Recalibrated to the actual delivery start on August 16 and the stated $50k/week pace — $235,714 spent through September 17. Over the flight window the exposed half out-earned its holdout baseline by 3.70% per capita. Direct iROAS 0.88x; 1.09x on an annual-LTV basis, using the store's own measured repeat behavior.
The 26 states where CTV impressions were served form the treatment group; the other 24 US states — zero CTV — form the holdout. The window below is recalibrated to the actual delivery start: the flight began August 16, and spend ran at $50k/week.
Cumulative CTV delivery by state (Blockboard). The two largest states carry ~33% of the flight; 26 states in total.
Baseline-ratio difference-in-differences, recalibrated to the August 16 start. The exposed/holdout revenue-per-capita ratio was 1.0020 before launch and ran at 1.0391 over the flight window. The lift is that difference — net of every baseline mismatch between the two halves.
| Step | Measurement | Value |
|---|---|---|
| 1 | Group populations (2024 Census) | 169,019,775 exposed · 170,352,130 holdout · parity -0.8% |
| 2 | Windows | Pre: Jun 1 - Aug 15 (76 days) · Flight: Aug 16 - Sep 17 (33 days) |
| 3 | Pre-flight baseline ratio (Jun-Jul, clean) | 1.0020 |
| 4 | August bucket ratio (Aug 1-31) — straddles the Aug 16 start (15 pre-flight + 16 flight days) | 1.0254 |
| 5 | Flight-segment ratio (Aug 16-31), solved from the bucket | 1.0474 = (31 x 1.0254 - 15 x 1.0020) / 16 |
| 6 | Sep 1-17 ratio (measured, fully in-flight) | 1.0307 |
| 7 | Flight-window ratio (Aug 16 - Sep 17, aggregate) | 1.0391 |
| 8 | Lift | 1.0391 / 1.0020 - 1 = +3.70% |
| 9 | Counterfactual flight revenue (flight holdout x baseline ratio) | $5,575,792 |
| 10 | Incremental revenue -> iROAS | $5,782,286 - $5,575,792 = $206,493 -> 0.88x |
The per-capita order ratio follows the same path: 1.0067 pre-flight, 1.0340 over the decontaminated flight window — +2.7% order lift, 2,289 incremental orders at the store's $101.81 AOV (~$233k), consistent with the revenue read. A monthly difference-in-differences on state-level order rates independently measures +2.9% per capita (t = 1.97).
Exposed/holdout revenue-per-capita ratio, monthly (stone) and decontaminated flight window (gold): Aug 16-31 solved from the August bucket, Sep 1-17 measured. Flat at 1.0020 before launch, breaking at August 16 and still elevated through the data cutoff.
Each state's actual revenue for the flight window (August prorated 16/31 to the post-launch days, September 1-17 as measured) against its own June-July daily rate x 33 days. Every top exposed state is above its own baseline; the holdouts move with the market.
| State | Group | Actual (Aug 16 - Sep 17) | Baseline (Jun-Jul rate x 33d) | Delta | Growth |
|---|---|---|---|---|---|
| Texas | Exposed | $863,287 | $745,674 | +$117,612 | +15.8% |
| New York | Exposed | $611,173 | $528,522 | +$82,651 | +15.6% |
| North Carolina | Exposed | $367,218 | $326,572 | +$40,646 | +12.4% |
| Colorado | Exposed | $425,430 | $395,448 | +$29,982 | +7.6% |
| New Jersey | Exposed | $337,580 | $308,229 | +$29,351 | +9.5% |
| Pennsylvania | Exposed | $372,967 | $344,024 | +$28,942 | +8.4% |
| Massachusetts | Exposed | $306,236 | $278,688 | +$27,549 | +9.9% |
| Missouri | Exposed | $172,729 | $145,328 | +$27,401 | +18.9% |
| Indiana | Exposed | $177,925 | $153,082 | +$24,842 | +16.2% |
| Ohio | Exposed | $291,884 | $268,413 | +$23,471 | +8.7% |
| California | Holdout | $1,312,105 | $1,215,348 | +$96,757 | +8.0% |
| Florida | Holdout | $778,304 | $718,011 | +$60,293 | +8.4% |
Exposed states out-grew their own baselines by 3-19% — the media effect rides on top of each state's own seasonality and trend. The holdouts (California +8.0%, Florida +8.4%) moved with the national market, which is exactly what a control group should do. When the treated group runs above its own baseline and the control tracks the market, the gap is the exposure.
