Forecasts are always wrong, so why bother?

Background

This often comes up in debates and discussions around the world for budgeting, planning and forecasting activities.

Preceded by Excel is dead or something crazy like that! 

The finance teams forecasting is often undervalued as a result. 

The opposite is also sometimes true, that finance teams sometimes measure their own performance based on how accurate they forecast future performance, rather than how well they influence future performance. 

The use of AI/ML to do highly accurate forecasting based on data is also being promoted as the measure of a highly advanced finance team.

In my opinion these concepts ie the lack of value and the overstatement of value specific are both flawed.

George Box "all models are wrong, some are useful".

It's not the forecast, in of itself, that is valuable but the actions, scenarios and activities that it can drive that are most valuable. 

The forecast is merely an output of a tool (financial or data model). It's only 1 cycle, 1 turn of the crank. Life just isn't like that.

If you running an advanced AI/ML model be careful how much time you are going to spend working out why the model was wrong, rather than spending time on how you can influence the outcomes (which you can).

Outcomes that are influenced by humans are very difficult to predict as its based on softer skills of delivering the message and adopting the action by humans who are running the business. 

Yes, humans run the business and make strategic choices, not machines.

Choices and outcomes that no AI/ML can reasonably predict. Predictive analytics is useful, but in such high degrees of uncertainty less so.

The problem lies in how the forecast is derived that is the most challenging but also most valuable.

When do forecasts lack value?

If you are presented with a forecast the first thing you should look out for is whether it is driver-based or growth-based. 

Whether the model has a scenario manager to play with what-ifs.

When a forecast is derived from a complex algorithm based on AI/ML and nobody knows how or why it was wrong. It's a black box. 

It can be useful for a very narrow context like revenue or costs. For strategic, whole of business, modelling it lacks the data and insight needed to be valuable.

In my experience, most forecasts that I review (then rebuild for clients) do not reflect non-financial drivers that enable active decision making on value creation that enables users to influence future performance.

The forecasts that often take last period times a percentage (%) or allocation method of averages. The average and averages problem becomes real.

Average revenue of last year plus growth and average margin and cost allocations means there is multiple averages of averages.

You cannot drive performance or changes in outcomes (the real value of a model) if there are no drivers in the model and only averages.

No matter how big or complex the planning system is, without a really solid understanding of these drivers the value of the outputs will be lost.

Being able to dynamically update those drivers (changes in logic) is equally important. 

Assumption changes are a given, but business model changes require changes to logic in addition to the assumptions.

Where forecasts fail you will often find growth % and averages scattered everywhere.

The pandemic has shown that this approach simply doesn't work.

When does a forecast have value? 

For me, a forecast that shows the real cash flow impacts has significantly more value. 

Sales = Vanity,

Profit = Sanity BUT

Cash = King.

You don't survive or build a sustainable business based solely on profit, you need to understand cash and cash flow.

However, to build a really accurate cash flow forecast you really have to ensure you factor in things like working capital (debtors, creditors,  inventory) and other balance sheet items like capital expenditure, debt etc.

The best way to do this is by using a 3-way integrated financial model containing the Income Statement, Balance Sheet and Cash Flow Statement (4-way if you really want to be fancy including Sources and Uses of Funds).

The next key ingredient is applying one of the recognised standards in building that 3-way model (FAST, SMART, BPM, ICAEW etc)

The last ingredient is 3s. Sensitivities, Scenarios and Simulations.

If a forecast is driver-based, 3-way, 3s' and the model is built based on one of the standards it's incredibly valuable.

This doesn't need to be a multi-million dollar system, you can build this in Excel if you have the know-how.

How to get started?

Hope the above has challenged your thinking when it comes to building a valuable forecast that enables relative comparison to different actions and strategies not just 1 or 3 cases (low, middle, high).

Step 1. Pick a standard and start to apply it.

Step 2. Learn how to build a 3-way (financial statement) model in Excel.

Step 3. Learning how to create a scenario manager in Excel to test different what-ifs.

Use this holiday break to make a start in developing your financial modelling skills.

If you want to learn more about financial modelling read some past blogs or take a look at www.modelcitizn.com for some example models.

If you want to find out more and see the recent article series on spreadsheeting to financial modelling it can be viewed in the Financial Modelling App

If you really serious about your modelling skills then maybe 2021 is your year for levelling up?