Three results: 79% ROAS, 16% external sales and almost 10% of brand equity

On 7 October, Circana and Best Buy Ads published new headline findings from a marketing mix model covering leading consumer-electronics and appliance brands. Circana’s estimate puts Best Buy Ads ROAS 79% above the other media studied when total U.S. Best Buy sales are counted. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026]

The model also attributed almost 10% of measured brand equity to Best Buy Ads. Another result extends beyond the retailer: nearly 16% of sales the model associated with Best Buy Ads occurred at competing sellers. Circana and Best Buy describe this as a halo effect — one retailer’s media influencing category demand across the wider market. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [2 · Best Buy · initial research findings at The Drop, 28 September 2026]

MediaPost previously reported that Circana analysed two years of sales, four major categories and $2.3 billion in media spend against the largest national platforms. That is broader than a click-and-conversion report within one retailer, but Best Buy commissioned the analysis and neither the full model nor the data are public. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [3 · MediaPost · Best Buy Ads pushes further into commerce media, 24 September 2026]

79% does not replace the earlier 17%: the denominator changed

On 28 September, Best Buy framed the finding differently: network ROAS was 17% above the national average, while 17% of media spend generated 35% of measured media contribution to Best Buy sales. The new release compares Best Buy Ads with “all other media” and counts total U.S. Best Buy sales. These are different denominators and possibly different views of the same model. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [2 · Best Buy · initial research findings at The Drop, 28 September 2026]

The gap between 17% and 79% therefore cannot be presented as campaign improvement over ten days or a correction of the earlier estimate. Comparison requires common definitions: which channels make up the national average and “other media,” whether the return is revenue or margin, whether carryover windows match and whether competitor sales are included. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [2 · Best Buy · initial research findings at The Drop, 28 September 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017] [5 · Google Research · Bayesian Methods for Media Mix Modeling, 2017]

Absolute ROAS is not disclosed. If one channel hypothetically returned $2 and another $3.58 in revenue per dollar, the relative difference would be 79%, yet economics after cost of goods, discounts and fees could differ. Without base levels, marginal ROAS and profit, an advertiser cannot reproduce a budget decision. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017] [5 · Google Research · Bayesian Methods for Media Mix Modeling, 2017]

MMM broadens the market view but does not replace a causal test

MMM can jointly estimate multiple channels, seasonality and delayed effects from aggregate data. Its weakness is dependence on functional form, variables, prior assumptions and correlation between spending and expected demand. Circana’s public description provides no specification, robustness tests or uncertainty intervals. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017] [5 · Google Research · Bayesian Methods for Media Mix Modeling, 2017]

Competitor sales may represent real category growth, but alternatives remain: popular brands can receive more advertising and sell better everywhere, while national promotions, availability and product launches move both series. The model may control for these factors, but that cannot be checked from the release. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [3 · MediaPost · Best Buy Ads pushes further into commerce media, 24 September 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017]

A practical validation would use pre-planned geo experiments or other control tests of incremental ROAS, then compare them with the MMM forecast. Gross profit, returns and category-level results are also needed: revenue at Best Buy and competitors does not by itself establish profitable new demand. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [6 · Annals of Applied Statistics · Robust Causal Inference for Incremental ROAS, 2022]

Expert commentary

The release matters because it tries to measure retail media across the market rather than inside one checkout system. For technology manufacturers, that is closer to the real question: does the network create brand demand or merely redirect a purchase from another seller? The 16% external-sales estimate surfaces an effect that closed-loop Best Buy attribution would miss. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [2 · Best Buy · initial research findings at The Drop, 28 September 2026]

For Best Buy, the result cuts both ways. The network can compete for national brand budgets if it helps sales everywhere. Yet competitors receive part of the created value while Best Buy pays for data, content and inventory. Its commercial model must monetize that external effect through media pricing rather than assume all returns will flow back into retail margin. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [2 · Best Buy · initial research findings at The Drop, 28 September 2026] [3 · MediaPost · Best Buy Ads pushes further into commerce media, 24 September 2026]

MMM is appropriate for this scale, but the word “model” is essential. It constructs a counterfactual from observed time series; it does not randomly switch advertising off for a comparable group. Research shows estimates are sensitive to specification, saturation, carryover and priors. Without those details, 79% is an analytical-system output, not a reproducible fact. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017] [5 · Google Research · Bayesian Methods for Media Mix Modeling, 2017]

Almost 10% of measured brand equity should not be confused with customer loyalty either. The release does not define the measure, how long the effect lasts or whether it connects to repeat purchases, retention or willingness to recommend. In technology categories with long replacement cycles, short-term consideration and an enduring customer relationship are especially easy to conflate. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [2 · Best Buy · initial research findings at The Drop, 28 September 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017]

One plausible reason for strong ROAS is proximity to purchase intent. That helps conversion, but also creates selection risk: ads appear where demand is already formed. Calibrating the MMM against randomized geo tests would help separate incremental demand from successful capture of existing intent and test the claimed halo at competitors. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017] [6 · Annals of Applied Statistics · Robust Causal Inference for Incremental ROAS, 2022]

Over the next 6–12 months, credibility would improve with uncertainty ranges, absolute and marginal ROAS, results by category and out-of-sample validation. Budget decisions need incremental gross profit more than attributed revenue. For now, the study usefully widens the measurement frame, but its headline percentages are best treated as hypotheses to test rather than universal benchmarks. [1 · Circana · new MMM findings for Best Buy Ads, 7 October 2026] [3 · MediaPost · Best Buy Ads pushes further into commerce media, 24 September 2026] [4 · Google Research · Challenges and Opportunities in Media Mix Modeling, 2017] [5 · Google Research · Bayesian Methods for Media Mix Modeling, 2017] [6 · Annals of Applied Statistics · Robust Causal Inference for Incremental ROAS, 2022]

Sources

  1. Circana · new MMM findings for Best Buy Ads, 7 October 2026 — Joint Circana and Best Buy Ads release with ROAS, brand-equity and competitor-sales estimates. It does not disclose absolute ROAS, the number and identity of brands, model specification, confidence intervals or profit.
  2. Best Buy · initial research findings at The Drop, 28 September 2026 — Primary source for the earlier framing: ROAS 17% above the national average, 17% of media spend, 35% of measured contribution to Best Buy sales, 9.9% of brand equity and almost 16% of sales at competitors.
  3. MediaPost · Best Buy Ads pushes further into commerce media, 24 September 2026 — Industry source describing the input scale: two years, four major categories and $2.3 billion in media spend; it also says Circana ran the model for Best Buy.
  4. Google Research · Challenges and Opportunities in Media Mix Modeling, 2017 — Methods paper on the challenges of obtaining reliable inference from marketing mix models; it does not audit the Circana model.
  5. Google Research · Bayesian Methods for Media Mix Modeling, 2017 — Research on carryover, saturation and uncertainty in MMM, showing estimates can depend on data volume, specification and prior assumptions.
  6. Annals of Applied Statistics · Robust Causal Inference for Incremental ROAS, 2022 — Peer-reviewed paper on causal iROAS estimation with randomized paired geo experiments; a methodological benchmark, not a test of Best Buy Ads.