It's hard to have a good forecasting algorithm if you don't know what happened in the past.
Everybody has different ways of approaching the topic. I like to use a Principal Components Analysis ... creating two "factors" ... each factor is then "segmented" into seven different classifications that yield 7*7 = 49 customer segments. By tracking the change in file counts and rebuy rates by segment, I can quickly see how a business is "evolving" ... and then I am better able to create a forecast that shows where the business is going.
Let's take a look at where a business has been over the past six years.
Each image below shows how the file evolved over the past six years ... red = many customers ... blue = small numbers of customers.
There are two trends you should pay attention to.
- Lower-Right corner changes in file counts.
- Upper-Right corner changes in file counts.
In the lower-right corner of the image, counts have been on the decline over time ... but in the past year, there is a significant increase in counts. I looked at the attributes comprising customers in this corner of the graph. What did I learn?
- Far lower right count declines are from customers who purchased one time per year, online-only, buying product not in the primary product category for this company.
- Upper right corner count declines are from customers who purchased one time per year, phone-only, buying product not in the primary product category for this company.
In other words, this is a company that appears to have greatly cut back on marketing of supporting product categories ... and customers simply disappeared instead of buying from the primary product category.
But then something changes in the past year ... around X = 1 / Y = -0.8 we see a significant increase in customer counts. What describes this segment?
- Customers who buy 1.5 times per year.
- Their purchases are split between the primary product category and the second-most popular product category.
This company changed strategy ... and is actively cross-shopping customers between the two most important product categories.
On Monday, I'll share rebuy rates across time ... and if you like geeky GIFs, here is a GIF of how file counts evolved over time.
"Forecasting outcomes are the sum of all analytics and marketing knowledge possessed by your company."