Marginal

Marketing Mix Modeling

last-click told you what happened. it never told you why.

marginal is the independent measurement platform that answers the questions last-click never could — what did your spend actually cause, and which parts of your media mix drove that outcome. including brand and upper funnel investment that last-click makes invisible.

start your pilot — first 3 runs free find out where to start →
[screenshot coming — MAR-479]
built on google meridian the same bayesian foundation used by enterprise teams
incrementality + mmm in a calibration loop — architecturally unique
no data science team required guided configuration for marketing teams
structurally independent not owned by a platform, mmp, or agency

your measurement stack has a conflict of interest.

every tool in the current stack profits from the number it reports. platforms inflate roas. mmps maximise attributed volume. last-click — whether in analytics on web or your mmp on app — records who was last in the room when a conversion happened. not who influenced the decision.

the signal it relies on is getting worse. cookie deprecation on web. ios privacy and inconsistent postbacks on app. and the question that actually justifies your budget — what would have happened if we hadn't spent? — remains unanswered.

see what every channel actually contributed — including brand.
bayesian mmm decomposes contribution across your full media mix simultaneously. brand spend, upper funnel, awareness — every channel that last-click makes invisible finally has a number attached to it. not a reach metric. contribution to outcomes.
run the tests that prove it.
geo-holdout and conversion lift tests establish causal ground truth — what your spend actually caused, not what the model inferred. built-in test design wizard with power analysis. results feed back into the model through the calibration loop.
allocate smarter. own the narrative.
scenario planner with hard budget constraints turns measurement into decisions. the equimarginal output tells you where the next pound generates the highest return. credible intervals, sensitivity analysis, and plain-language calibration narratives give you the evidence to walk into any board meeting and stand behind the numbers.

Trusted by teams at

Acme Brands
[LOGO: Company B]
[LOGO: Company C]
[LOGO: Company D]
[LOGO: Company E]
[LOGO: Company F]

for the marketing director / cmo

brand spend finally has a number you can defend.
replace last-click reporting with independent evidence. walk into any board or cfo conversation with credible intervals, causal test results, and a methodology you own — not numbers coming from someone who profits from making them look good.

for the head of growth / performance director

find the saturation curve you can feel but can't see.
stop optimising on last-click attribution you don't trust. see what every channel is actually responsible for. quantify diminishing returns. scale systematically toward what's working — not spray budget and hope the right channels benefit.

for the marketing analyst / data lead

the statistical rigour — automated.
bayesian mmm and incrementality testing done. you focus on interpretation and recommendation, not computation. transparent methodology, exportable outputs, no walled garden. move from data wrangler to the person who owns measurement.

how it works

  1. 1

    find out where to start

    answer a few questions about your channels, spend history, and conversion data. we'll tell you exactly what to pull together.

  2. 2

    bring your data

    upload spend and conversion data from analytics on web, mmps on app, ad platforms. 12+ months of weekly data across 3+ channels.

  3. 3

    configure with guidance

    guided prior setup grounded in industry benchmarks. no statistician required.

  4. 4

    see what your spend actually caused

    first model run decomposes contribution by channel, including brand and upper funnel. interaction effects visible. uncertainty shown, not hidden.

  5. 5

    test and compound

    marginal suggests which incrementality tests to run next. each cycle sharpens the model. you're never left to figure out the next step alone.

  6. 6

    make an impact

    scenario planner with hard budget constraints. the equimarginal output tells you where the next pound generates the highest return.

  7. 7

    own the narrative

    credible intervals, sensitivity analysis, plain-language calibration narratives. independent evidence you can stand behind in any room.

stop optimising on the wrong signal. start scaling what's actually working.

your first three model runs are free. no commitment required to see what last-click has been missing.

start your pilot