Build customer → category → row counts

profile = {}
foreach bill in bills {
    customer = bill.Customer
    if (not isKey(profile,customer)) {
        profile[customer] = {}
    }
    foreach item in bill.Items {
        category = broadCategory(item.Category)
        if (not isKey(profile[customer],category)) {
            profile[customer][category] = 0
        }
        profile[customer][category] = profile[customer][category] + 1
    }
}

The lab pre-fills all five category keys with zero for a readable table. The pseudocode above creates only observed category keys. They are equivalent for the displayed profile if missing category counts are treated as zero when forming the five-column vector. Category order is [Apparel, Food, Toiletries, Utilities, Household].

A zero here means no such row was observed in these supplied bills. It does not establish that the customer never buys that category elsewhere. Customer → category and customer → customer tables answer different questions.