I like Blueberry PopTarts

I like Blueberry Pop Tarts, or I How I Learned to Stop Worrying and Love Demand Planning
Advantage ERP Group

Ok, yes, I like Blueberry Unfrosted Pop Tarts.  But here’s the rub, my local store doesn’t sell my favorite, instead Strawberry Unfrosted finds favor on their shelves.  And when confronted with this comfort food letdown what do I do?  I buy the Strawberry instead and try to pretend it’s my flavor of choice.  By now you might be thinking ‘come on Dave, you usually talk about topics like Three-Tiered Problem Solving and Managing Corporate Intelligence, who cares about Pop Tarts, quit the rant already!’ 

Ok, I’m with you, but this is a classic example of the age-old challenge every company faces:

How do we make sure we sell and stock the items that our customers actually want to buy?

Let’s come back to my Pop Tart fixation.  When I compromise and buy the Strawberry version, that purchase registers as demand in my local store’s demand planning / replenishment system.  My purchase of an item that I really did not want makes it more likely that the store will continue to stock the undesirable product, at the expense of the desired product.  My store appears to have no way to capture the actual item that their customer really wants to buy.  Worse for my local emporium, I discovered that another store some distance away stocks my beloved pastry, so I’ve now started to shop there instead.  If I’ve driven out of my way for something as silly as Pop Tarts, why not get all my shopping done since I'm already there?


Now back to our regularly scheduled ERP blog…

Most ERP systems can generate a forecast by projecting sales history into the future, and they use this forecast in the Material Requirements Planning (MRP) module.  Some systems generate forecasts by simply extending recent sales history into the future.  Other systems have a more sophisticated approach using algorithms such as exponential smoothing.  And then the next level of forecast generation uses a system such as Forecast Pro, which applies a series of advanced statistical models to the data and automatically chooses the model that produces the best results.  Each of these approaches rely on sales data that correctly reflects the market’s true historical demand.  Consider:

  • Is the sales history skewed by substitute items? Using my example, the customer ordered Blueberry Pop Tarts on their PO but we shipped a substitute item – Strawberry.
    • This is sometimes best addressed by basing forecast generation on sales order data instead of invoice data. Sales orders are often the best representation of what the customer wanted to buy.
    • Be careful – many ERP systems allow shipments and invoices without a sales order. If this occurs then forecasting from sales order data will understate actual demand!
  • Was the customer short shipped or over shipped?
    • Does the customer regularly order in non-standard quantities and then accept shipments with a different quantity? If so does it make sense to change the standard quantity?
    • How often does this happen?
    • This is like the substitute item scenario, so consider basing the forecast on sales orders instead of invoices
  • Do the dates in the sales history accurately corelate to the date when the customer actually wanted to receive the shipment? And what's the big deal about a few days difference? 
    • Maybe the customer’s requested ship date was not honored because of a supply chain issue
    • Or possibly the customer agreed to an earlier / later shipment to take advantage of a freight discount
    • This can have a big effect when using systems that determine seasonality from the data, especially if the date difference spans planning periods.  When forecasting we need to focus on what the customer wanted to happen, and less so on our actual performance.
  • Does the sales history contain unusual or one-off customer shipments?
    • Consider excluding closeout and new store opening shipments from the sales history before generating the forecast
  • Does the sales history include ‘drop ship’ (or back-to-back) shipments, where the customer sales order initiates a vendor purchase order and the vendor ships goods directly to your customer? In this case the material isn’t processed in your warehouse.  This may seem obvious, but it’s best to decide to include / exclude drop ships based on how will the forecast be used. 
    • If the forecast is used to plan stock replenishment, then yes, exclude drop shipments!
    • If the forecast is used for financial or managerial planning, then look at including drop ships as a separate sales category (i.e. warehouse shipments and drop ships)
    • Consider sending ‘drop ship only’ forecasts to vendors so they can improve turnaround of your orders by optimizing their planning and supply chains


Ok Dave, I’m with you on the need to feed good data into the forecast generation process, but what about things that should impact the forecast but are just not in the data?

Tough question, this is what makes forecasting an art, not a science.  Each forecast for each business is unique.  Factor in these ideas when building forecasts:

  • Consider big-picture factors such as the national / world-wide economy, industry trends, changes in fashion, buzz at trade shows
  • Carefully listen to sales reps, customer service teams, and the customers themselves
  • Review tweets, LinkedIn comments, and Facebook postings about your company and your products
  • Survey customers to determine what they would purchase from you if they could.  But be careful here, customers sometimes will give you wish list items that they can’t or won’t actually buy.
  • Use or deploy a win / loss analysis process to best understand what made prospective customers decide to buy from you. More importantly use this process to gain insight into why prospects chose to not buy from you!
  • Deploy a computer screen on the Pop Tarts shelf asking customers ‘which flavor would you rather buy?’


To sum it up – companies often can improve their forecasts by doing the heavy lifting with the sales history data.  Understand the data – where it comes from and when / how it is collected.  After that we’ll be ready to dig into the statistical and software tools that transform sales history into forecasts which are ready to be used by supply chain planners and financial teams.  Sounds like the subject of the next blog!


1 comment

  • Well done Dave. All valid and important points to consider.

    Michael Blair

Leave a comment

Please note, comments must be approved before they are published