What Features Actually Increase Playtime?

What features actually increase playtime?

Turbine analyzed how feature launches correlated with playtime in successful casual mobile games —

And our client gave permission to share the results.

What the data suggests

3 features most strongly correlated with playtime lift:

1. PVP Race Events (the ubiquitous 'flash races')
2. 'Dense' Daily Task systems (4+ tasks)
3. Same-Day Sessioning Incentives (come back in 2 hrs!)

These seem pretty intuitive, right?

Always nice when data (occasionally) supports intuition.

What features didn't correlate with playtime lift?

• Battle Passes
• Achievement Systems
• Guilds
• (many others)

One could rightly argue that features like Battle Pass SHOULD intuitively improve daily playtime.

BUT

1. These features aren't explicitly designed to do so
2. And, they don't punish by removing rewards TODAY

So, not potent / acute enough to inflect PT behavior.

• Battle Pass = mostly monetization, little time pressure
• Achievement Systems = rarely time-bound
• Guilds = activity minimums affects a small % of users

My takeaways

For games that already have proven product–market fit

• User behavior resists change (more than is intuitive!)
• We've found playtime is particularly difficult to move.
• A 'big win' usually looks modest (4-8% lift)*.
• So, to move playtime, use a surgically designed feature.
• Assume other features will have ~0 playtime impact

*These takeaways are informed by this analysis, and the many a/b tests Turbine has done targeting playtime.

Caveats

• This is a noisy dataset & correlation is not causation.
• Playtime data are ST estimates, not actuals.
• Given above, I'd interpret results conservatively, i.e. 'these 3 features are probably more likely than others in the list to move playtime'
• Turbine's extensive a/b testing does support these recos, but we haven't shared that data here

Methodology

• Feature launch dates cobbled from multiple sources
• Manually reviewed implementations for quality
• Playtime data (estimates, per geo) from Sensortower
• Looked at PT 30d before vs. 30d after 1st launch date
• Slightly mitigated noise by analyzing per country across Geos that should be uncorrelated re: featuring
• Tried for at least 20 casual games per feature.
• Note: 'Pet system' only n=6 so confidence low