← Back to Blog

How Ex-Meta Engineers Think About Creator Metrics: Retention, Watch Time, and What Really Matters

When I joined Meta's Ads Data Foundation team, I thought I understood metrics. I'd built data pipelines, tracked KPIs, analyzed dashboards. But working on systems that processed 500B+ events per day taught me something different: most people are tracking the wrong things.

Now, building tools for creators, I see the same mistakes everywhere. Creators obsess over follower counts and likes while ignoring the metrics that actually predict success. Let me show you how we thought about metrics at Meta and how that applies to creator content.

The Hierarchy of Metrics

At Meta, we organized metrics into three tiers:

Tier 1: North Star Metrics

  • The single metric that best predicts long-term success
  • Usually retention or revenue-related
  • Changes slowly but matters most

Tier 2: Leading Indicators

  • Metrics that predict changes in North Star
  • Change faster, give you early signal
  • Where you focus day-to-day optimization

Tier 3: Health Metrics

  • Track quality and sustainability
  • Prevent you from gaming the system
  • Warning lights, not steering wheel

For creators, here's what this looks like:

North Star: Expected watch time per video (retention % × video length)

Leading Indicators: 3-second retention rate, click-through rate on thumbnails, share rate

Health Metrics: Follower growth rate, comment sentiment, unfollow rate

Most creators track only Tier 3 metrics (followers, likes) and wonder why their account isn't growing.

Why Watch Time > Everything Else

At Meta, we spent months debating which metric best predicted content success. We tested:

  • Total views
  • Like rate
  • Comment rate
  • Share rate
  • Completion rate
  • Average watch time

The winner? Expected watch time.

Here's why: watch time is the currency of social platforms. Instagram, TikTok, YouTube—they all optimize for keeping users on the app. Content that delivers more watch time gets more distribution.

But here's the nuance: it's not just raw watch time. It's expected watch time per impression.

The Math That Matters

Let's say you post two videos:

Video A: 60 seconds long, 20% average retention = 12 seconds watch time Video B: 15 seconds long, 70% average retention = 10.5 seconds watch time

Which one should the algorithm promote?

Video A delivers more watch time per view (12s vs 10.5s), so it gets more distribution. But Video B has better engagement rate (70% vs 20%), which signals quality.

The algorithm balances both. In testing thousands of videos, I found:

  • Short videos (15-30s) need 50%+ retention to compete
  • Medium videos (30-60s) need 35%+ retention
  • Long videos (60s+) need 25%+ retention

The key insight: longer videos can win with lower percentage retention if they deliver more total watch time.

Retention Curves Tell the Real Story

At Meta, we didn't just track average retention. We tracked the retention curve—second-by-second drop-off.

Here's what healthy vs. unhealthy curves look like:

Healthy Retention Curve

100% |●
 80% | ●
 60% |  ●
 40% |    ●
 20% |      ●
  0% |________●
     0s  5s  10s 15s

Gradual decline, people leaving at natural breakpoints.

Unhealthy Retention Curve

100% |●
 80% | 
 60% | 
 40% |●
 20% |  ●●●●●
  0% |________●
     0s  5s  10s 15s

Cliff at 3-5 seconds, then flat. Hook failed.

When I analyze a creator's content, I look for:

  1. The Hook Cliff - Drop in first 3 seconds. Should be <60%.
  2. The Mid-Roll Plateau - Stable retention 5-15s in. Good sign.
  3. The End Spike - Slight increase near end. People rewatching or the call-to-action working.

You can see these curves in Instagram Insights or TikTok Analytics. Most creators never look at them.

Leading Indicators: What to Watch Daily

Average watch time and retention are lagging indicators. They tell you what happened yesterday. To improve, you need leading indicators that predict tomorrow.

1. First-Second Retention Rate

Percentage of people who watch past the first second.

At Meta, we found this was the single best predictor of overall retention. If you lose 70% in the first second, you'll lose 90% by second 5. If you keep 60% in the first second, you'll keep 35% to the end.

How to track it: Most platforms don't show first-second data, but you can estimate it:

  • Look at your retention curve at 0s vs. 1s
  • The drop is your first-second loss rate

What good looks like:

  • Great: >60% retention after 1 second
  • Good: 40-60% retention
  • Needs work: <40% retention

2. Shareability Rate

Shares per view, not total shares.

At Meta, shares were worth 10x more than likes in the algorithm. Why? Shares = distribution. When someone shares your video, you're reaching their audience, not just yours.

How to track it: Shares / Total Views × 100

What good looks like:

  • Great: >3% share rate
  • Good: 1-3% share rate
  • Average: 0.3-1% share rate

Shareability is the metric that takes you from 10K to 100K followers. Retention keeps you growing, shares make you explode.

3. Click-Through Rate (CTR) on Feed Impression

How often people click your video when they see it in their feed.

This is pre-watch behavior. High CTR means your thumbnail/first frame is working. Low CTR means people scroll past before the video even plays.

