great thread and follow on, shomik! For pre/early product- I'd add that when you consider building the low generalizability features its important to not only consider the potential learnings, but also what the ongoing costs (tech debt, ops, etc) could be. If the learnings are high, but the debt is also high it may not be worth it (even if upfront cost is low). It can be very difficult to sunset a feature that was directly paid for- even if it has high ops cost and is non-core.
great thread and follow on, shomik! For pre/early product- I'd add that when you consider building the low generalizability features its important to not only consider the potential learnings, but also what the ongoing costs (tech debt, ops, etc) could be. If the learnings are high, but the debt is also high it may not be worth it (even if upfront cost is low). It can be very difficult to sunset a feature that was directly paid for- even if it has high ops cost and is non-core.
Very true. It’s a complex decision in the early days that requires a lot of thinking and tradeoffs and potential benefits. Great point!