Forecasting

It is practically impossible to know what is going to happen tomorrow. As much as the future is unknown one can make a reasonable assumption that it would be influenced by what is happening today. Conditions are always changing, uncertain, volatile and unplanned eventualities are more likely than anything else.

 

While one of the most powerful ideas is that decisions in themselves create certainty – there is also an important value to forecasting. Forecasting is a technique that we use consciously or unconsciously to predict what will happen here and what the likelihood is of specific events. Forecasting weighs the risk, with an aim to model the overall impact of decisions and known eventualities in a specific role of function. As such forecasts form the basis of sales planning, production and inventory planning, manpower planning, financial planning and budgeting, research and development planning, acquisition and merger analysis and strategic planning.

 

With something so important and so fundamental to understand if a business will survive and be profitable it would be expected that there are brilliant techniques and systems by which we can predict the future. It would be great if there were a big data repository in which everything that is happening in the world goes into a central feed and I could pull up a prediction of how many people would like product A vs product B in market X. If I then decide to put more money into advertising – it should be clear what comes out as net profit, while taking into consideration that suppliers will offer me prices at different levels. The big data machine should also take into consideration that labour is unstable in this market and that the government is busy making policy that will result in the financial markets going up or down. If all the complexity can be analysed and built into a micro model for my decisions it would be lovely. Some has explored this idea and some micro-modelling techniques based on advanced artificial intelligence have this level of complexity built in. It is not something that is generally done easily on your version of Excel.

 

Unfortunately most data and analytical models are just not advanced enough to bring this type of modelling into practical use and the executive or manager is expected to be aware of the macro and micro factors that drive the business model of the enterprise and to be able to predict what will happen to the business given a specific set of variables.

 

The millions of bits of information that flow in from all sources add to an understanding of what will happen and influences either intuitively or quantitatively the direction of decision making. This information shapes the perception of management of what to do next and may fundamentally shape the outlook of the business.

 

So how do we practically forecast?

 

Step 1: Decide the time period of forecasting

 

A forecast should be made for a period that is practical. Lets say that you are forecasting something that changes often – then at best you will be able to make a short term forecast that has some level of accuracy. Updating such a forecast more regularly is important as it is likely to change often.

 

So you have to decide on what period this forecast is for and how regularly it will be updated.

 

Step 2: Break up the period of forecasting into shorter time windows.

 

If you are forecasting for a year you may break your forecast up into shorter periods of months. If you are forecasting for a month, then you break it up into days.

 

It may be useful to break up a year into days if you understand your process well but it may require you to have more variables.

 

Lets build a spreadsheet that has as its columns each one of these periods. Column 1 would thus look at period 1, column 2 at period 2 etc.

 

Step 3: Map the inputs, process and outputs

 

If you understand the process or business model of a specific flow in your business then you can look at the inputs, the process that they need to go through and the outputs.

 

These become the rows of your spreadsheet.

 

So the idea is that you look at how many people there are, if there are those many people, how much can the sell, if they sell that much, how much is produced, if we produce that much, how much do we have to order, if we order that much, how long is the delays, if the delays are that long then how much do we have in inventory. If we have that much in inventory, how much can we sell? How does impact what we said before about the number of people that we have?

 

So your spreadsheet would have inputs (people, inventory), processes (sales, ordering, delays), outputs (inventory levels, sales, profits)

 

  Time 1 Time 2 Time 3 Time 4  
Input A 1 2 3 4  
Input B 3 3 3 3  
Input C 1.1 1.2 1.3 1.4  
           
Process A A + B A + B A + B A + B  
Process B ( Process A ) * c / 5 ( Process A ) * c / 5 ( Process A ) * c / 5 ( Process A ) * c / 5  
Process C Process B * 1.1 Process B * 1.1 Process B * 1.1 Process B * 1.1  
Result D A + B – C A + B – C + Result D (Time 1) A + B – C + Result D (Time 2) A + B – C + Result D (Time 1)  
           
Result A 4 5 6 7  
Result B 1.1 /5 = 0.22 2.4 /5 = 0.48 3.9 /5 = 0.78 5.6 /5 = 1.12  
Result C 0.242 0.528 0.858 1.232  
Result D 3.978 8.93 14.852 21.74  

 

Step 4: Building the micro model

 

The micro model is how you get to the results and is usually captured in the equations you put into your spreadsheet. It is the tools by which you map out the links between parts of the process.

 

How the processes impact each other can become a matter of some voodoo and experimentation. Sometimes it is as easy as deciding if it is a plus or a minus and if we times it by the inflation rate or not.

 

Sometimes it is more complicated and requires statistical analysis or investigation into the existence of theoretical model in this knowledge area that explains how something works. It may also be possible to use your experience or a clear process definition to get to the equation that drives it all.

 

To give an example we all know that Sales Price = Cost price x mark-up but your taxes may be different, and you may have commissions and other factors that goes into your cost price and a million other factors that goes into mark-up. You only have to build the micro model to the level of complexity that is required. It can be as simple as putting in the cost price and the mark-up into your spreadsheet or to build complex models on how to get to these numbers.

 

Step 5: Testing the micro model

 

The test is if these numbers are observed in the real world. If your forecast works for historical inputs and is robust enough to cater for what happened in the past – it is usually a good indicator for the future. A lot of people forget to model their micro models on the past and then find themselves trying to explain post the even why things went wrong and they blame their model. This is largely where forecasting fails.

 

If no past data exist it may be beneficial to do parameter tests. What if a parameter was this, what is the outcome? This type of testing, although possibly expensive may refine forecasting models and give a very clear indication of what will happen, given specific conditions.

 

Step 5: From model to execution

 

A forecast is only useful it if gets implemented. It is recommended that forecasts are shared, discussed and regularly updated. Some successful sales organisations forecast weekly and production businesses automate the process to the level where they can forecast daily, weekly and further ahead.

 

Once a forecast is established and proven it becomes a tool for monitoring variance. If the actual is compared to the forecast, the variance starts explaining deviations that occurred. By understanding re-occurrences it allows us to either eliminate them, or to build them into the forecast and understand their impact on the bottom line and top-line.

 

Step 6: Breaking the mould

 

Forecasts as a tool aids in perfecting the outcome of a process – but it also needs to become a source of innovation. When we know what we budgeted and what actually happened, it allows us to look for innovation and improvement. It may require us to relook the business model fundamentally and to come up with interesting and new ways to do things. This process of breaking the mould should translate into new measures and disassembling old measures. So forecasts while useful as a tool to know what the future holds, must also be a tool to keep the business competitive and on the path to innovation.

 

Conclusion

 

Whether you believe in intuition or forecasting, there is a constant process of weighing that happens in predicting the future. A view on the future is necessary to allow businesses to embed strategy and achieve prosperity and sustainability. Being better at forecasting and making it a daily discipline will help in driving a view of the future that is practical and that is built on innovation.