How to Read Video Metrics in Meta Ads Manager (Thumbstop, Hold & ThruPlay)

In Ads Manager → Ads (creative-level view works best for diagnostics):

Save the preset. The goal is repeatable creative reviews — not one-off exports when an ad "feels" weak.

Step 2: Create custom hook and hold metrics (recommended)

Meta does not ship thumbstop or hold as default columns. Vaizle and Madgicx both document the same setup path — Columns → Customize columns → Create custom metric:

Use Number format (not Percentage) for these custom metrics, or Ads Manager double-multiplies the ratio. The ThruPlay-funnel hold (ThruPlays ÷ 3-second views) reads ~40–50% — materially higher than the 15s÷3s definition — for the same creative.

Step 3: Compute the three diagnostic rates

Meta's raw columns map to practitioner metrics like this:

A ThruPlay counts when a viewer watches a video to completion (under 15 seconds) or reaches at least 15 seconds of view time (for longer videos). It does not count ad clicks — a widespread misconception. Practitioners compute thumbstop as 3-second views ÷ impressions — Meta exposes the raw column but not always the rate. On the 15s÷3s definition, the cold-traffic hold band is ~15–25%, with a kill floor below ~10%.

Use the free Hook Rate Benchmark and ThruPlay Rate Benchmark tools to compare your computed rates against 2026 segmented ranges — or see the consolidated Video Creative Metrics Benchmarks 2026 guide.

Step 4: Segment like a benchmark report would

Public benchmarks fall apart when you skip segmentation. Before declaring a creative "bad," lock:

• Length bucket — sub-15s, 15–30s, 30–60s, 60s+ (ThruPlay mechanics differ wildly)
• Placement — Reels/Stories vs Feed (autoplay and screen real estate differ)
• Audience temperature — cold prospecting vs retargeting (warm audiences inflate engagement)

Our ThruPlay Rate Interpretation guide walks the diagnostic sequence when aggregate numbers look fine but ROAS does not.

Step 5: Connect metrics to creative tests

Video diagnostics tell you where to iterate; they do not replace experiment design. When you promote a hook variant:

1. Run a controlled test (Meta creative testing playbook)
2. Pre-define sample size for the metric you are optimizing (impressions for thumbstop, conversions for ROAS)
3. Document hook vs body learnings separately so the next brief inherits evidence

Step 6: Read the retention curve (not just the headline %)

Hold rate is one number; the per-second drop-off curve is the diagnostic. adlibrary.com describes three recurring shapes:

• Cliff at 3–4s — hook resolves but the pivot to the body is too sharp; bridge tension for one second before introducing the product
• Slow bleed 5–15s — cuts too long; insert pattern interrupts at ~7s and ~11s
• Cliff at 14–15s — strong build, weak offer framing at the CTA

Pull video_p25, p50, p75, and p100 watched actions from Ads Manager (or your analytics stack) and chart second-by-second drop-off. Two ads with identical 18% hold can need opposite edits — optimize the curve, not the aggregate.

Quick reference: 2026 anchor bands (cold prospecting, 15–30s, Reels/Stories)

These anchors come from the same cited sources as our benchmark research hub — always compare within your segment:

• Thumbstop: ~20–30% Feed, ~25–40% Reels/Stories; practitioner tiers in Sepia Lab’s 2026 hook roundup
• Hold (15s ÷ 3s): ~15–25% cold; below ~10% → structural mid-roll problem before scaling
• Hold (ThruPlays ÷ 3s): ~40–50% average if your dashboard uses Vaizle’s formula — not comparable to the row above
• ThruPlay: ~18–28% on 15–30s cold Reels/Stories; strong above ~25%, hook problem below ~10%