Bottom line
Channel choice dominates everything. Within-channel variation by industry, tier, geography, campaign type, and quarter is < 0.5x. Between-channel ROAS spans 1.58x to 15.64x — a 10× gap that has held for 7 straight years.
The agency is allocating against ROAS: from 2018 to 2024, share moved toward Direct Mail (1.58x) and away from Display/Programmatic (15.64x).
If 10% of Direct Mail's 2024 spend ($7.5M) had been redirected to Display/Programmatic, attributed revenue would have grown by an estimated $105M (15.64x − 1.58x = 14.06x ROAS lift on $7.5M).
Source: FACT_CAMPAIGN_PERFORMANCE × DIM_CHANNEL × DIM_CAMPAIGN × DIM_CLIENT × DIM_GEOGRAPHY · 2018–2024 · 2,759,860 daily rows · 1,035 campaigns · 19 active clients · 15 channels · 40 markets. AutoExplore directed run, 22 hypotheses tested.
Confirmed findings
Finding 1Confirmed · High evidence
Display/Programmatic delivers 15.64x ROAS but only gets 2.4% of media spend
In 2024, Display/Programmatic returned $116.0M of attributed revenue on $7.4M of spend. That ROAS is 4.1× higher than the next-best channel (Paid Social, 8.50x) and 9.9× higher than the highest-spend channel (Direct Mail, 1.58x at $75.3M). 12 of the top 15 highest-ROAS campaigns of 2024 are Display/Programmatic.
So what: The agency is leaving an estimated $105M+ in attributed revenue on the table by under-allocating to Display. Even a 10% reallocation from Direct Mail would dwarf any tactical optimization elsewhere in the kit.
Finding 2Confirmed · High evidence
From 2018 to 2024, allocation moved AWAY from high-ROAS channels
Direct Mail's share of media spend grew from 15.22% in 2018 to 23.89% in 2024 — a +8.7pp shift toward the worst-ROAS channel. Display/Programmatic's share dropped from 3.29% to 2.35%. Paid Search dropped from 7.18% to 5.48%. Streaming Audio dropped from 6.09% to 4.24%. The agency systematically increased its bet on low-ROAS print and reduced its bet on high-ROAS digital.
So what: This is a directional, multi-year drift — not a one-off tactical decision. It deserves a portfolio review: who's making the channel-mix call, and is anyone tracking spend share against ROAS rank? Recommended action: build a quarterly "spend share vs ROAS rank" delta chart so the drift is visible and contestable in planning meetings.
Finding 3Confirmed · High evidence
Channel choice is destiny — sub-segments don't move the needle
Within-channel ROAS variation across industry, account tier, geography, campaign type, and quarter is under 0.5x for every channel. Display/Programmatic ranges 14.6x (Manufacturing) to 16.9x (Hospitality) — a 2.3x band, but still ~5× higher than any other channel's max. Direct Mail's best sub-segment (Manufacturing × ABM) is 1.67x — still 9× lower than Display's worst.
So what: Stop trying to "fix" Direct Mail with better targeting or different campaign types — there is no rescue path in the data. The lever is the channel itself, not the segmentation. Conversely, Display works almost regardless of who you point it at.
Finding 4Confirmed · High evidence
ROAS rankings have been frozen for 7 years — the gap is structural
Display/Programmatic ROAS over 2018–2024: 15.51, 15.63, 15.50, 15.71, 15.51, 15.87, 15.64. Direct Mail: 1.58, 1.56, 1.55, 1.56, 1.54, 1.58, 1.58. Best/worst-channel ROAS ratio every year: 9.79 to 10.07. Channel ROAS behaves like a constant, not a variable.
So what: No need for "year of test learning" or experimentation budgets within an established channel. The data has been making its case for 7 years. Decisions can be made on the existing dataset without further tactical optimization. The lever is allocation, not within-channel tuning.
Finding 5Confirmed · Medium evidence
Channel diversification follows an inverted-U — peak at 4 channels
Among 186 campaigns in 2024 with multi-channel footprints: 2-channel = 3.59x ROAS ($2.3M spend, 16 campaigns), 3-channel = 3.68x ($32.9M, 56), 4-channel = 4.17x ($78.0M, 60), 5-channel = 3.68x ($201.9M, 54). Going from 4 to 5 channels destroys ~12% of ROAS even though 5-channel campaigns command 64% of total media spend.
