A six-step playbook for measuring marketing ROI defensibly in 2026 - using modern Bayesian Market Mix Modelling, calibrated with experiments and refreshed monthly so the numbers drive live planning, not retrospective reporting.
Pick the number the board actually cares about - revenue, gross profit, contribution or units. Avoid clicks, sessions and platform-reported conversions as the headline KPI; they conflate cost, demand and platform self-attribution.
Pull weekly spend and exposure (impressions, GRPs) by channel and market, plus price, promotions, distribution, brand metrics and macro variables (weather, CPI, competitor activity). Two to three years of weekly history is the practical minimum.
Decompose the KPI into a base plus the incremental contribution of every driver. Model adstock (carry-over) and saturation (diminishing returns) per channel. Use Bayesian priors informed by experiments and prior knowledge to constrain unidentifiable parameters.
Run geo-holdouts, platform lift tests and switchback experiments. Feed the measured incremental lift back into the MMM as priors. Uncalibrated MMM drifts - calibrated MMM stays honest.
Marketing ROI = incremental contribution divided by cost, per channel and per pound. Use the saturation curves to find the marginal ROI of the next pound spent in each channel, and reallocate until marginal ROIs equalise. Typical uplift: 10-20% more value at the same total spend.
Move the model from annual deck to monthly decision system. Build scenario simulators on top so commercial teams can pressure-test plans before signing them off. The deliverable is decisions, not a model.
ROAS is revenue divided by ad spend. It is the platform metric, and it overstates value because it includes sales that would have happened anyway and ignores margin.
ROI is incremental profit divided by cost. It is the CFO metric, and the number that should drive budget decisions. MMM produces ROI; platforms produce ROAS. Always know which one you are reporting.
Define the commercial KPI (revenue, gross profit, contribution), collect weekly spend and exposure by channel together with price, distribution and macro data, then fit a Bayesian Market Mix Model that decomposes the KPI into a base plus the incremental contribution of each driver. Calibrate with geo-experiments and platform lift studies, divide each channel's incremental contribution by its cost for ROI, derive saturation curves to find the optimal mix, and refresh monthly so the numbers inform live planning rather than retrospective reporting.
There is no universal benchmark - it depends on category, margin and channel role. A high-margin DTC brand might run paid search at 5-8x ROAS; a low-margin grocery brand might run TV at 1.5-2x and still be value-accretive once long-term brand effects are included. The right comparison is always marginal ROI - the return on the next pound spent - not blended ROAS.
ROAS (return on ad spend) is revenue divided by ad cost - a useful platform metric but it ignores margin, cannibalisation and the base level of sales that would have happened anyway. ROI (return on investment) is incremental profit divided by cost - the number the CFO cares about. MMM produces ROI; platforms produce ROAS.
For pure-digital, fully-trackable channels you can approximate ROI with incrementality tests and multi-touch attribution. For TV, OOH, brand, sponsorship and any channel with offline impact, MMM is the credible answer. Most businesses end up with both.
A first defensible MMM typically takes 8-12 weeks from data sign-off to first results. Modern platforms compress that to 4-6 weeks once data pipes are in place, and then refresh in days, not weeks, every month.
Most businesses do not, day one. The pragmatic path is a 12-week data-foundation sprint alongside model build: reconcile spend across platforms, harmonise channel taxonomy, fill gaps with proxies, and lock a weekly cadence going forward. The Clarity Score diagnostic identifies the biggest gaps in 10 minutes.
Monthly for in-year planning and reallocation, with a deeper annual review for budget setting. Quarterly is acceptable for slower-moving categories. Annual-only MMM is no longer fit for purpose - the market and platforms change too fast.
Start with a free Clarity Score - a 10-minute diagnostic of where your marketing-ROI capability sits today and where the biggest gains are.
Start your Clarity Score