Having custom activities set up isn't so bad. if you have a spare data scientist lying around, I'm sure they'd appreciate the amount of data you're collecting and might give you a nice model in return.
Barry Dupont To the problem at hand, regardless how you identify the the SKUs that a person hasn't purchased, what makes this a hard problem is the amount of SKUs you have. At every step of the way you need to be thinking about scale. So one thing you haven't mentioned, if you have data of a person who e.g. has visited pages but has not purchased products 123,546,2398... What do you plan to do specifically?
Without knowing what you plan to do with the data, the first question that comes to mind is, can you not use something a few levels up like a product family instead of individual SKUs?
Other than creating smart lists (e.g. visited webpage for product 123 AND NOT Purchased product 123), Marketo is horrible at comparisons between fields - and I get it you don't want to be creating 2500 smart lists. Basically, I agree with Sanford that you need to use webhooks, i.e. send fields to be compared, via webhook, to a service that can. In fact for most things that need scale, webhooks is the way to go (typing some lines of code beats creating multiple assets in Marketo)
So, to even get to the point where you can identify a set of products that individuals have not purchased but have viewed, I would set up 2 custom fields. One for storing the visited/quoted/purchased data, one purely for marketable SKUs. Because dates matter, the data in these should be json format. To actually populate these fields and keep them updated, use webhooks OR if you have a custom CRM integration, your middleware might be able to.
e.g. sku raw data
{
"visited":
[
{"sku":123,"date":"2019-04-23"},
{"sku":345,"date":"2019-05-28"},
{"sku":666,"date":"2019-09-11"},
{"sku":9001,"date":"2019-11-12"}
],
"quoted":
[
{"sku":345,"date":"2019-06-13"},
{"sku":666,"date":"2019-11-19"}
],
"purchased":
[
{"sku":666,"date":"2019-12-01"}
]
}
business rules: marketable within 6 months of visited or quoted (whichever latest)
marketable SKUs = [345,9001]
when an input changes, i,.e. however you're collecting visited, quoted, puchased sku data, you'd need to call a webhook to update the json data and spit out the marketable SKUs based on what ever date range business rules you've decided
.png)