"How would you measure the success of Google Photos?"

The most common failure is treating Google Photos like a generic consumer app and listing DAU, photos uploaded, and photos shared as north-star candidates before settling on DAU. That answer fails because it has nothing Photos-specific in it, ignores the business model entirely, and would describe any engagement app equally well. A strong answer starts from what the product actually is in 2026 and what jobs it is doing for both the user and Google.

What the product actually is in 2026

Google Photos has 1.5 billion monthly active users as of its tenth anniversary, with approximately 4.3 billion photos and videos uploaded to Google’s servers daily. It has been downloaded over 10 billion times on Android. At that scale it is not a photo storage app competing on gigabytes. It is Google’s highest-trust AI surface: users hand it their most personal archive (relationships, health, travel, milestones) and Gemini reasons over it.

Gemini now powers natural-language search through Ask Photos, allowing queries like “hiking, spring 2025, with Emma.” Veo 2 powers a Photo-to-Video feature that animates still images into six-second clips. Memories surfaced on Samsung TVs in early 2026, showing themed slideshows in living rooms. Users can toggle between AI-assisted and classic keyword search, a control Google added after user complaints about the initial AI search rollout quality. That toggle is not a footnote. It tells you Google understands the core tension the product must navigate: the AI must feel like it knows you without feeling like it is surveilling you.

On the business side: the free tier is 15 GB shared across Photos, Gmail, and Drive. That shared storage cap creates real pressure and a clear upsell path to Google One. Photos is the primary conversion hook for Google One subscriptions. A 1.5-billion-user product with no direct transaction is only viable if the storage paywall converts.

Eliminating the wrong north-star candidates

Photos uploaded per week. This measures backup instinct, not value delivered. A user who auto-backs up 300 photos from a vacation and never opens the app again looks great on this metric. It captures the behavior of a file cabinet, not a product that earns ongoing trust.

DAU. Inflated by iOS and Android auto-backup notifications and passive opens. A backup utility can have high DAU with zero moments of genuine engagement. It is a lagging indicator that tells you whether people have the app installed and unlocked it, not whether the product delivered something meaningful.

Storage fill rate alone. A leading indicator for Google One conversion, but it tells you nothing about whether users value the product or are simply accumulating files they never look at.

The north star: Weekly Memories Engagement Rate

The right north star is the percentage of users who open a surfaced Memory (not just dismiss the notification) at least once per week. Call it Weekly Memories Engagement Rate.

This metric earns its position because it sits at the intersection of the two things Google Photos must do to succeed. First, it measures whether Gemini correctly identified something meaningful: a curated Memory that users actively open signals the AI model understood the archive, not just indexed it. Second, it captures the habit loop that makes storage feel indispensable: a user who returns weekly to a surfaced Memory is not a backup-and-forget user, they are someone for whom Google Photos is the place memories live. That user is far more likely to hit the storage cap and convert to Google One. Third, it catches the trust failure mode: if the rate drops, it means Memories are being surfaced that feel irrelevant, intrusive, or poorly timed, which is the exact risk the product team must monitor as AI indexing deepens.

The supporting metric tree

AI quality and trust:

• Ask Photos query satisfaction rate: the share of queries where a user clicked a result and did not immediately retry with different wording. A retry signals the model returned something irrelevant.
• Memory dislike or hide rate: a direct signal that the AI surfaced something the user did not want to see. Counter-metric to Weekly Memories Engagement Rate.
• AI search toggle rate: what share of users switch from Ask Photos back to classic keyword search per session. A rising rate signals the AI layer is losing credibility.

Retention and habit:

• D30 and D90 retention, cohorted by user type: backup-only users (no downstream engagement with surfaced content) versus active-engagement users (open Memories, use Ask Photos, create shared albums). The cohort split matters because the two groups have very different conversion curves to Google One.
• Return visit rate excluding auto-backup opens: strips passive installs from the retention signal.

Monetization:

• Storage fill rate to 80% of free tier: the leading indicator of Google One conversion. When a user crosses this threshold and sees a storage prompt, their conversion probability spikes.
• Google One conversion rate from Photos storage prompt: the direct business outcome. This is a lagging indicator but must appear in any complete answer because it is the actual revenue mechanism the product exists to serve.

Sharing and social proof:

• Shared album creation rate and external share link clicks. Sharing distributes trust: when a user sends a Photos link to someone outside the ecosystem, it is a form of endorsement that can drive new user acquisition.

Counter-metrics

Counter-metrics protect the north star from being gamed or from hiding second-order damage:

• AI-generated content that gets immediately deleted: if Veo 2 Photo-to-Video clips are being created and then deleted within the same session, the feature is generating noise, not value.
• Memory notification opt-out rate: rising opt-out means the Memories surface is annoying rather than delightful. At scale this degrades the core mechanism the north star depends on.
• Storage upload rate with zero downstream engagement over 90 days: the backup-and-forget user segment. These users consume storage costs, will not convert to Google One (they have no emotional attachment to the product), and inflate MAU without contributing to the trust loop. Tracking this cohort separately prevents the aggregate numbers from obscuring churn risk.
• Privacy complaint volume related to AI indexing: a qualitative signal that the “knows you without surveilling you” balance has tipped. Not a hard metric, but it belongs in the answer because it names the product’s existential risk.

Strong vs weak

The PM judgment

The interviewer is checking whether you understand the product’s actual job in the Google ecosystem, not just its features. Google Photos at 1.5 billion MAU is only viable as a business if the trust it earns converts to Google One revenue. The lovable standard is harder: the product must be the place memories live, which means the AI layer must surface the right memory at the right moment without feeling like a surveillance tool. A strong answer names both the trust metric and the monetization metric, explains the tension between them, and picks a north star that sits at their intersection.

For the framework behind north-star selection, see north star metric. For a parallel metrics question with a different business model dynamic, see measure success of Instagram Stories. For when two metrics point in opposite directions, see two metrics conflict: engagement vs revenue.