User flow

In previous posts, we analyzed how different product features can impact user retention and monetization. This information could help us to prioritize product features and focus on the most impactful ones.

But as it usually happens usage context could change our understanding a lot. That’s why it’s quite reasonable to learn usage context:

• what happened before a particular product feature was used (will be touched on in this post)
• what happened after a particular product feature was used

Let me remind you of the result of product features' impact on retention from one of the previous posts.

MCC coefficient — the coefficient for measuring impact on retention.

Product features are sorted by decreasing [# users]. For each product feature, we calculated the MCC coefficient — how this particular product feature impacts the user retention to the product.

A few examples from the table above:

• feature41 is very popular ( [# users] = 11,598 ) and has an almost natural impact on user retention ( [MCC = -0.0093] ).
• feature24 is often used ( [# users] = 5,121 ) and has a quite strong positive impact on user retention ( [MCC = 0.0954] ).
• feature18 is very popular ( [# users] = 10,571 ) but has a quite strong negative impact on user retention ( [MCC = -0.1532] ).
• feature43 is less often used ( [# users] = 2,541 ) but has a very strong positive impact on user retention ( [MCC = 0.1593] ).

Let’s figure out what users did before they start using these product features (current user step — column event_type2, previous user step — column event_type2_-1).

• feature41

feature41: flow analysis.

As we can see a lot of users come to feature41 (column event_type2) from feature41 (column event_type2_–1). An interesting moment here is the sign and magnitude of MCC coefficient from the previous step:

• feature41 at current step = -0.0093
• feature41 at previous step = -0.0251

It means that if users continue using feature41 then they start receiving a little bit more value and their friction is decreasing.

But the most important learning is this:

MCC coefficients for independent product features are different from MCC for product features in the usage context.

If we take independent MCC coefficients from the table above we will get these figures:

• feature4 = -0.1319 (negative, strong)
• feature42 = 0.0352 (positive, light)
• feature2 = -0.0996 (negative, strong)

If we take usage context MCC coefficients from the flow above we will get these figures (I used italic to highlight changes):

• feature4 = 0.0079 (positive, light)
• feature42 = 0.0938 (positive, strong)
• feature2 = 0.0495 (positive, medium)

It means that depending on usage context (user’s previous experience) some product features can change the sign and impact strength.

• feature24 / feature18

feature24: flow analysis.

I would like to point out to the readers that the feature24 impact on retention is strongly positive as all 4 main previous usage cases are positive as well.

feature18: flow analysis.

I would like to point out to the readers that the feature18 impact on retention is negative as almost all 4 main previous usage cases are negative.

Positive past user experience (positive mood) is usually followed by positive user experience, and conversely, negative past user experience is usually followed by negative user experience.

• feature43

feature43: flow analysis.

I would like to point out to the readers that the strength of the feature43 impact on retention is related to the very strong positive impact of the main previous usage cases (feature41 and feature43).

The level of user experience is the cumulative value of past user experiences.

SUMMARY:

The user flow charts that I presented above should demonstrate two important aspects of product development:

• user experience is not just a bag of product features, it’s a flow.
• user experience within your product is accumulative.