Co-movement does not prove that experience metrics cause business outcomes

What it is

Experience and business metrics rising or falling together only says they co-moved in that period. Co-movement is not causation: “satisfaction and revenue moved together” does not entail “raise satisfaction and revenue will follow.” Common causes are ordinary—peak season, spend, price, a breakout item—and they lift both columns at once. Treat co-movement as cause, and later work will optimize the easier experience column, then feel betrayed by measurement when business does not follow.

Why it happens

Two series share a calendar. Marketing calendars, holidays, release trains, and acquisition mix all leave marks on the same axis, so many columns look in step. Correlation and synced charts do not distinguish: experience causing business, business causing experience (service improves after payment), or a third factor driving both. Experience work falls especially easily into the first story, because it fits professional identity. Composition also manufactures co-movement: when new users flood in, experience scores and revenue can rise together simply because those people have not yet hit deeper failures. Without contrast, without handling common causes, without a mechanism beyond order in time, co-movement supports only “they did not split in this period,” not “this column is why that column moved.”

Studying it

List every concurrent spend, price, version, and seasonal event in the co-moving period as common-cause candidates. Test whether an experience intervention moves the business column with contrast, an instrument, or at least a staggered ship—not with site-wide time-series correlation. In the report, keep “co-moved,” “led,” and “caused” as three grades of claim; a correlation chart must not write “therefore.” A negative control helps: take an experience change that, by mechanism, should not affect business; if the business column still co-moves, the co-movement is the calendar, not that experience.

Where it stops holding

On short paths with a clear mechanism and little interference (fixing an error that blocked payment), co-movement can be a weak hint of cause; contrast should still confirm. A stable long-horizon lead raises the prior for causation; it does not complete the proof. Rejecting co-movement as proof is not rejecting experience work; it only rejects co-movement as acceptance. When contrast is impossible, stop the claim at “common causes have not been separated,” rather than upgrading it to “experience has been shown to drive business.”

Applying it

• When two columns co-move, open a concurrent-event list first; only an empty list may enter a causal discussion.
• Experience proposals that claim business impact need contrast or an explicit quasi-experiment, not a site-wide correlation screenshot.
• Split “moved together” and “caused” into two kinds of sentence in external narrative; do not mix them.
• When common causes cannot be ruled out, accept the experience column against the experience goal and watch the business column separately; they do not prove each other.

Related

• Same group: Q6.08.2 Experience metrics with no corresponding business impact should be re-examined · Q6.08.3 The causal chain from experience metrics to business outcomes must be modeled, not assumed · Q6.08.4 Optimizing only business metrics lets experience decline silently before the problem surfaces
• Adjacent: Q6.04 Experience and business metrics · Q3.13 Statistical significance and practical significance
• Search terms: co-movement is not causation · confounding · UX-business correlation