Shift concentration
How concentrated the shifting is: largest shift over total shift.
At a glance the headline numbers for Shift concentration
How concentrated the shifting is: largest shift over total shift.
On the typical site, the largest single shift causes 1 of all layout shift.
Distribution & median CLS site count and median CLS at each level of shift concentration
Passing CLS by shift concentration which level passes the CLS most often
Shift concentration 1.0. p75 1. p99 6.2. At the low end (0–0): CLS 0.00. At the high end (1+): CLS 0.00. computed
Why this matters for the Core Web Vitals, and where to start fixing it
Two sites can have the same CLS for opposite reasons. One has a single big jump. The other moves a little, everywhere, all the time. This ratio separates them: the largest shift divided by the total. Close to one means one event causes nearly all the damage. Find that element, reserve its space, and the score follows.
A low concentration is the harder case. No single shift looks bad, but they add up. That pattern points at scroll-triggered animations, CSS transitions on layout properties, or content that streams in piece by piece. There is no one bug to fix, so it needs a structural pass instead of a patch.
How does this affect the Core Web Vitals?
Shift concentration is part of the CLS itself, so this is arithmetic, not correlation. Where the shift concentration is low: 84% pass the CLS. Where it is high: 91%.
Chrome field data from 189,915 sites, representing millions of real page loads. How we measured.
Live queries (2) — admin only
SELECT COUNT(*) AS count,
quantile_disc(m.cls.concentration, 0.10) AS p10,
quantile_disc(m.cls.concentration, 0.25) AS p25,
quantile_disc(m.cls.concentration, 0.50) AS p50,
quantile_disc(m.cls.concentration, 0.75) AS p75,
quantile_disc(m.cls.concentration, 0.90) AS p90,
quantile_disc(m.cls.concentration, 0.99) AS p99
FROM sites WHERE m.cls.concentration IS NOT NULL;SELECT CASE WHEN m.cls.concentration >= 0 AND m.cls.concentration < 1 THEN 0 WHEN m.cls.concentration >= 1 THEN 1 END AS bucket_idx, COUNT(*) AS n,
quantile_disc(crux."phone".cls, 0.5) AS median,
COUNT(*) FILTER (WHERE crux."phone".cls IS NOT NULL AND crux."phone".cls <= 0.1) * 1.0 / NULLIF(COUNT(*) FILTER (WHERE crux."phone".cls IS NOT NULL), 0) AS good_pct
FROM sites WHERE m.cls.concentration IS NOT NULL GROUP BY bucket_idx;