One of the major data projects I worked on was building a "customer journey" asset to stitch together and sequence each interaction a customer had with the company: sales, billing, processing, call center, website.
Each department was huge (there were many millions of customers over decades), and so each was bucketed by function. There was no universal "customer id" and this made it incredibly hard to create that sequence of data. We relied on fuzzy matching and other algorithms to get our best possible results.
Beyond the data, for the employees of those divisions they didn't really have perspective of the entire customer experience. No idea that problems in the billing department transformed into calls to the service center (which transformed to operational expense on the company's bottom line).
At any rate, some mental model of bucketing is useful for all the reasons you mentioned. But it's up to the savvy mind to create buckets well fit for the end-game solution.
One of the major data projects I worked on was building a "customer journey" asset to stitch together and sequence each interaction a customer had with the company: sales, billing, processing, call center, website.
Each department was huge (there were many millions of customers over decades), and so each was bucketed by function. There was no universal "customer id" and this made it incredibly hard to create that sequence of data. We relied on fuzzy matching and other algorithms to get our best possible results.
Beyond the data, for the employees of those divisions they didn't really have perspective of the entire customer experience. No idea that problems in the billing department transformed into calls to the service center (which transformed to operational expense on the company's bottom line).
At any rate, some mental model of bucketing is useful for all the reasons you mentioned. But it's up to the savvy mind to create buckets well fit for the end-game solution.