What happens to organic and direct traffic when paid is paused for a test window? Quantifies incremental lift.

MetricIncremental lift attributable to a paid channel, measured as the change in organic and direct volume when paid is paused in a test window

Unit% incrementality, and incremental cost per acquisition against reported CPA

Two lines (organic plus direct) before, during, and after a paid spend pause. Quantifies actual incremental lift of paid. Build by holding paid spend at $0 for 1-2 weeks and tracking unpaid channels. Small dips during pause confirm real incrementality. Example: organic dips 14% during pause week, direct 12% (real incrementality).

Randomized and geo-split pause experiments, single-advertiser and multi-advertiser: eBay field experiments across Yahoo and MSN with Google as control (Econometrica 2015), 17 Meta conversion lift studies (Brainlabs), Google conversion lift on client accounts (Pearmill 2026), five geo-split CausalImpact experiments (Blackbird PPC) · The spread here is genuine and very wide, and averaging it would be the mistake. eBay found near-total substitution on brand terms, while Google's own search ads pause meta-analysis reports roughly 89% of lost paid clicks are not recovered by organic. Both can be true: incrementality falls as organic strength on the query rises, so a dominant brand with the top organic result sees low brand-search incrementality and an unknown brand sees high. Read your own result, not the average. The eBay paper is 2015 and predates the current SERP layout and AI Overviews, so its magnitude should not be transplanted, though its mechanism holds. Sample sizes for the practitioner studies are small, in the range of 5 to 17 experiments.

The organic line rises by almost exactly the amount the paid line falls during the pause window, so the two curves mirror each other and the combined total never moves, meaning the channel was buying traffic that was already coming.

Paid brand clicks fall by roughly 4,100 during a three-week pause while organic clicks rise by roughly 3,900 and total conversions move less than 2%, against a reported CPA of $38 that implies an incremental CPA above $600.

• Run the pause as a geo-split rather than a global switch-off, holding matched control regions live and reading the difference with a structural time-series model; this is what produced a 25% combined paid plus organic conversion drop at 99% confidence in the Blackbird experiments, a result a global pause could not have separated from seasonality.
• Run brand and non-brand separately and never pause both at once; the measured gap is 19% incrementality on brand against 59% on non-brand plus PMax, so a combined pause produces a blended number that justifies the wrong decision on both.
• Reprice the channel on incremental CPA rather than reported CPA after the test; in the Blackbird case a $40 reported CPA became a $100 incremental CPA once substitution was removed, which is the number that belongs in the budget model.
• Hold the pause for one to three weeks and track paid plus organic as one combined entity per engine for the duration; shorter windows do not clear the conversion lag and separate tracking hides the substitution that is the entire point of the test.

1. Blake, Nosko and Tadelis, EconometricaConsumer Heterogeneity and Paid Search Effectiveness: A Large Scale Field Experiment · 2015 · large-scale field experiments at eBay across Yahoo and MSN with Google as control Brand keyword ads showed no measurable short-term benefit; almost all forgone paid click traffic and attributed sales were recaptured by natural search, so substitution was nearly complete faculty.haas.berkeley.edu ↗
2. Econometrica via RePEcConsumer Heterogeneity and Paid Search Effectiveness (journal record) · 2015 · Econometrica vol. 83, pages 155 to 174 Journal-of-record listing confirming the eBay result, including negative average returns on non-brand keywords for frequent users ideas.repec.org ↗
3. PearmillA Practitioner's Guide to Paid Search Incrementality, Part 2 · 2026 · Google conversion lift studies run on client accounts Brand search measured at 19% incrementality, meaning roughly 4 in 5 attributed conversions would have occurred anyway; non-brand search plus PMax measured at 59% incrementality with 182% conversion lift pearmill.com ↗
4. Brainlabs with MetaProving Paid Social's Halo: Meta Conversion Lift · 2026 · 17 Meta conversion lift studies across client accounts Paid social drove +19% incremental search visits at $6.78 cost per incremental visit, of which 71% were organic rather than paid search brainlabsdigital.com ↗
5. Blackbird PPCHalo Effect in Digital Marketing: Measure Incremental Impact Without Cookies · 2026 · five geo-split incrementality experiments using Bayesian structural time-series (CausalImpact) Pausing Google brand search cut combined paid plus organic conversions by 25% at 99% confidence; organic recaptured 300 of 500 conversions, moving reported CPA of $40 to an incremental CPA of $100 blackbirdppc.com ↗
6. Geeks For GrowthIncrementality or Illusion: Getting Past Last-Click in 90 Days · 2025 · synthesis of the eBay field experiment and Google's search ads pause meta-analysis (Chan et al.) Google's search ads pause meta-analysis reports roughly 89% of lost paid clicks are not recovered by organic on average, against eBay's near-total substitution, so the spread across advertisers is very wide geeksforgrowth.com ↗
7. Search Engine LandWhy incrementality is the only metric that proves marketing's real impact · 2025 · methodology overview of RCT, geo-split and platform lift tests Geo-based randomization is the workable substitute when individual randomization is not possible; the eBay pause showed sales largely unchanged, indicating demand capture rather than creation searchengineland.com ↗
8. Search Engine JournalHow Paid Search Incrementality Impacts SEO · 2021 · single-advertiser SEM shutoff and restart test With SEM and SEO both in market, incremental traffic ran 30 to 40% above SEO alone, mostly on brand keywords; total site visits fell 19% during the shutoff searchenginejournal.com ↗
9. Search Engine LandSearch incrementality: How paid and organic work together for better performance · 2022 · SMX Next practitioner methodology session Recommends suspending SEM campaigns for one to three weeks and tracking paid and organic as a single entity per engine during the test window searchengineland.com ↗