Affinity analysis, also known as “Market Basket Analysis” is used to identify the two products who have a high probability of being sold together.
Let's take an example. A new mom explores feeding bottles for her little one on a shopping website. While she's still exploring, a suggestion pops up for related products like a pair of bibs and a swaddle cloth.
These suggestions are not just assumptions but a result of structured analysis that is based on the historical data of customer’s buying behavior. This analysis can be utilized for upsell, cross-sell, promotional offers, loyalty programs and even a store layout. Another reason why this analysis becomes important is because it helps to identify if a marketing tactic is worth its implementation. For example, a buyer who comes to buy an aerated drink will mostly buy a pack of wafers therefore a discount on wafers is just a margin loss.
Affinity analysis helps to identify a pattern of buying behavior and the items bought could either be related or not. For example, a study revealed that beer was sold with baby diapers, though they are not related but the analysis is about finding a pattern and not the reason behind it.
How do you calculate Affinity Analysis?
So this analysis is all about making association rules between products based on the transactional data in the following way:
IF[Item A] THEN [Item B]
So the above rule implies that Item B has a huge probability of sale if item A is purchased. Item A is called the antecedent and item B is the consequent.
The rules may become complex depending upon the level of analysis and data available.
A real life association rule may look like the below:
IF[ Burger, Coke] THEN[Fries and cheese dip]
These association rules are based on 3 important factors or measuring parameters:
Confidence
Support
Lift
Confidence: Confidence is the measure of how strong an association rule is.
Support: Support is the frequency of the rule in the transactional data
Lift: Lift is the ratio of Expected confidence to real number of transactions occurring g as per the expected rule. This measures the performance of the rule in terms of the confidence expected. Lift also plays an important role in determining rank and prioritizing association rules.
Association rules: how are they generated?
So the next obvious question is how do we arrive at these rules when the dataset is huge. It can be done through various programming languages and also a famous algorithm used for this purpose called the “Apriori Algorithm”. These are run on a transactional dataset to arrive at the strongest rules based on their confidence, support and lift.
Affinity analysis can also be conducted on various other platforms like Tableau and Google Analytics.
Let's take a look at an example for a stationery store who did an affinity analysis for a set of his products:
In the table above, the numbers denote the distinct count of purchases made in combination like pen and pencil were purchased 5 times out of a given 50 transactions.
The highest number of purchases were made for a combination of Binding material and name stickers i.e. 27 whereas the next best combination was pencil and eraser at 25. But another factor that comes into play is the profit percentage that these combinations give. So we saw even though the binding material and name stickers had the highest number of combinations, their percentage profit was very low, whereas, the pencil eraser combination gave the best profit percentage and therefore won the highest rank for the association rule.
Why Affinity analysis is an important metric?
Affinity analysis, undoubtedly, becomes an important growth metric for marketers for their number of advantages
There could be certain product combinations which may pull down the sale and are easily identifiable with the help of affinity analysis.
Increased cross sell and upsell opportunities
Helps to prioritize the combinations and therefore take better informed decisions
Helps in making promotional offers or discounts on the right product and removes where not required.
Makes your client's life easy as they get it all they need in one place
Better allocation of marketing budget
Helps marketing campaigns with high quality data
Comments