Who Forecasts?

The final article in a set of 6 entry-level discussions about the Whys & Wherefores of Forecasting. As with the previous articles the focus is really on Demand Planning, but I like to think that the principles are generic across business functions.

Who Creates the Forecast?

The clear answer, of course, is that everyone forecasts and within Business there can be dozens, nay thousands of forecasts created by all sorts of departments and individuals. Excel is both a boon and a bane as predictions proliferate everywhere! The maturity of your planning, the ownership bias your plans contain and the strategies to improve are discussed in the article Why Forecast? but it is worth delving more specifically into who creates forecasts.

It could be someone's primary job, or it might be a side-task. It could be a team in one department or perhaps delegates from a group of departments. The more involved, the longer it takes, though perhaps collaboration produces forecasts with less bias. Forecast creation can be organised into 5 types:

  1. Individual: Each person builds and uses their own forecast
  2. Independent: Each Department creates and uses its own forecast
  3. Concentrated: One Department creates the forecast, and all others use it
  4. Negotiated: Each Department creates a forecast and then representatives gather to create one agreed version
  5. Consensus: A Planning Committee is formed with individuals from each Department who create the forecast together.

The right creation approach depends on the purpose of the forecast, the skills of the people involved, the business culture and the desired speed of output. How many Forecasts do you have? How are they created? Who owns them? Who uses them? What Bias do they carry? What is the impact of Allocation and Aggregation? Should they inter-connect? Should you try and create a single source of truth Forecast? Answer these questions and as Yoda might say "be close to understanding, you will".

Biodegradable Hammocks?

Let's take a mile-high look at the flow of forecast information through an imaginary company called Fictitiously Swing Ltd who make biodegradable hammocks. Their hammocks have taken the world by storm! Sales are rocketing and the company has recently expanded into biodegradable garden furniture, fences and trellises. Product Development are urgently working on new lines of biodegradable masks, bags, clothes and pet toys.

No alt text provided for this image

The Fictitiously Swing Annual Plan is a once-a-year activity at the highest levels of Product Brand, Region & Month (but then rolled up to Quarter) for the next two years. It only includes value and is used to inform Company Strategy and support the creation of the Budget. It is created on a whiteboard, transferred to PowerPoint and owned by the Company Chief Officers. They are very fond of it and trust their instincts.

The Budget is a once-a-year activity that applies the Annual Plan numbers to Product Category, Country and Month. This Plan is built in Excel by the heads of all the key departments (actually, it is built by a few Excel gurus who work tirelessly to allocate and aggregate the values back to the Annual Plan numbers). The Budget is value based and is ultimately owned by Finance. The Plan is monitored and reported against on a regular basis. How are we compared to Budget?

The Sales Forecast is a monthly activity by each Salesperson in Excel at Product Group, Channel, Territory and Customers levels. Discussions with Customers, Product Development, Marketing and Finance over Promotions, Pricing Catalogues and Marketing activities and refine the plan. The Sales Forecast is value based and compiled from dozens of individual Sales Spreadsheets on a monthly basis by the Sales Admin team. There is no need to make the Sales Forecast match to Budget (aside from the first month when the Annual Budget is released) though the two Forecasts are compared on a regular basis.

The Demand Forecast is created once a week at the level of Organisation, Item, Customer and Weeks by a team in Operations. One a month the planners manually convert the latest Sales Forecast to volume and import it into their spreadsheets along with the latest 12 Months Sales History which is then lagged by 52 weeks. The Planners compare the Sales Forecast and the lagged Historical Data and using insight from meetings with Sales & Marketing and Customer calls, merge and adjust the numbers into the Demand Forecast. The Spreadsheets are compiled to create a Final Demand Forecast which is reviewed, approved and owned by Sales.

The Supply Plan is created daily at the Organisation, Item Component and Day level using the weekly Final Demand Forecast alongside Daily extracts of Open Orders, Inventory & Incoming Supply. The Supply Plan horizon is 3 Months, is owned by Manufacturing, is used to create the Manufacturing Resource Plan and a daily Production Schedule.