How to track it: Most platforms don't expose this directly, but you can infer it:

  • If your views are low but retention is high = CTR problem
  • If your views are high but retention is low = content problem

What to test:

  • Different first frames
  • Different thumbnail text
  • Different visual contrast
  • Different facial expressions

Health Metrics: The Warning Lights

These metrics don't tell you what to optimize, but they warn you when something's breaking.

1. Follower Efficiency

New followers per 1,000 views.

At Meta, we tracked "conversion rate" for every action. For creators, followers are the conversion.

How to track it: New Followers This Week / (Total Views This Week / 1000)

What good looks like:

  • Great: >5 new followers per 1K views
  • Good: 2-5 per 1K views
  • Needs work: <2 per 1K views

If this drops, your content is getting distribution but not converting to followers. Usually means:

  • Content is viral/entertaining but not "followable"
  • No clear reason to see more from you
  • Missing call-to-action or profile optimization

2. Unfollow Rate

People leaving after they follow.

How to track it: (Unfollows This Week / Total Followers) × 100

What good looks like:

  • Great: <0.5% unfollow rate per week
  • Good: 0.5-1% per week
  • Problem: >1% per week

High unfollow rate means:

  • Inconsistent content (they followed for X, you're posting Y)
  • Posting too frequently
  • Quality decline

3. Comment Sentiment

Not comment count—sentiment.

At Meta, we had ML models to detect comment sentiment. You can do this manually: read 20-30 comments, tag them positive/negative/neutral.

What good looks like:

  • Great: >60% positive sentiment
  • Good: 40-60% positive
  • Problem: <40% positive or lots of "I don't get it" comments

Negative sentiment kills distribution. The algorithm can detect it and will stop showing your content.

Common Pitfalls: How Creators Game Themselves

Pitfall 1: Optimizing for Vanity Metrics

I see creators celebrating 1M views on a video, but the retention was 12% and they gained 200 followers. That video hurt their account more than helped it.

Why? The algorithm learned that people don't want to see your content. Next video gets less distribution.

Better approach: Celebrate high retention, even on lower view counts. That's what compounds.

Pitfall 2: Chasing Viral Over Consistent

Viral videos are lottery tickets. They're great when they hit, but you can't build a strategy around them.

At Meta, we distinguished between:

  • Viral content: Breaks out to new audiences, high shares, low follower conversion
  • Core content: Performs well with existing audience, high retention, high follower conversion

You need both, but if you optimize only for viral, you'll have a feast-or-famine account.

Pitfall 3: Not Segmenting by Content Type

All your videos aren't the same. Track metrics by content category.

Example from a tech creator I work with:

  • Tutorials: 45% avg retention, 4% share rate, 6 followers per 1K views
  • Hot takes: 32% avg retention, 8% share rate, 2 followers per 1K views
  • Storytelling: 58% avg retention, 3% share rate, 8 followers per 1K views

Each format has a purpose. Tutorials build authority, hot takes get distribution, storytelling converts followers. You need all three, but measuring them together hides the insight.

How to Actually Use This

Here's the weekly routine I recommend:

Monday: Review Last Week's Data

  • Calculate expected watch time per video
  • Identify highest and lowest retention videos
  • Tag each video by content type

Wednesday: Deep Dive One Video

  • Pull retention curve for your best performer
  • Identify what worked (hook, mid-roll, end)
  • Screenshot and save to a swipe file

Friday: Plan Next Week

  • Choose content types based on last week's performance
  • Test one new hook style based on retention data
  • Set a retention target for each video

Monthly: Check Health Metrics

  • Follower efficiency trending up or down?
  • Unfollow rate acceptable?
  • Comment sentiment still positive?

The Meta Framework: Instrument → Observe → Change → Review

At Meta, we had a mantra: IOCR (Instrument, Observe, Change, Review).

Instrument: Set up tracking for the metrics that matter Observe: Watch for patterns over time Change: Test one variable based on what you learned Review: Did it work? Why or why not?

Most creators skip straight to "Change" without Observing. They post randomly, see random results, learn nothing.

The creators who grow consistently are the ones who treat content like experiments. They form hypotheses ("I think shorter hooks will improve retention"), test them, measure results, and iterate.

That's how we operated at Meta with billions of users. It works the same with 1,000 followers.

What to Do This Week

  1. Find your retention curves. Instagram and TikTok both show them in Insights. Look for the patterns.

  2. Calculate expected watch time. For your last 10 videos, multiply retention % by video length. Which format delivers the most?

  3. Track shares per view. This is probably your most underrated metric.

  4. Set up a simple spreadsheet. Date, Video Title, Length, Retention %, Watch Time, Shares, Followers Gained. Fill it in weekly.

  5. Focus on one metric to improve. Don't try to optimize everything. Pick the weakest link (probably hook retention) and improve it by 10% over the next month.

The difference between stagnant and growing creators isn't talent or luck. It's knowing which metrics matter and actually tracking them.

Most creators track nothing. If you track the right things, you're already ahead of 90% of the competition.

← Back to All Articles