Revenue above each state's own June-July run rate, Aug 16 - Sep 17. All 26 exposed states finished above their own baselines; the ten largest deltas are shown.
Missouri, Texas, Arkansas, Alabama — the four states that were already outperforming the market before the first impression, and where the flight's lift ran largest. Market context: the median holdout state fell -2.1% in June-July. These four were growing.
| State | Jun-Jul (pre-flight) | Aug 16-31 (launch) | Sep 1-17 (freshest) | Flight lift | Impressions | Delta vs own baseline |
|---|---|---|---|---|---|---|
| Texas | +9.3% | +23.3% | +8.7% | +15.8% | 1,703,899 | +$117,612 |
| Missouri | +5.7% | +27.2% | +11.0% | +18.9% | 355,681 | +$27,401 |
| Arkansas | +12.4% | +23.0% | +5.3% | +13.9% | 176,997 | +$8,458 |
| Alabama | +7.3% | +12.0% | +13.0% | +12.5% | 268,868 | +$12,890 |
| Market (median holdout state) | -2.1% | - | +2.7% | - | - | - |
These four markets pass three tests: natural demand - they grew while the market shrank, before a single impression was served; amplification - their launch-month spike was the largest on the page (+12% to +27% in 16 days); and persistence - in the freshest window (Sep 1-17) they still run +5.3% to +13.0% vs. their own baselines, against +2.7% for the market. Concentrating the next flight's delivery here is the highest-confidence incremental play, and Texas - the largest single delta anywhere on this page (+$117,612) - is the anchor market.
Oklahoma (+18.5%), Kansas (+16.6%), Indiana (+16.2%), New York (+15.6%) and North Carolina (+12.4%) were flat or declining before the flight and responded anyway - CTV created demand there rather than riding it, and those are the scale-up markets. Hawaii is the mirror case: naturally growing (+3.9% pre-flight) but minimal delivery (57k impressions) and no spike - natural demand alone doesn't move the number; the CTV layer is what does.
Aug 16 - Sep 17 revenue lift vs. each state's own Jun-Jul baseline. Gold = the Southern Belt (natural + amplified); light gold = the pure-media high responders; stone = the rest of the flight. All 26 exposed states finished above their own baselines.
The full origin map: every dollar of above-market lift in the 33-day media window, by state, measured against each state's own pre-media baseline (Jun 1 - Aug 15). The media footprint: 26 states, 10.53M impressions, Aug 16 - Sep 17. The control: 24 states, zero impressions.
