The Dude has seen some bizarre shit in his days, but not too long ago, he saw a thread on Twitter that really poured sand in his gearbox.
The thread was about how a product leader interviews candidates. In general, the Dude is not adverse to what if, and scenario based questions. But this one really stuck in his craw.
Paraphrasing, the question was:
You are the product manager for the Netflix app main page. How would you define and construct an algorithm to personalize the selections a viewer sees.
That is a fine question, but the clarifications were inane. Upon clarifying questions, it became clear that:
- The expected responses were whiteboard algorithm development
- That the product manager be amidst the various algorithmic decisions (as in, playing the role one would expect of a software architect)
- That the product manager not only be able to code (or p-code) the desired algorithm, but to then apply business spidey-sense to do some a priori product-market fit simulation (in their head).
(I should add, it wasn't for a job at Netflix, it was some other startup of the day)
You might think this is a reasonable ask of a product manager candidate, but in reality, the product manager is to provide the high level design goals, translated from customer insight, market intelligence, and domain knowledge.
If you are expecting your product manager to be a über software architect with business savvy, you will never find what you are seeking. Full-stop.
If this is your interview process, and the Dude makes it to that first interview, he will be taking a hard pass. Like, get up out of his seat, saying “Good day” and leaving the building.
The only places that have these expectations of their product managers are companies that really have no fucking clue what product management is even if it smacked them in the face. They may find some success in making a senior SW Architect into a product manager, but they will struggle long term.
The Dude’s thoughts on the Netflix Recommendation
The only real answer is:
- Buy fuck-loads of personal data on your users and potential users - Data brokers, abuse the Facebook APIs (a-la Cambridge Analytica and the Trump campaign) and other sources
- Build affinity maps of their interests, grouping in to key classes
- Ignore this affinity map, and then just push your shitty shows in the recommendations anyway, knowing that your garbage “Netflix Originals" suck hairy donkey balls (with a couple of notable exceptions - Narcos being one of them)
- Hide all the things that a subscriber really likes (for instance, the Dude has a soft spot for BBC police drama shows (Broadchurch, Hinterland, and the like) ) so that they must search for them instead of discover them organically, and also force them to scroll through 15 pages of your 'Netflix Original' trash before giving up and just going to Amazon Prime or Hulu.
Then flip them the bird and storm out.