rca · hard
RCA: Instagram MAU is flat but DAU is down
Instagram MAU is flat but DAU has been declining for the past several weeks. Walk me through how you would diagnose this.
Most candidates treat “MAU is flat” as background context. The strong answer treats it as a logical eliminator: if users hit the monthly threshold, they have not churned. The diagnostic space shrinks immediately, and the candidate who states this out loud in the first minute signals they can impose structure on ambiguous data.
What the divergence actually tells you
MAU flat means every user counted this month logged in at least once in the 30-day window. That rules out three entire hypothesis classes before any data pull: acquisition pipeline failure (new users are still arriving or existing users are still returning), hard technical blockers severe enough to prevent login, and total churn. What remains is a frequency problem. Users are visiting, but less often. The question is why daily visit rate dropped while monthly visit rate held.
This framing collapses the brainstorm. You are not diagnosing “something wrong with Instagram.” You are diagnosing “something that reduced the number of days per week a user opens the app, without being severe enough to stop them opening it at all.”
Structure a strong answer
strong
"Before I hypothesize user behavior, let me clarify the scenario and then check measurement. On clarification: is the DAU decline a sudden cliff or a gradual slope over weeks? Is it global or concentrated in a specific market or platform? And what is the DAU definition here, unique openers or users who perform a qualifying action like a like, comment, or post? The shape and scope of the decline tells me whether to look at a product change, an external event, or a measurement artifact.
On measurement: I want to rule out a phantom divergence before touching behavioral hypotheses. Three checks. First, did the iOS or Android SDK update during this period? A logging regression can undercount daily opens while monthly counts stay stable because monthly aggregation tolerates gaps. Second, was there a bot or spam account purge? Purging inflates real MAU while reducing counted DAU if bots had high daily activity but humans count toward MAU at a lower rate. Third, did the DAU definition or time-zone bucketing change? A definition tightening reduces DAU without affecting MAU. Meta does not publish Instagram DAU as a standalone figure. It reports Family Daily Active People across Instagram, Facebook, WhatsApp, and Messenger, roughly 3.27 billion as of Q1 2026. Any Instagram DAU we are working with is an analyst estimate in the 500 to 700 million range. I want to flag that the number itself may carry measurement uncertainty.
Assuming measurement is clean, I decompose the frequency problem in two ways. First, breadth: is the share of monthly users who open the app on any given day shrinking, meaning fewer unique users cross the daily threshold each day? Second, depth: are the same users opening the app but not crossing whatever action threshold defines 'active' in our DAU definition? These require different fixes. Breadth down means the re-engagement trigger is broken. Depth down means the session is not generating qualifying actions.
My highest-likelihood hypothesis given a gradual slope and clean measurement: habituated mid-tenure users (90 to 180 days on platform) are shifting from daily to every-other-day use because casual text-based check-in behavior has migrated to Threads. Threads reached 300 million MAU by 2025. A user who previously opened Instagram to post a quick status or check what friends are saying now opens Threads for that. They still open Instagram for photos, Reels, and DMs, which keeps them in MAU, but they do it three days a week instead of seven. This is within-Meta cannibalization, not competitive loss, and I can test it without running an experiment: pull cross-app behavioral data for users whose Instagram daily frequency dropped and check whether their Threads opens increased in the same period.
Two secondary hypotheses worth naming. Reels recommendation degradation: if the recommendation system is serving lower-relevance content, users feel done faster and return less often. This would show as flat or declining session length alongside the DAU drop. If time-per-session is stable, this hypothesis weakens. If time-per-session also dropped, it strengthens. Second, notification strategy change: Instagram's daily return rate is heavily driven by re-engagement notifications. A policy change, delivery failure, or iOS permission revocation at scale could reduce the daily trigger without affecting monthly reach.
At 3 billion MAU, a 1% shift in daily return rate is 30 million users. The scale makes both real behavioral shifts and measurement artifacts plausible at the margins. I would sequence the investigation: measurement check first, then cross-app Threads data pull, then session-length analysis by cohort tenure, then notification delivery audit. The first two data pulls require no experiment and can be done in days. If Threads cannibalization is confirmed, the recommendation is not to defend Instagram's session count at the expense of Threads growth. The right product response is to evaluate whether a cross-surface daily digest or integrated content experience serves both products, because Meta's business-level goal is Family DAP, not Instagram DAU isolation."
weak
"It could be a bug, or maybe the algorithm changed, or a competitor is taking users, or it's seasonal, or there are privacy concerns." The candidate names eight hypotheses with no prioritization, no connection to the MAU-flat constraint, and no measurement check. When pushed to pick one, they say they would look at all of them. They skip frequency decomposition entirely and move to solutions. This is the brainstorm-dump failure: it tells the interviewer the candidate cannot reduce ambiguity, only expand it.
The 2026 framing interviewers look for
In 2026, “Instagram DAU is down” is not automatically a product crisis. Meta operates Instagram, Threads, Facebook, WhatsApp, and Messenger as a coordinated family. A user migrating daily check-in behavior from Instagram to Threads shows up as Instagram DAU down and Threads DAU up, while Meta Family DAP stays flat or grows. The interviewer at a Meta-caliber company is specifically checking whether you distinguish the product-level metric (Instagram DAU) from the business-level metric (Family DAP), because the fix and the urgency are completely different. Defending Instagram’s session count by suppressing Threads could be the wrong call if Family engagement is healthy. The candidate who raises portfolio-level thinking rather than metric-defense instinct is the one who clears the bar.
Scoring breakdown
Meta analytics rounds in 2026 weight this question roughly as: structured decomposition 30%, hypothesis prioritization with stated reasoning 30%, measurement validity check 20%, proposed next steps 20%. The measurement check is the highest-impact move because most candidates skip it, and checking it first signals that you know data can lie before users can.