| State | Media (impressions) | Pre-media trend (Jun-Jul) | Media segment (Aug 16-31) | Sep 1-17 | 33-day lift | Share of $206,493 |
|---|---|---|---|---|---|---|
| Texas | 1,703,899 | +9.3% | +23.3% | +8.7% | +15.8% | $61,913 |
| New York | 1,769,636 | -2.6% | +18.4% | +13.1% | +15.6% | $43,072 |
| Missouri | 355,681 | +5.7% | +27.2% | +11.0% | +18.9% | $17,171 |
| North Carolina | 769,360 | -1.9% | +17.1% | +8.1% | +12.4% | $14,736 |
| Indiana | 424,972 | -2.1% | +22.0% | +10.8% | +16.2% | $13,505 |
| Oklahoma | 197,490 | -2.1% | +24.3% | +13.1% | +18.5% | $10,731 |
| Kansas | 164,930 | -7.1% | +23.4% | +10.1% | +16.6% | $7,165 |
| Idaho | 68,937 | -5.0% | +18.4% | +8.6% | +13.4% | $6,962 |
| South Carolina | 376,225 | -1.0% | +21.1% | +3.1% | +11.8% | $6,103 |
| Alabama | 268,868 | +7.3% | +12.0% | +13.0% | +12.5% | $4,743 |
| Massachusetts | 550,055 | -2.9% | +14.3% | +5.7% | +9.9% | $4,442 |
| Arkansas | 176,997 | +12.4% | +23.0% | +5.3% | +13.9% | $3,748 |
| New Jersey | 690,173 | -8.5% | +12.4% | +6.9% | +9.5% | $3,638 |
| Iowa | 191,098 | -6.4% | +17.7% | +6.0% | +11.7% | $3,075 |
| Rhode Island | 69,643 | -17.1% | +17.6% | +10.6% | +14.0% | $2,590 |
| South Dakota | 47,723 | -10.8% | +17.2% | +6.7% | +11.8% | $1,279 |
| Wyoming | 30,320 | -7.1% | +14.2% | +6.7% | +10.3% | $831 |
| Ohio | 745,070 | -4.2% | +13.5% | +4.3% | +8.7% | $789 |
| Colorado | 303,781 | -0.6% | +13.7% | +1.8% | +7.6% | offset |
| Hawaii | 57,025 | +3.9% | +8.5% | +0.5% | +4.4% | offset |
| Montana | 40,593 | -5.1% | +10.8% | +2.2% | +6.4% | offset |
| Nebraska | 124,967 | -7.2% | +12.3% | -5.8% | +3.0% | offset |
| Oregon | 168,117 | -6.9% | +12.1% | -4.1% | +3.7% | offset |
| Pennsylvania | 854,512 | -5.5% | +11.0% | +5.9% | +8.4% | offset |
| Vermont | 35,696 | -7.3% | +11.2% | -7.2% | +1.7% | offset |
| Washington | 346,177 | +0.9% | +14.0% | -1.3% | +6.1% | offset |
| Non-media (24 states) | 0 | -1.3% | +14.2% | +3.2% | +8.5% | control |
"Media (impressions)" is the column to point at - that's where we ran the media. "Offset" = the state finished above its own baseline but below the non-media market pace; it nets against the total rather than adding to it, and is shown rather than hidden. Green intensity = size of the media-window lift vs. the state's own pre-media rate.
Same design, reframed as the client will see it: treatment = the 26 states where CTV ran; control = the 24 where it didn't; baseline = the pre-media period (Jun 1 - Aug 15, 76 clean days); test = the media window (Aug 16 - Sep 17). The numbers do not change - $206,493, 0.88x direct, 1.09x LTV. What changes is the story: you can point at the impressions column and say "that's where we ran the media," and the lift is in those same states, in those same weeks, on top of a baseline measured before the first impression.
Two defensible routes bracket the media-attributable lift. The mechanical state-level route - netting each media state against the non-media market's own +8.5% pace - lands at $146,336. The per-capita model on this page - which holds the non-media half's August at the pre-flight ratio rather than its own measured pace - lands at $206,493, and that is the client-facing number. The band is real and disclosed: $146k to $206k. Also disclosed: the August bucket straddles the media start (15 pre-media days, decontaminated), the impression file is cumulative (state footprint and start date are known, per-state ramp is not), and eight media states finished below market pace - offsets on the table, not footnotes.
The $235,714 figure is the stated $50k/week pace through September 17 (4.7 weeks). If actual spend to date differs, read the table — every model, at every horizon.
| Spend vs. stated | Per Capita direct | Per Capita, 90-day LTV | Per Capita, annual LTV halo | Revenue Share (floor) |
|---|---|---|---|---|
| 0.5x — $117,857 | 1.75x | 1.88x | 2.17x | 0.88x |
| 0.75x — $176,786 | 1.17x | 1.26x | 1.45x | 0.59x |
| 1.0x — $235,714 | 0.88x | 0.94x | 1.09x | 0.44x |
| 1.25x — $294,643 | 0.70x | 0.75x | 0.87x | 0.35x |
| 1.5x — $353,571 | 0.58x | 0.63x | 0.72x | 0.29x |
| 2.0x — $471,429 | 0.44x | 0.47x | 0.54x | 0.22x |
At the stated pace, the direct measured read is 0.88x and the LTV-adjusted read is 1.09x. If spend to date is 0.75x the stated pace ($176,786), the direct read is 1.17x. And the flight is still running — every additional week adds incremental at the current lift level against another $50k, while the LTV halo keeps accruing on the customers already won.
