We need to talk about Forecasting

Part of a mini-series of blogs looking at Planning

Planning starts with forecasting therefore makes sense to start this mini-series of blogs at the first stage. Much has been written about forecasting over the years, but it has been surprising to me to find that many companies still have an overreliance on “basic” methods and don’t make time and embrace learning to improve the process and forecast.?

To start with what do I mean by “basic”? Process starts with na?ve forecast which can be last year’s sales (not demand in many instances) and from there goes through several value adding / detracting activities and so on as shown below.

Forecast process:?

Appreciate I have oversimplified it but when you evaluate your process it won’t be vastly different. Technology advancements are making it more efficient with buzz words like AI and autonomous planning which are great theories, and their time will come in planning, but never forget a forecast is still a forecast and will be inherently wrong whatever certain software providers promise you!

Now we have put down a marker, we need to get to a place where planners aren’t having to justify / defend misses (and focus on short-term) but use it to empower knowledge sharing, med – long term planning, and revenue growth. Collaboration is a strong word but all roads lead to breaking silos internally and the development of ecosystem that brings suppliers tier 1 – X into the fold. Appreciate not new thinking but in the same breath still misunderstood and poorly implemented or even established in many organisations today.??

Also how explainable is the forecast and the changes that occur to inform decision making throughout the journey from x weeks out to consumption? If the planner isn’t seeing it or has the time, what chance have others and usually by the time it is spotted it’s too late to react.??If we then overlay what has happened during the pandemic, we see extremes whereby it was difficult to differentiate between real demand and pandemic related spikes leading to several winners and losers:?

• Lockdown led to demand going down / stop / move fully online / small extreme peaks then reduce to high ongoing peaks. Driven due to changes in what consumers needed aligned to the new situation they found themselves in. Lululemon and Peloton were clearly winners at this point and now facing tough market conditions due to consumers coming out of lockdown and normalised demand returning.??
• Opening of the economy released pent up demand leading in some instances to out of stocks because forecasters didn’t have valid data points and / or confidence to put inventory on the books and / or had dedicated supply points in locations that weren’t coming out of lockdown at similar times. (KFC and Nandos are examples that made the news during this time)?

In most scenarios, planners couldn’t win, and this is best reflected by matrix opposite which was adapted from Gilliland et al in their business forecasting book where 2 and 4 were faced at various times during these last two years. My main observations are that whilst this has been a very challenging time it has elevated the role and importance of planning.

Therefore, as we emerge from hopefully the last significant lockdown it is a good time to reflect and to set a new path for forecasting. Below I have a few suggestions which you’ll be glad to read aren’t moon shots and this has come from discussions and engagements with clients and academics on the topic:

• Get acquainted with forecast value-added analysis and use it – this is where you start to measure the na?ve forecast and how adjustments to the forecast over time add / remove value to better inform future forecasts.
• Technology is yesterday’s knowledge as Drucker use to say so what is the knowledge that you can feed the forecast to improve? Reassess the data sets you are using to inform forecasts and what value are they adding remembering that products / categories react differently, and that more data is not always better. (diminishing returns)
• Take snapshots of forecast accuracy at different intervals from x weeks out to consumption – forecast will have more confidence nearer to consumption but how good is accuracy compared to week 4, week 8, week 12 and time periods that align to your supply chain requirements. (all are different) Ask yourself continuously are there any significant changes being driven by over corrective behaviour in the system / planners and what are the ramifications, get into a feedback loop, learning helps inform and improve the future.?
• Work with Procurement to identify those outliers that are causing you most challenges and rather than talk about them, get action plans in place, and execute to put in place additional capacity or alternative supply. (nearshore is coming back in fashion you may have heard)?
• Products and categories are different so understand which create most noise in the Supply Base and not just the ones Commercial and Marketing are fixated on.?
• KPIs are helpful but are you refreshing what you are measuring and keeping them relevant to explain business performance? No point if they are always green or red and nothing changes, they must drive change that leads to performance improvement.
• Frequent Supply Base engagement which is dedicated to forecasting and not part of overall discussions – get into what went well and what didn’t through forecast process and make it a two-way conversation since collaborating breaks down barriers and helps to minimise bullwhip effect. (if Supply Base not informed how can they help you?)
• Have you looked at the tools you are using to manage the process including engagement in the S&Op cycle and are they still fit for purpose? There are a number of tools available in this space which can improve the process and ability to collaborate and can easily plug into your current technology stack. (also doesn't have to be painful long implementations now!)

In conclusion, forecasts are inherently wrong and will always be therefore the goal should be for a planner to have a strong process in place where learning from what has happened better informs the future and therefore a better forecast. Building a collaborative view shouldn’t just be words but have all relevant parties energised and aligned to one set of numbers and plan. There aren’t any silver bullets sadly but there is plenty of enjoyment and satisfaction to be had especially with the advancement of data and getting into the machine learning and artificial intelligence realms which I will touch on in a future blog to advance the role and importance to the business of the planning function. ?

Thanks for stopping by and reading, don’t forget to share any suggestions you may have and let me know if you liked reading this short blog.