T6

See the scoped evidence_register and qualification_protocol. Structural plausibility, descriptive surveys and laboratory benchmarks do not alone identify a local commercial policy effect.

See the scoped evidence_register and qualification_protocol. Structural plausibility, descriptive surveys and laboratory benchmarks do not alone identify a local commercial policy effect.

A convolution describes a proposed distributed response to earlier activity. A linear stock with decay has an equivalent exponential convolution solution. Last-touch credit is not incremental contribution. Self-reported attribution can suggest discovery mechanisms but does not identify a lag kernel or causal brand effect by itself.

• Wei, Bencina, Nguyen, Ribeiro and Zelditch — From Prediction to Incrementality — 10 August 2026. Full preprint. Field intervention; full methods/limitations inspected. Eight-week member-randomized 50/50 LinkedIn comparison of causal scoring, exploration, constrained allocation and withholding versus propensity targeting, with a matched delivery envelope. The reported value lift concerns the complete policy. Static audience definitions and segment lineage protect assignment. Source: https://arxiv.org/html/2608.10182v1
• Li, Yuan, Yang, Chen and Song — Attributed But Not Incremental — 25 June 2026. Full preprint. Production calibration against 18 rounds of channel experiments across up to eight TikTok markets, with rolling forward evaluation and daily hierarchical correction. Source: https://arxiv.org/html/2606.26690v1
• Heusch — Structural Estimation of MMM Parameters from Geo Experiments — 21 August, revised 7 September 2026. Full v2. Synthetic structural study: 156-week fictional retailer, three channels, four four-week go-dark tests plus cooldown. Uses time paths to estimate effectiveness, carryover and saturation. Source: https://arxiv.org/html/2608.21128v2
• Filippou, Quach and Jha — Pseudo Incrementality Testing — 16 September 2026. Full preprint. Changepoint detection plus counterfactual forecasting, two GeoLift benchmarks and synthetic stress tests. Only one of three nominal 90% intervals covers truth in the stress test; weak controls yield large bias. Source: https://arxiv.org/html/2609.18257v1
• Dubé, Morozov, She and Tuchman — The Effects of Ads on Beliefs and Implications for Consumer Search — written 31 July, posted 7 August, BFI release 17 August 2026. Paper and full PDF. Preregistered experiment, 17,775 Prolific participants, simulated bookstore; randomized ads/content/price and separate belief/shopping samples avoid priming. Data are from 2025. Source: https://bfi.uchicago.edu/working-papers/the-effects-of-ads-on-beliefs-and-implications-for-consumer-search/

• No fitted local coefficient or complete-policy deployment evidence is supplied.
• one platform, short experiment, no isolation of individual components, no autonomous B2B seller economics. C3/C5/C6.
• normalized confidential errors; child-level credit is allocated rather than independently causally identified. A later cannibalization decline is before/after deployment evidence, not a randomized business-impact estimate. Complementarity and spillover remain limitations. C3/C5.
• stable parameters, specified adstock/saturation, no spillover and tested spend support. It is not an empirical estimate of SaaS brand decay. C3/C5.

Last-touch credit is not incremental contribution. Self-reported attribution can suggest discovery mechanisms but does not identify a lag kernel or causal brand effect by itself.