the thumbnail personalizing is cool and all but it’s been 9 days since the new year
A few more ideas:
Adding / removing thumbs
- Would be nice if there was a way to add / remove thumbs without having to restart with all the thumbs once again. The issue now is restarting gives new thumbs equal allocation of impressions for a period (seems 12-24 hours).
- I’ve found myself wanting to test more thumbs, but hesitant in case any new thumbs performance is significantly worse and could end up messing the games performance with the main ranking algo. Currently I have two great performing thumbs, but if I add 3 new ones to test and they turn out to be duds, it’s going to really impact my overall performance, because for a period they’d be receiving equal allocation as the existing proven thumbs.
- What would be useful is if I could add more thumbs and be able to choose the % allocation of impressions they get (I’d probably give them 5-10% so they don’t pull too much traffic away from the top performing ones). Once the algo is getting enough data to decide how it should rank, it can then decide its own impression allocation. Or a simpler approach could be the algo just providing a minimum number of impressions it needs to start making allocation decisions and then slowly increase / decrease as needed once enough data starts coming in.
Optimizing towards Revenue / LTV / custom actions
- Revenue is pretty down funnel and there might not be enough data gathered to do quick ranking decisions, but it would still be nice to optimise towards other more up-funnel LTV proxies such as QPTR of premium users, QPTR of high GDP countries etc. For example, I may have two thumbs - the first has a higher overall QPTR but a second thumb may have a lower overall QPTR but for USA it’s higher. In this case I would probably want to algo to pick the USA thumb with the lower overall QPTR instead.
- Choose to optimise towards other metrics - retention, engagement time etc
- Optimise towards a custom event inside the game. I know that certain behaviours from users within my experience will give me certain results. For example, users that reach level X, play for X minutes etc might be more likely to spend in the game. Or users who meet certain events may just perform in a particular way I want (may be more likely to retain for example).
3 Likes
Another piece of feedback:
Returning vs New personalization.
- Over time, the proportion of home recommendation traffic coming from returning users tends to increase. Often what motivates existing users to return is quite different to what motivates a new user.
- Currently I use icons with imagery showing new content which generally informs existing users to return back, but those icons wouldn’t be suitable for convincing new players to join.
- Similar logic needs to be considered for thumbnail personalization. A high QPTR thumbnail for new users might be very different to a high QPTR thumbnail for existing users. Some way to select the intent here (new vs existing) would be good. Perhaps a completely new tab where I can upload thumbs for the specific purpose of showing existing users.
3 Likes
It would be nice to know what a “good” qPtr is, unless that’s visible somewhere, I can’t really tell if my thumbnail is performing well compared to thumbnails from other games. And it seems to be very different based on how popular the game is. Or is that the “Unique User Qualified Play Through Rate (Recommendations)” from experience overview? But then why is the number so different from the qPtr on the thumbnails? I find this all very confusing.
Suggestion: Make the thumbnail personalization visible from the Experience page, it’s annoying having to swap between them all the time.