Case: Product Recommendation

Purpose: To increase sales from existing customers, many time vendors use product recommendation to existing customers in a way such as "Customers who bought this also bought that..", "Viewers who viewed this also viewed that.."


First, the user can pair up two items and count number of people buying both by building an association.


Then, the user can recommend the customer products to purchase based on what he just ordered. For example, the user can recommend the customer who just bought potato chips with one of the following commands.

GET TOP 10 FREQ('Potato Chips') FROM prod2prod

GET TOP 10 PROB('Potato Chips') FROM prod2prod

Furthermore, if the user would like to generate a graph of association between products. He can do following:

GET TOP 200 FREQ() FROM prod2prod

Choose 'Chord Chart' from Chart tab to get visualized associations.

Case: Customer Based Recommendation

Purpose: Identify customers with similar profile can be an efficient way to up-sell. Customers' purchase history may achieve such purpose.


First, the user can build up an association by pairing up customers who bought the same products.


Then, the user can query customers with similar profile from the customer association.

GET TOP 10 FREQ() FROM cust2cust

If the user wants to recommend some products to Theresa Campbell, he can use above query and finds Judy Nelson sharing the most interest with Theresa with 39 same products purchased in the past. Things Judy likes may also attract to Theresa. The user can push recommendation of products that Judy has bought but Theresa has not by first finding what Judy has bought.

BUILD TABLE theresa_judy AS (FIND TOP ALL FROM sales {sortby: count(qty) & filter: = 'Judy Nelson'})

Then excludes what Theresa has bought from the list.

FIND TOP 10, count(qty), sum(qty), sum(total_price) FROM sales, theresa_judy {filter: <> 'Theresa Campbell'}

The user will get below list to recommend to Theresa Campbell.