Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Glossary term

Mix Shift

A change in a blended metric caused by a change in the mix of segments, such as devices or sources, rather than by any segment performing differently.
AnalyticsconceptUpdated

Definition

A blended metric is a weighted average. A site's conversion rate, for example, is the conversion rate of each segment of traffic weighted by that segment's share of sessions. It can change for two reasons: the rates within the segments changed, or the shares changed. A mix shift is the second case. The resulting movement in the blended figure is called the mix effect; movement caused by segments performing differently is the rate effect.

The same logic applies to any ratio built from groups: cost per conversion across campaigns with different goals, average order value across product categories, revenue per customer across plans. In its extreme form, every segment can improve while the blend gets worse, a pattern statisticians call Simpson's paradox.

In practice: decomposing a change into rate and mix effects

The guide to explaining metric changes sets out the arithmetic, which fits in a spreadsheet. Choose segments defined identically in both periods, record each segment's share and rate in each period, and calculate two totals: the mix effect from the changes in share, and the rate effect from the changes in rate. Together they account for the change in the blended rate. A quicker check is to reweight: apply the earlier period's shares to the later period's rates. If the result sits close to the earlier blended rate, the segments held steady and the traffic moved between them.

Segment choice decides what the decomposition can see. Device and channel are the usual starting point; the e-commerce conversion hub adds new against returning visitors, product category and landing page type, each of which can carry its own mix shift. The conversion rate calculator handles the per-segment arithmetic once the segments are chosen.

Why a mix effect is not a performance problem

A mix effect describes where results came from, not how well anything worked. A campaign that brings a large number of colder visitors lowers the blended conversion rate while adding orders; treating the fall as a conversion problem leads to a checkout redesign nobody needed, or to cutting the campaign that added sales.

Mix also creeps into figures that look like single measures. The guide to conversion windows and attribution warns that a campaign whose goals mix actions under different attribution models has a cost per conversion that is not comparable month to month if the mix shifts. In pricing tests, the pricing page layout experiment reports the plan mix beside the primary metric, because a layout can move buyers to a cheaper plan. And Goodhart's law explains why mix shifts towards easy cases once a blended number becomes a target.

Illustrative example — not a real result

An office chair seller's conversion rate falls after a social campaign

Swivelmarch Office Chairs saw its site conversion rate fall from 2.4% to 1.9% in the month a social campaign ran, and the team began planning a product page review. The analyst split both months by device and source. Desktop search visitors converted at 3.6% before and 3.6% during; phone visitors from social converted at 0.8% before and 0.8% during; every other segment was within a tenth of a point. What changed was the share: social phone visitors rose from 12% of sessions to 40%. Reweighting the campaign month's rates to the earlier shares gave 2.4%, the same as before, and orders had risen from 1,150 to 1,330. The product page review was dropped.

CVR is the blended metric where mix shifts are most often misread. Segmentation supplies the groups a change is decomposed into, and the decomposition is only as good as their definitions. Channel Grouping defines one of the most common segment sets, so a change to the grouping rules can itself look like a mix shift. A blended rate reported without its segments drifts towards the Vanity Metrics it was meant to avoid, because it can move without anything meaningful changing.

Common confusions

  • A falling blended rate means something broke. Check the segments first; if each holds, the mix moved.
  • Mix effects are noise. They are real changes in where traffic or customers come from, often caused by the team's own campaigns.
  • Any split will reveal it. A mix shift is visible only in the dimension whose shares changed; the wrong split hides it.
  • Mix and rate effects are exclusive. Both usually happen at once; the decomposition says how much of each.
  • Segments must be redefined to match each period. They must be defined identically in both, or the comparison mixes definitions with behaviour.

More terms to look up

Calculators

Where this term is used

Related