The Finance Plan is compiled once a Month from the most recent Final Demand Plan at the level of Organisation, Item, Customer and Month and with additional feeds from Business Intelligence so that volumes are converted back into values for Revenue, Cost and Margin Plans and comparison to the Sales and Budget Forecasts.

No alt text provided for this image

Who wills the end wills the means

In the Fictitiously Swing Ltd scenario above everybody is forecasting from the CEO to administrative staff. Multiple versions of the various forecasts are in use at the same time. They interconnect (sometimes years apart) and the plans have different sources, destinations, levels of granularity, horizons, functions and owners. The integration between them is manual and the effort of allocating and aggregating the data is considerable and made particularly difficult when having to make the numbers add up to the Budget and Annual Plans.

Master Data impact is everywhere and controlling it is a permanent nightmare. Excel guru's with imports, formulas, lookups and advanced techniques are at a premium and the spreadsheets built by those experts (many who are no longer with the company) are stressful security risks and can impact cycle times when they error.

The Forecasts are often running at different intervals. The Sales Forecast is refined and ready once a month and is used by the Demand Planners as a baseline (it has to be since Sales 'own' the Demand Forecast) but within a few weeks the Demand Forecast is based on old and inaccurate data. In a long month and with changeable Sales & Marketing tactics the Demand Plan can be actively wrong.

Trust degrades quickly and there are high levels of stress for "the latest forecast". The Annual Budget month is particularly high profiled with long weeks of re-work to get it "just right" even though it is a losing battle. Does any of this sound familiar?

Centralised or De-Centralised?

A decentralised planning structure where Sales are dealing with customers locally and need to convert their forecasts quickly and efficiently requires a local planning team but planning expertise in one location with Engine Tuners and Data Scientists communicating quickly & sharing segments of data also makes a great deal of sense. Both options can exist together.

The pandemic world has accelerated the use of cloud systems and remote working to the extent that de-centralised or location planning is perhaps less of a problematic or contentious issue. However, beware situations where forecast ownership changes over local and central boundaries as conflict almost inevitably follows.

Product or Customer Focussed

Another subject worthy of consideration is how to align the forecasting resources. Should they be organised by Product? or Customer? or Manufacturing Plant? Territory? These and many other ways of slicing the data are probably more cultural but it is worth reviewing your product set and changing things around - especially if accuracy is not improving. Ask your planner how they would do it if they could choose! You might get some interesting insight.

Instead of traditional Product or Channel planners, how about grouping your forecast team by the forecasting suitability - resources who look after sets data according to how difficult they are to predict. Those that are easy can be managed in larger volume by fewer people while the more complex and perhaps, business critical are managed by more senior planners who have fewer combinations to manage.

Scientist or Artist?

The skillset of planners varies enormously. It depends on a host of factors from the System being used and the teamwork method to the complexity of the dataset, the horizon, predictability and the business knowledge required. If you run a complex Statistical Tool it might be that you need Data Scientists and Advanced Mathematical skills to tune your engine, or it may be that Marketing knowledge or methodical spreadsheet management is more important.

Who's getting better?

Whether building a solution from scratch or refining existing people and processed you need to create and follow a strategy to improve. Quality forecast generation is more than a PHD in box, it requires the right people empowered to do the right things with the right data in the right places at the right time.

And the people forecasting should not stand still. As the world changes, so the people predicting should adapt and evolve with new skills, datasets, systems, processes, measurements and objectives. You can't refine a forecast without also refining the people who make it.

Your Planning Strategy

Now that you are considering optimising your forecasts:

  1. Understand existing Maturity, Ownership and Bias.
  2. Strategise options from 'one number forecast' to multi solution
  3. Select the right Strategy and define resources and ownership
  4. Engage, Communicate, Harmonise and get a better future going!
No alt text provided for this image