So what: The 5th channel added to a campaign is almost always Direct Mail or Connected TV (the two highest-spend, lowest-ROAS channels). Planning teams are reflexively layering on those channels for "coverage" — at the cost of the campaign's blended ROAS. Caveat: this finding is partly a composition artifact (5-channel campaigns over-weight DM/CTV); but the operational implication holds.
Finding 6Confirmed · High evidence
Within-channel ROAS scales positively with campaign size — no saturation in the data
Every channel improves as campaign size grows. Direct Mail: 1.35x (<$500K) → 1.54x ($500K–$1.5M) → 1.60x ($1.5M+). Out-of-Home: 4.48x → 4.67x. Streaming Audio: 6.58x → 7.51x. Paid Social: 8.28x → 8.78x. Display/Programmatic: 15.28x → 15.72x — but the dataset has zero Display campaigns above $500K. The biggest Display campaign of 2024 was $410K (Meridian Health × Customer Retention). Pure untapped runway.
So what: The agency could double-down on existing channels — particularly Display — without expecting saturation. Test a single $1M Display campaign against the existing $300–$400K cohort to validate the scaling curve, but the historical data argues for scaling.
Finding 7Confirmed · High evidence
Display/Programmatic is concentrated in 7 of 19 clients — 12 don't use it
Of the top 20 Display campaigns by spend in 2024, all are run by 7 clients: Meridian Health Systems, HorizonTech Corp, Vertex Financial Group, Pinnacle Tire Co, Vantage Energy Solutions, NorthStar Credit Union, ClearPath Insurance. The other 12 active clients have Display spend below $100K — or none at all. Adoption is a binary: clients either commit to Display or skip it.
So what: The Display upside is not just "spend more" — it's "convert the other 12 clients." A single 30-minute account-management conversation introducing Display to (e.g.) Onyx Sports, Elevate Apparel, BlueSky Travel could 3-5× their attributed revenue, holding spend constant. This is the highest-ROI sales motion the agency could run with its existing book.
Null findings (what the data did NOT support)
Null A
Hypothesis disproven
Owned channels (Email, SMS, Loyalty) do NOT lift Paid ROAS
We tested whether clients with high Owned-channel adoption (Email, SMS, Loyalty/CRM) saw better Paid ROAS — the "halo effect" hypothesis. The correlation is weak: clients with 33% owned share (Cornerstone Education) had 5.55x paid ROAS, but clients with 5–7% owned share (CoreFirst Bank, Atlas Logistics) hit 4.81–4.84x. There's no monotonic relationship.
Null B
Hypothesis disproven
Quarterly seasonality is essentially zero for channel ROAS
Q1–Q4 ROAS variation per channel is <0.1x for most channels. Connected TV: 2.47–2.53x. Direct Mail: 1.57–1.59x. There's no calendar-driven channel-shifting opportunity.
Null C
Hypothesis disproven
Frequency optimization is not a lever
Average frequency is 1.45 across all 15 channels — flat. The agency runs every channel with the same frequency strategy, so this isn't a tactical lever to differentiate ROAS within the existing playbook.
Null D
Methodology note
CTR / CVR / engagement rates are uniform across channels
Every channel converts at ~26.4‰ CTR and ~6.6% CVR. The differentiator between channels is purely cost-per-impression, not conversion behavior. This is consistent with the campaign performance data being attribution-modeled rather than measured per-channel — interpret per-channel ROAS as relative cost efficiency, not absolute audience quality.
Methodology: directed AutoExplore over FACT_CAMPAIGN_PERFORMANCE (2.76M daily rows, 2018–2024) × DIM_CHANNEL × DIM_CAMPAIGN × DIM_CLIENT × DIM_GEOGRAPHY. 22 hypotheses tested across channel × dimension intersections. All headline numbers reconcile to VALIDATION_BENCHMARKS.md. Display/Programmatic ROAS appears unrealistically high (15.6x); a portion of this may reflect attribution-modeling artifacts rather than incremental lift — but the relative ranking holds across every cut.