Question

Here’s an obfuscated version of something I’ve been trying to do at work. Say I have been given this month’s data for customers in my shop – how much they’ve spent split by the food type:

CUSTOMER    FOOD_TYPE      FOOD_TYPE_VALUE
1           SWEET          52.6
1           SAVOURY        31.0
1           DAIRY          45.8
1           DRINKS         12.1
2           SWEET          15.1
2           SAVOURY        44.1
2           DRINKS         23.4
3           SWEET          95.7
3           SAVOURY        20.0
3           DAIRY          10.8
3           DRINKS         57.1

It has been decided that Customer 3 is our ideal customer as he fits the demographic profile, and we want to track month by month, how everybody’s distribution of shopping preferences differs from his.

I can get the percentage allocation for each food type for each customer with:

SELECT
  c1.customer,
  c1.food_type,
  100 * c1.food_type_value / sum(c2.food_type_value)
FROM
  mytable c1 INNER JOIN mytable c2
  ON c1.customer = c2.customer
group by c1.customer, c1.food_type, c1.food_type_value

But I am having trouble constructing a query that will give me a further column with the matching percentage values for my ideal customer. ie:

CUSTOMER    FOOD_TYPE      FOOD_TYPE_PERC  IDEAL_PERC
1           SWEET          37              52
1           SAVOURY        22              11
1           DAIRY          32              6
1           DRINKS         9               31

Any tips on how I can achieve this without too much mess?

Answer 1

Join it on the subset of the customer table that contains your ideal customer:

SELECT
  c1.customer,
  c1.food_type,
  100 * c1.food_type_value / sum(c2.food_type_value),
  c3.FOOD_TYPE_VALUE / sum(c2.food_type_value) as IDEAL_PERC
FROM
  mytable c1 INNER JOIN mytable c2
  ON c1.customer = c2.customer
  INNER JOIN (
    SELECT FOOD_TYPE, FOOD_TYPE_VALUE
    FROM mytable
    WHERE customer = 3) c3
  ON c2.FOOD_TYPE = c3.FOOD_TYPE
group by c1.customer, c1.food_type, c1.food_type_value, c3.FOOD_TYPE_VALUE

Your comment to my answer suggests you need to divide by the FOOD_TYPE_VALUE of the ideal customer, so do this:

SELECT
  c1.customer,
  c1.food_type,
  100 * c1.food_type_value / sum(c2.food_type_value),
  c3.IDEAL_PERC
FROM
  mytable c1 INNER JOIN mytable c2
  ON c1.customer = c2.customer
  INNER JOIN (
    SELECT s1.FOOD_TYPE,
           100 * s1.FOOD_TYPE_VALUE / sum(s2.FOOD_TYPE_VALUE) IDEAL_PERC
    FROM mytable s1
    INNER JOIN mytable s2
    on s1.customer = s2.customer
    WHERE s1.customer = 3
    GROUP BY s1.FOOD_TYPE, s1.FOOD_TYPE_VALUE) c3
  ON c2.FOOD_TYPE = c3.FOOD_TYPE
group by c1.customer, c1.food_type, c1.food_type_value, c3.IDEAL_PERC