The concept of data triangulation
The concept of triangulation is nothing new or difficult.
We can define ecommerce simply as using a stack of technology to put products in front of customers in order to generate profit.
Then if we look at sales in the light of the Pareto principle (or 80:20 rule) we can see that:
- 20% of customers account for 80% of sales.
- And 20% of products account for 80% of sales.
This means that 64% of sales come from top buyers purchasing top products. In light of this generalisation, the concept of data triangulation is finding that sweet spot, knowing what best customers buy which best products to generate the most profit.
Of course, the long tail is important for a retailer like Amazon, with lots of customers and lots of products. But not every retailer has infinite shelf space.
The concept of ignoring the long tail customers in order to spend time on the most valuable is what Tim Ferriss expounded in his immensely popular book, ‘The 4-Hour Work Week’.
Mail order companies have for a long time been trying to master data triangulation, for obvious reasons. Nobody wants to send a £5 catalogue out to a customer that isn’t interested, or indeed to print a catalogue with the wrong product mix in it.
With the advent of the web and the maturation of ecommerce there is now enough data at sufficient resolution to do something useful with it.
What are your best products?
- Products clicked most often.
- Products bought most often.
- Products with the highest order revenue (that’s total order revenue, not simply the product in question).
- Products that encourage most return visits to your site.
Using the AIDA model of customer journeys (awareness, interest, decision, action) as illustrated below, we can give different definitions of ‘best product’ depending on what stage the customer is at.
Click propensity – promoting off site
Which products should you promote away from your website (via display, paid search etc)? This is all about the ‘awareness/interest’ stage of the customer journey.
Fairly obviously, one should promote products with high click propensity, or proportion of clicks to impressions, in order to attract as many new visitors as possible.
Purchase propensity – promoting on site
Which products should be prioritised in your merchandising?
Take a category page with 30 products showcased. We want to assess how each product performs against the null hypothesis of random clicks and random purchases, or an equal share of clicks and purchases for each product.
With 30 products the null hypothesis is 3.3% CTR, 3.3% purchase.
Mapping how often each product (in this case three completely hypothetical examples) is viewed and sold as a percentage of the null hypothesis can be visualized as follows:
Those products that are under-viewed and yet oversold are the ones to focus on. In Mike’s words:
..adjust traffic flows through your site to match product-views with product-purchase-propensity.
Ensuring highest margin
Aside from focusing on highest purchase-propensity items, and factoring in total order value, there are other factors that affect profit.
Ensuring the highest margin is achieved involves looking at each product’s profit margins but also at the product’s return rate and any discounting that may in place.
That concludes this short and simplified version of ‘finding your best products’ in Mike’s presentation. I’ll be following this up with ‘finding your best customers’ so do stay tuned.