Who Was the Click For? Action Address as a Missing Variable in Digital Brand Analytics

Digital brand analytics records what a user acted upon, but not the consequence the user expected the action to produce. A heart on a brand post may evaluate the brand, tune a recommender toward a desired category, preserve an item for later, or address other people. This paper theorizes action address: the perceived locus or loci of consequence whose response, state change, or future availability an actor treats as instrumental to a platform action. That users strategize toward recommenders is well established, including by randomized experiment (Cen et al., 2024). We are not aware of work asking what this does to a brand inference: whether a trace logged against Brand A keeps its relation to affinity for Brand A when the operative consequence moves from brand feedback to category-level recommendation. Two things follow. First, an identification result. Let ω be the share of action opportunities governed by a non-exact-brand address regime. Given observed action rates and a ceiling on ω, the exact-brand contrast is sharply bounded — an application of contaminated-sampling bounds (Horowitz & Manski, 1995) — and its sign is identified if and only if that ceiling falls below the absolute observed contrast. A four-rung ladder shows how experimental estimates of the recommender regime can tighten the interval under three explicit assumptions — an assignment to-perceived-address bridge, a composition assumption, and lab-to-field transport — and a projected confidence region carries sampling error through it. Second, a design. We are not aware of a study that randomizes the downstream consequence of a brand-visible action, enacts it, and locks an independent brand criterion before any consequence is delivered. A construct-discovery study and a three-arm experiment crossed with instruction strength are specified at a scale one group can run, with instruction with held rather than supplied in the arm that carries the argument.