Developing a Solid Forecasting Strategy

What is your forecasting strategy?  Chances are you’ve never been asked that question and the answer is probably not on the tip of your tongue.  Don’t worry.  You’re not alone.  In the rush to implement a workforce scheduling system, few retailers give much thought to forecasting unless (or until) something goes wrong.  Even fewer organize those thoughts into a formal strategy.

Don’t get me wrong, any retailer who uses a workforce management system or even manually plans labor has a forecasting strategy.  However, that strategy may not be something that is articulated.  I’ll get into the benefits of articulating your forecasting strategy in a minute.  For now, let’s talk about what a forecasting strategy is.

A forecasting strategy describes a company’s approach to predicting business demand at its stores as part of its weekly scheduling process.  There are three big parts to a forecasting strategy.

What are you Forecasting?

The first part describes what you’re forecasting.  This part focuses on the business activities that drive labor in your stores.  For example, a women’s apparel department may be the number of items sold or foot traffic. Or a bakery labor may be driven by the number of orders placed. Not surprisingly, these business activities are typically called “drivers”.  However, some workforce management systems use unique terminology.  RedPrairie, for example, calls them “metrics”.  Similarly, StaffWorks refers to them as “business indicators”.

Beyond simply identifying the drivers, you’ll need to understand how detailed your forecasts will need to be.  For example, do you need to forecast to the 15-minute level or can you forecast to the 30-minute level?  Are you going to forecast individual items or can you forecast product groups?

If you have well modeled labor standards, they will explicitly tell you what drivers you should be using and can help you figure out how detailed your forecasts should be.  If you don’t, a little self-reflection on the nature of the work in your stores can get you started.

How are you Forecasting?

The second part describes how you’re forecasting.  At the highest level, this is the forecasting method or algorithm being used to forecast each driver.

To find the right forecasting method, you’ll need to pull some sample data – data that represents a reasonable cross section of your stores – for each format or store type to test various forecasting methods to determine which method most accurately forecasts each of your drivers.

This leads us into five important topics that will affect how you forecast:

1. Forecast Accuracy.  There are several aspects of forecasting accuracy that need to be considered when developing your forecasting strategy.  You’ll first need a measure of – accuracy, such as Mean Absolute Percentage Error (MAPE) to compare the forecast results back to actuals.  You’ll also need to understand how accurate your forecasts need to be.  Finally, you’ll likely need to find ways to improve the accuracy of your forecasts.
2. Limitations of your WFM Systems.  If you already have a workforce management system, you’ll need to work within the limits of the system.   I know it sounds basic but it bears reminding you to focus on the specific algorithms and capabilities your WFM system offers first.  Only after you determine that you cannot achieve your required level of accuracy would you want to consider a custom or external forecasting process.
3. Data Sources and Data Quality.  Forecasting algorithms require historic data…often lots of historic data.  You need to know where to find data for each of the drivers in your corporate and store systems.  In some cases, you may have more than one source of data, and you may have to make a trade-off between timeliness and quality of data.  Having a reliable, consistent source of clean data is paramount to producing good forecasts.
4. Special Days.  How do sales, promotions or holidays affect labor in your stores?  How does road construction or severe weather affect your business?  Your forecasting strategy should consider these special days to ensure that semi-regular or unique events are properly accounted for in your forecasts.
5. Impact of the Budget and/or other Company Forecasts.  You’ll also want to consider how you reconcile your WFM system generated forecast with the budget or other forecasts created by finance. Even in situations where the finance-provided budget or forecast is gospel, you’ll need to reconcile this with the more detailed WFM forecast required to accurately forecast workload and generate a schedule.

As you can see, this part of your forecasting strategy is more than just picking a forecasting method, and you’ll spend a lot of time iterating through variations of your forecasting strategy while analyzing the results of many tests.

Who is Creating the Forecast?

In the not too distant past, store managers hand wrote schedules.  To write those schedules, the store manager used his experience and guidance by finance to decide when and where to place his people and shifts.

Even when client-server scheduling systems were introduced in the 1990s, store managers would generate their own forecasts as part of the scheduling process.

However, things changed when centrally-installed, web-based scheduling systems were introduced around 2000.  Retailers had the ability to generate forecasts centrally, and many retailers took control of the forecasts out of the hands of the managers.  Some gave managers the ability to edit the schedules but most simply started giving the manager his weekly forecast.

While this shift made up for a lot of mistakes that managers made when they had total control of the forecast, retailers also lost critical input as to local happenings and the individual aspects of a retail store that local store leadership knows.

So, the third part describes who is involved in the forecasting process.  Namely, how involved is store leadership in establishing the forecast?  It also describes the role that field/district management and corporate users play in the process.

Documenting your Forecasting Strategy

As I said earlier, most retailers struggle to articulate what their forecasting strategy is, much less have it documented, but there are some very good reasons to write your strategy down.  Let me give you four of them:

1. Obtain Buy-in.  Documenting your forecasting strategy explains the decisions that went into the system configuration.  This will be critical to get buy-in from finance and other stakeholders that may feel that “forecasting” is their domain.
2. Document Decisions.  It lets you (or the person that takes over after you’ve moved on to your next opportunity) remember what you did and why you did it.  Workforce management is complex and it is easy to forget the good reasons why something was done when it comes time to evaluate if it needs to change.
3. Define Requirements.  During implementation, a documented forecasting strategy helps establish business requirements written in business terms.  The good news is that any documentation can be pretty simple.  It doesn’t have to be a long, verbose document.  You just need to write it down.
4. Increase Adoption.  Documentation creates transparency which leads to better adoption as you start to rollout your scheduling process.  When users understand why they are getting the results they are getting, they are going to be more willing to use the system and funnel questions to your help desk.

Speaking from Experience

Now that we have a common understanding of what a forecasting strategy is, I’d like to share some do’s and don’ts that you should keep in mind as you develop your forecasting strategy:

• Don’t work in a vacuum.  Understand who the stakeholders are across your organization and get them involved early in the process.  Along these lines, don’t forget to include the field and store leadership in this process.
• Do educate the team members so that they understand the role of forecasting in WFM and aggressively work to resolve any confusion around how the WFM forecast reconciles with the forecast and/or budget.
• Don’t put too much pressure on yourself or the team to have a clear strategy quickly.  You should certainly put a straw man together as soon as possible; however, as you vet the strategy with others within your company, their input – and the results of your testing and analysis (see next bullet) will help the strategy evolve.
• Do test your strategy.  You can’t develop a good forecasting strategy without doing a lot of number crunching.  I mentioned this above but I can’t stress it enough.  Use a separate instance of your WFM system or a spreadsheet to quantify the quality of your forecasts and objectively understand the impact of forecast accuracy on your schedule.
• Don’t underestimate the amount of time involved testing the strategy. Each driver must be measured for accuracy across various months of the year and across varying volume stores.
• Do revisit your forecasting strategy periodically.  Overtime, you’ll need to fine tune, and in some cases, make material changes to your forecasting strategy to reflect the natural evolution of your business.
• Don’t be afraid to develop your forecasting strategy even if you already have a WFM system deployed.  Improving your forecast accuracy can result in better schedules.  This delivers additional business benefit from your existing system.

Get Started!

Now, what’s holding you back from developing your forecasting strategy?  As always, feel free to leave a comment below or e-mail me your thoughts.

This post is part of the Axsium Retail Forecasting Playbook, a series of articles designed to give retailers insight and techniques into forecasting as it relates to the weekly labor scheduling process.  For the introduction to the series and other posts in the series, please click here.