Every number Kinect reports starts with an A/B split

First, the split. Your store runs behind an A/B test against its own baseline from day one, and nothing is declared until it is statistically significant. Second, attribution. An order is Kinect-attributed only when the shopper sent a message or clicked a recommended product before buying. Widget presence alone never counts, which removes the most common inflation trick in this category.

Brands with Kinect see 3–6% more revenue, measured against their own baselines. That is the store-level read from the split. Separately, 6–21% of shoppers who engage with Kinect go on to purchase, which is multiples of a typical site average. That second figure is engaged-cohort behavior, presented as such. Shoppers who choose to engage were always likelier to buy, so cohort conversion is never presented as store-wide causal lift. The published case studies keep the same discipline, engaged-cohort framing on every stat.

Guaranteed lift percentages. Kinect considers guaranteed CVR claims a category-wide bad habit and does not publish them. Behind the answers themselves, Kinect runs hundreds of evals daily, grading its agents' responses against real store conversations. The full methodology is public at how we measure, and it is written to go into a board deck as-is.