Most people in retail can't agree what a forecast is - no wonder we’re not very good at it.
Supermarkets have been selling things and collecting data for years, so why is it so hard to forecast accurately? And what is a forecast anyway?
Suppliers to UK grocery retailers deserve to receive timely and accurate forecasts - so says the Grocery Code Adjudicator. Their code of practice (GSCOP) defines what is expected, and their 2016 best practice guide suggested that all regulated retailers achieved this.
Did they? Forecasting and promotional planning formed 2 of the top 3 issues in the 2017-2018 GCA annual report, so what is the real story?
Background - why is this such a hot topic?
The ten largest UK supermarkets are all subject to GSCOP, which requires them to
“… fully compensate a Supplier for any cost incurred … as a result of any forecasting error … unless … prepared … in good faith and with due care …” and to
“… ensure that the basis on which it prepares any forecast has been communicated to the Supplier.”
Further advice from the GCA, in March 2016, suggested that all of the ten GSCOP retailers were (at that time) compliant with the code, yet forecasting was highlighted as one of three Top Issues (along with payment delays and promotions) in the 2017-2018 annual report. That report (on page 35) highlights suppliers reporting poor forecasts from retailers, significant variations between forecasts and orders, and penalties for failing to meet service levels whilst some cases result in suppliers left with significant amounts stock.
It might help if we all knew what we meant by ‘forecast’
The problem is that it means many different things to different people, and unfortunately GSCOP isn’t clear about what it means. Consider the following forecasts:
- An annual prediction of the volume of every product which will be purchased from a supplier (a joint business plan?)
- A 3-month prediction of weekly product sales in store (a seasonal forecast?)
- A 6-week aggregate weekly order prediction
- A 14-day daily prediction of short-term product sales, allowing for weather and promotions
- A 7-day demand forecast (not the same as a sales forecast, and more important to a supplier). Is this by depot?
- An order, for delivery tomorrow
Which of these is the ‘forecast’ that GSCOP requires retailers to share with suppliers? Which of these is the most useful to the supplier? Which should be measured against actual orders to determine ‘forecast accuracy’?
If a retailer makes the following ‘forecast’ for demand of one SKU for 7 days from now, and amends the ‘forecast’ each day like this:
7 days out - 100 cases
6 days out - 100 cases
5 days out - 100 cases
4 days out - 80 cases
3 days out - 80 cases
2 days out - 60 cases
1 day out - 40 cases
And then places an order for 40 cases - does that mean they get 100% forecast accuracy, or 40%?
It would be easy to argue that NONE of the ‘forecasts’ described above is of real use to the supplier. Why? If the point of sharing a forecast with a supplier is to ensure that they meet order expectations (this is implied in the latest GCA annual report) then the supplier needs:
- Forecast order volume
- By product (not by category, and not a total)
- By day (important for short-life goods and/or just-in-time logistics)
- By depot (essential for appropriate inventory in the right part of the country)
Without this, any other ‘forecast’ is at best a guide to what might be required and when, but won’t help the supplier ensure that appropriate inventory is in the right location at the right time such that orders can be fulfilled to high service-level targets - typically 98% or above.
Why is it important?
- A supplier needs to know what to produce, and what not to (and plan all the associated knock-ons such as labour, raw-materials and transport)
- A collaborative retail-supplier relationship should be using all available data to ensure that the balance between 100% availability and 0% waste is as good as it can be
It must be concerning that in the GCA June 2018 publication, it states that:
“It was found that retailers adopted a range of approaches, and used the word “forecast” in a variety of ways. Some made a clear distinction between a forecast and an order; others did not see forecasting as a discrete activity but rather, as an integral part of supply chain management, often proceeding close to real time;”
The GCA has made 17 recommendations as to how to improve the current process. Such as:
- Closer collaboration between retailers and their suppliers
- Regularly reviewing forecasting performance
- Ensuring that suppliers are able to get access to supply chain or buying teams to share intelligence and discuss forecasts or orders
- Ensuring that retailers have adequate systems and processes which learn from and take account of known or past issues
- Ensuring that suppliers are able to access adequate sales data
Sounds easy doesn’t it? To be fair, these are all good suggestions, and achievable...IF both sides can make better use of available data, and are willing to share and discuss insights in an easily accessible way. This requires:
- Consistency of definition across retailers
- Access to timely and robust data
- Suppliers having the ability to quickly interpret data
- The ability to challenge forecasts and hold retailers to account
Some suppliers have hundreds (even thousands) of products going into multiple depots each day. Not only do they need accuracy, but they need to be alerted to changes in forecast.
How data visualisation helps
If it was easy for retailers to provide accurate daily product-by-depot order forecast then they would - it is in their interest to advise suppliers quickly, efficiently and automatically. The fact that such forecasts don’t exist suggests that the process of forecasting demand - in the form of predicted orders - is complex and error-prone. It’s also quite simply a lot of data (not to be confused with ‘big data’!)
Working out what will happen in the future, can only be based on information that we have about the past, mixed with known changes (weather, promotions, seasonal events etc.)
The good news is that some retailers make most of the necessary data available - and whilst they may not have the time and resource to look at all of their ~40,000 SKUs daily, suppliers can. By using best-practice visual analytics tools and techniques, supply-chain and sales data can be blended to highlight indicators for demand (such as sales spikes, low-depot stocks etc.), and provide an easily understood picture of recent and historic sales patterns. Contrary to the caveats given to you by your financial advisor, short-term history is often the best indicator we have of future performance.
In an increasingly competitive retail environment, those retailers that can provide accurate forecasts to their suppliers, will benefit from more efficient supply chains, lower costs, better availability, and happier customers. (And a happier Grocery Code Adjudicator).
Ian started his career as a buyer at Tesco, followed by several years in grey-market procurement and supply, as well as branded sales roles. He has been Chief Operating Officer at Atheon Analytics for 12 years.