What are the differences between subscription-based and “basic” IAP business from a UA perpective?

I agree 100% with @cfarm's answer, these businesses look completely different across all dimensions, not just UA.

From a UA perspective, there are a few fundamental differences between subscription-driven and IAP-driven apps:

1. LTV is a lot easier to model for subscription businesses. Instead of thinking about some stochastic conversion process with decay, you only really need to think about the percentage converstion to paid and churn. That's a huge relief for UA / marketing analytics teams. And a lot of subscriptions are yearly, which is even better.
2. Subscriptions usually happen pretty early in the user lifecycle, so the UA campaigns get the payment event feedback with the entire LTV baked in very quickly. This is incredibly valuable for VO and AEO campaigns on Facebook because the optimization signals are very strong and come early. 
3. As a corrolary to #2, VO campaigns can be hard to run efficiently for subscriptions because there is no / very little magnitude to user monetization: a user either converts (subscribes) or doesn't, and if they do, it's generally at the same level as other users. The exception here is if the subscriptions are offered in tiers of eg. month, 6-month, year with large differentials in pricing. But for apps that only offer one subscription package (which I believe is most of them), there are no "whales," only purchasers.

I would also add that subscription businesses operate more like SAAS businesses and the metrics will shift. They care more about subscription trials, conversions, MRR Net Churn, upfront payments etc.

one key difference is in the nature of the LTV modeling. for a subscription-based business, the LTV is a function of the subscription price and the subscription retention curve - so you can get a basic estimate of LTV from the retention curve in iTunes(but you can certainly get more sophisticated).

for a 'basic' IAP business, the LTV modeling can be more choppy - and typically requires more rigorous analysis of IAP data.