Time spent is flat but Reels is growing — what's going on? — Meta Instagram Data Scientist Interview

Time spent is flat but Reels is growing — what's going on?

Diagnose a problemMedium

Problem. Over the last quarter, Instagram's Reels watch time grew nicely, but overall time spent on Instagram is flat. Your PM asks: is that a problem? Diagnose it.

Before you reveal: say your answer out loud, as if you were in the real interview — get your reasoning across clearly first. There is no single correct answer: reading what the interviewer is really after and defending your own thinking is what makes an answer strong.

Clarify the question first

"When we say time spent is flat, is that total across all surfaces, and measured the same way all quarter — no logging or attribution change between Reels and Feed?"
"Flat versus what — flat versus last quarter, or flat versus a growth target? Flat can be fine or alarming depending on the baseline."
"Do we care about time spent itself, or is it a proxy for engagement and revenue? That decides whether flat is actually bad."

Lay out your approach

The obvious hypothesis: cannibalization. Reels time is up but total is flat, which is the signature of Reels taking minutes from Feed and Stories rather than adding them. I'd confirm by decomposing time spent per user across Reels / Feed / Stories / Explore.
Distinguish substitution from addition. If Feed and Stories fell by about what Reels gained, it's a mix shift. If they held and Reels still didn't move the total, something else is capping total time (a ceiling, a different surface declining).
Weight by value, not minutes. If Reels currently monetizes at a lower rate than Feed, a flat total that shifted toward Reels is actually a revenue decline hiding behind a flat engagement number. That reframes "is it a problem."
Segment. New/casual users may be genuinely gaining on Reels while core users substitute.

Then check in: "My lead hypothesis is cannibalization — Reels pulling minutes from Feed/Stories — so I'd decompose time per user by surface and weight by monetization. Want me to work through the decomposition?"

Simulate the back-and-forth

Interviewer

Say the decomposition shows Feed and Stories down by roughly what Reels gained. Is that bad?

You

It depends entirely on value per minute. If a Reels minute monetizes and satisfies as well as a Feed minute, a mix shift with flat total is neutral — we moved people to a surface they prefer. But if Reels monetizes lower today, flat total time actually means falling revenue, which is a real problem even though the engagement headline looks calm. So I'd immediately pull monetization-weighted time, not raw minutes.

Interviewer

Reels does monetize lower right now. So it's a problem?

You

It's a problem for revenue this quarter, but not necessarily a strategic mistake. Short video is where attention is going industry-wide; defending Feed minutes by suppressing Reels would lose long-term. The right framing for the PM is a trade: we're accepting near-term monetization dilution to hold engagement on the surface users are migrating to, and the real work is closing the Reels monetization gap. I'd size that gap in dollars so leadership sees the actual stakes.

Interviewer

Anything that would change your read entirely?

You

Yes — segments. If the Reels growth is mostly new and casual users gaining time while core users are flat, Reels is additive at the top of the funnel and healthy. If core users are the ones substituting Feed for Reels, that's the monetization concern. Same flat total, different story, so I wouldn't conclude without the segment cut.

Land a clear conclusion

You

My read: flat total with Reels up is almost certainly cannibalization, so I'd confirm it by surface, then judge it on monetization-weighted time, not raw minutes, and split by user segment. The honest conclusion for the PM is likely "engagement is being defended on the surface users prefer, at a near-term revenue cost — the priority is closing the Reels monetization gap," with the gap sized in dollars. Flat isn't automatically bad; it's bad only if value per minute is dropping and we're not on a path to fix it.

Key summary — what really matters.

Flat total + one surface up = cannibalization until proven otherwise.
Decompose time per user across Reels / Feed / Stories.
Weight by monetization — a mix shift to a lower-value surface is a hidden revenue drop.
Segment (new/casual vs core) — it can flip the conclusion.
Reframe for the PM as a trade, and size the Reels monetization gap in dollars.

Surface decomposition. Minutes per user per day:

Surface	Last quarter	This quarter	Change	$ / min (illustrative)
Feed	18	15	−3	$0.010
Stories	10	9	−1	$0.008
Reels	12	16	+4	$0.006
Explore	5	5	0	$0.007
Total minutes	45	45	0

Raw time is flat. But monetization-weighted:

$\text{Value}_{\text{before}} = 18(0.010)+10(0.008)+12(0.006)+5(0.007) = \$0.367$

$\text{Value}_{\text{after}} = 15(0.010)+9(0.008)+16(0.006)+5(0.007) = \$0.353$

So flat engagement hides a −3.8% monetization move, because ~4 minutes shifted from higher-value Feed to lower-value Reels. That's the number the PM actually needs, and it reframes "flat is fine" into "there's a real, sized revenue leak driven by mix shift."

Sizing the gap. If Reels reached Feed's $0.010/min, the after-value would be $15(0.010)+9(0.008)+16(0.010)+5(0.007) = \$0.417$ — closing the Reels monetization gap is worth more than defending Feed minutes, which is the strategic point.

Dimension	What interviewers look for	Below the bar ❌	Above the bar ✅
Problem framing	Whether you ask if "flat" is even bad, and versus what.	You assume flat is automatically a problem or automatically fine.	You tie it to value/revenue and a baseline before judging.
Methodology and math	Whether you decompose by surface and weight by value.	You look at total time only.	You split per-user time by surface and monetization-weight it.
Business judgment	Whether you see the strategic trade in short video.	"Reels is cannibalizing, kill it."	You frame near-term dilution vs long-term attention migration.
Deployment thinking	Whether you segment and act on the finding.	One global number.	New/casual vs core split; size the monetization gap in dollars.
Communication	Whether the PM gets a decision, not a data tour.	You describe the tables.	You deliver "engagement defended at a revenue cost; close the Reels gap."

Interview tips

Strong-answer signal: spotting cannibalization, converting flat minutes into a monetization-weighted revenue read, and framing it as a strategic trade with the gap sized. Common trap: declaring "flat time spent is fine" (or "is a disaster") without decomposing by surface and weighting by value.