Metrics for a new AI Photo Editor in Google Photos

Ishant Juyal’s approach

For AI-powered editor feature, the primary metric would be:

• The number of images edited with this filter/feature
• The number of users who use this feature
• The percentage of DAU who used this feature
• Average time spent
• Number of people who started editing but did not complete the editing
• Number of iterations people used

We can have a feedback model after the photo has been edited and then we can track the feedback if people are giving positive feedback or negative feedback on the photo.

If there is any limit, how many times the limit was hit could be one metric.

If we are giving an upgrade up to the limit hit, we can also track a number of times people updated. The conversion percentage.

The primary metric would still remain the number of photos edited per day.

Pavlo’s approach

• Usage time (from entering the editing mode until completion)
• Processing time (how much time it takes to process and apply the editing)
• Number of reiterations
• Number of declines (Redo)
• Frequency of usage in relation to number of photos taken/viewed
• Path the user takes to open the photo when they initiate editing (when strolling through the gallery, right after taking a picture, etc)
• Types of photos (portraits, objects, outdoor/indoor, etc) - if possible
• Types of things to each the editing is applied - if possible
• Who are the users who use the feature (age, interests, etc)
• Number of sharing after editing
• Device type/model