I think there are 2 issues in your approach -
1. You started first with age group and then shifted to households
Households will have people less than 15 and more than 50 years of age. In your approach, it looks like 40 Millions households have only young population and rest lives in other household which you are not counting.
2. There is a difference between who buys the pizza and who eats the pizza. There is a big portion of population (less than 15 years) which does not buy pizza themselves but probably their parents buy it for them
A better way to approach is to find the total consumption - you can do this by age or by household but stick to one. I would have multiple filters -
1. Age groups
2. People who want to eat pizza (Interested population)
3. People who have the capacity to buy and eat pizza (Income levels)
4. People who will actually buy pizza (combine interest and income level)
5. Put some cyclic factor for weekdays and weekends to make it more interesting.