Buyers need forecast for multiple reasons. This article deals with why buyers need to be good at forecasting and how buyers can create quickly a demand forecast. Before reading further this post, you should read our central article on forecasting. You can also have a view at our forecasting solution demonstration.
3 reasons why buyers need to be good at forecasting
Demand forecasting helps determining the right moment to pass a command
Buyers should build at product category level a demand forecast. This allows them to pass the order to suppliers in a timeline maner and make sure the right quantity of products will arrive at the right moment to be sold. This is called "product or category lifecycle curves". These curves are the graal for every person in charge of supplies.
Let's take an example.
You run a fashion for kids company. The buyer responsible for "boys trousers" product line will need to anticipate the peak in demand, when customers will massively come in store to purchase a trouser or a jean for their kids:
- Plan 6-8 weeks ahead to get the product shipped from India or Bangladesh for a delivery in Europe. Customs, shippement, transport, paperwork. Not too early, not too late.
- Anticipate the sales season, when shops need to be filled up at 110% of their capacities and a lot of products are being ordered just for that.
Based on the sales forecast, provided that the forecast is accurate, buyers will be able to build a purchasing schedule and dispacth the orders to the right supplier at the right moment in time. Some orders will go to factories located nearer, faster to produce and ship the product but more expensive, some will go for the mass production.
Satistyfing the customers, by getting the right size, color or material
Product attributes matters for buyers:
Upload a CSV and start generating sales forecasts at scale with time series analysis.
- What is the % of product to buy per size?
- What are the colors the most purchased - and those for which quantities ordered need to stay low?
- Is a product available in multiple material? And if yes, what should be the mix to order to meet demand?
This appears more to be a breakdown of the general product forecast. However, in some cases, each product attribute could also be forecasted separately. This would allow purchasers to cross-check their order planning.
Command grids per size, color are critical to get the right mix of product. Usually, this information is known, as you analyze your weekly performance. A good buyer should know what are the trends, what are the products customers look for. Data, analysis and reporting are the key to success to be a good buyer. Justin O'Shea, the famous buyer for MyTheresa, is well-know to use extensively reportings and number to keep track on what product works - and which not.
Buyers have direct impact on the company's financial results
Order too much and the stocks will stay: Inventories consume cash and is just sleeping money. CFOs hate having a high non rotating inventory. Inventories have another problem: they depreciate. Fast consumption make products become quickly obsolete. What is not being sold now will need to be sold in the future with a discount. Buyers need to order just the right amount of product with the right type of product. It's something between art and science.
Inventories also generate additional costs: expensive warehouses, security costs, logistic costs. Having a buffer to meet demand is good, having products piling up on shelves are harming the business. These are additional costs.
Order too little and you will impact the sales team. Sales people usually attribute the success to themselves and make two departments responsible when results are not good: marketing and purchasing. "Product not available": this is the nightmare of every sales leader. There is nothing more frustrating knowing that 1) the demand is there but 2) the product is not.
CFOs always keep a close eye on stocks and created accounting methods to evaluate them: FIFO (First in, first out), LIFO (Last in, First out). The reason why is that inventories is a key driver of the WCR - Working Capital Requirement.
How buyers can make a forecast?
When it comes to forecasting, there are no secrets.
1) Start by getting historical sales
Data gathering comes first. Use your datawarehouse or whatever data system you have at hand to extract past sales. A critical step will be to clean the dataset and remove the outliers. The top 1% and bottom 1% of your dataset might create noise and introduce deviation in your forecast. Hence, this is why you need to go through a data validation and data cleaning step before proceeding with your demand forecast.
You can also use whisker box and percentiles methodologies to accurately remove these outliers. Python, and the SciKit module, can help you doing it.
2) Build a monthly / weekly forecast with Excel
Once the dataset ready and outliers removed, you can get an idea of the demand by using a quick pivot table and group the data per month or per week, over the past few years. This will allow you to have an idea of the demand seasonality.
This plotting step will allow you, from a high level perspective, to have an idea of the general trend, to determine if there are any peak or repetitive cycles (seasonalities) and detect anomalies.
What you should pay attention to are the "shortened peaks". Those are the peaks where you see a sudden flattening of the sales curve, following by another peak. This indicates you experienced product shortage in the past. In other words: sales missed opportunities. When reading an historical sales curve at a high level, you will not see. However, if you double-click at a product category level or SKU, you will quickly figure out whether or not you missed sales opportunities by not having the right quantity of product.
Take time to read them carefully and take notes separately as you go through the analysis of the data. These might be starting points to trigger a discussion with all directly departments: sales, marketing, design, logistic. Was there something odd last year? Where should we stir the demand?
Also include the customers feedback. Customer service has a lot of information, spend some time with the CS leaders to understand what can be improved. Return rate of the online shop are also a good proxy to detect errors and mistakes.
3) Establish a demand forecast
Proceed as you would generally do to build a forecast.
You should use a combination of the two following procedures:
- Use a statiscal model as a general guideline. Our solution make it very easy to get a forecasting, just by uploading a CSV file with historical data. This will use statistical models to reveal trends, seasonal pattern, include holidays and bank holidays components
- Adjust the statistical forecast manually on Excel and factor-in your management team "guidelines".
Still considering the kids fashion example, imagine now that your direction wants to push the "teeanager" categories. Global strategy wants to increase their market share among the teenagers segment. Last year, 14 y.o where accounting for 5% of the total boys trousers sales. Next year, this number might go up to 8%. New design, maybe some marketing actions dedicated to the teen segment, introducting a new line of cool jeans. This is what you need to adapt to fit the general company's plan.
4) Adjust timing to purchase in a timely manner
This is the final and most critical step.
Once you know when sales will happen and in which attribute proportion, as a buyer, you will need to anticipate on the timing. Check with the marketing and designers when collections plan will be ready. Determine a schedule and fix deadlines to order a product in order to meet the demand you have just forecasted.
This general planning should be a north star calendar to align purchase, sales and marketing departments.