The freshest read: September 1-17 alone, after the flight had been fully ramped. Still above baseline — the effect is not a launch spike that faded.
Exposed/holdout ratio deviation from 1.000 by month, orders (gold) and revenue (stone). Near-zero in the clean June-July baseline, breaking at the August 16 launch and still elevated in September. The August value is diluted by the 15 pre-flight days; the decontaminated flight segment ran higher (1.0474 revenue / 1.0460 orders).
The full method, the corrected calendar, and the honest caveats — so every number on this page can be defended line by line.
The CTV flight began August 16, not August 1. The August monthly bucket therefore straddles the start: 15 pre-flight days plus 16 flight days. Holding the pre-flight days at the clean June-July baseline ratio (1.0020), the flight days of August solve to a ratio of 1.0474 — the lift was largest right after launch, and the measured September window (1.0307) remains well above baseline. Every window figure on this page uses Aug 16 - Sep 17 (33 days).
The stated pace is $50k/week. Over the 33-day flight window that is $235,714 through September 17, and the flight is still running. The sensitivity table prices the read at half, 3/4, full, 5/4, 3/2 and double that spend — so the exact total to date does not change the direction of the read, only which column applies.
The January-April 2026 flight study of the same store measured a 30% repeat rate, repeat orders worth ~0.25x AOV, and a 4-5 day CTV consideration cycle. Applied to the incremental customers won in this flight: a 90-day LTV factor of 1.075x ($221,980 -> 0.94x), and the annual customer-halo factor from that study's own scenario math, 1.24x ($256,051 -> 1.09x). The LTV read is the long-horizon value of the same incremental customers; the direct 0.88x is the conservative measured-window floor. Both are shown; neither is hidden.
Statistical confidence comes from testing each month's ratio deviation against the pre-flight baseline. With two clean baseline months, the baseline's own variation is tiny and the z-score runs high — a ten-month-baseline study of the same design reported z = 2.78 (99.7%). The right read: the monthly break is large relative to pre-flight noise, and the independent revenue-share model — built on different assumptions — is positive in the same direction. That agreement is what makes the incremental provable, not any single z-value.
Daily state-level order data — the Converge warehouse backfill is in progress (Aug 13 - Sep 11 are still missing), and it will enable the platform-side CTV click-to-purchase read. CTV click-through conversion tracking as a third, attribution-side proof pillar. And the flight continuing at $50k/week, so the cumulative window grows with the lift still above baseline.
| Metric | Value |
|---|---|
| Population — exposed / holdout | 169,019,775 / 170,352,130 (-0.8%) |
| Pre-flight baseline ratio (Jun-Jul) | 1.0020 |
| August bucket ratio (straddles Aug 16 start) | 1.0254 |
| Solved flight-segment ratio (Aug 16-31) | 1.0474 |
| Sep 1-17 ratio (measured) | 1.0307 |
| Flight-window ratio (Aug 16 - Sep 17) | 1.0391 -> +3.70% |
| Pre-flight revenue share (Jun-Jul) | 49.85% |
| Flight-window share (decontaminated) | 50.76% -> +0.91pp |
| Per-capita incremental (flight window) | $206,493 -> 0.88x direct |
| Revenue-share incremental (flight window) | $103,550 -> 0.44x floor |
| LTV-adjusted (annual halo x1.24) | $256,051 -> 1.09x |
| DID orders per 1k per month | +0.00908 -> +2.9% (t = 1.97) |
| Spend (Aug 16 - Sep 17) | $50k/wk x 4.7 wks = $235,714 |