hierarchies significantly enhances the agility of the modelling.
✅ Flat data imports - we have discovered that using flat data structures to house your external data and using aggregation functions to make use of it will add significant flexibility in how you can maximise the use of your data across the model.
✅ Tidying up naming - predominantly cosmetic we have learnt that good model hygiene not only enhances the modelers user experience it also significantly improves development efficiency as navigating the model is simplified.
✅ Simplify calculation logic - where changes in dimensionality occurs we have discovered that it is more efficient to break out the calculation over more stages allowing us to remove elaborate formulas and therefore significantly reduce complexity.