Digital Marketing Academy · Lesson

MMM vs Attribution

Top-down vs bottom-up.

MMM vs Attribution is a free Digital Marketing Academy lesson on CoddyKit — lesson 1 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the Digital Marketing Academy learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.

Two Lenses on the Same Question

Marketing Mix Modeling (MMM) and multi-touch attribution (MTA) both try to answer "what drove sales?" but from opposite directions.

MMM is top-down: it uses aggregate, time-series spend and sales data to estimate each channel's contribution. MTA is bottom-up: it stitches together individual user touchpoints toward a single conversion.

Advanced practitioners use both, knowing each has blind spots the other partially covers.

What MMM Actually Measures

MMM regresses sales against marketing and non-marketing drivers over time. It produces channel-level elasticities and a decomposition of sales into base and incremental contributions.

Crucially, MMM captures offline channels (TV, radio, OOH) and long-term effects that user-level tracking cannot see.

sales_t = base
        + b1*TV_t
        + b2*Search_t
        + b3*Social_t
        + b4*OOH_t
        + seasonality_t
        + error_t

What Attribution Actually Measures

MTA assigns fractional credit to each tracked touchpoint a user encountered before converting. Models range from last-click to data-driven Shapley-style credit.

It excels at digital, cookie-trackable journeys and gives granular, near-real-time channel and campaign credit.

journey: Display -> Search -> Email -> Convert

last_click:   Display=0.0  Search=0.0  Email=1.0
linear:       Display=0.33 Search=0.33 Email=0.33
data_driven:  Display=0.18 Search=0.55 Email=0.27

Granularity vs Coverage

The core trade-off: attribution is granular but narrow; MMM is broad but coarse.

MTA can tell you which ad creative converted a user, but ignores TV, word-of-mouth, and anything outside its tracking window. MMM sees the whole budget but cannot drill to a single campaign or audience.

dimension      MMM            MTA
----------     -----------    -----------
data           aggregate      user-level
offline media  yes            no
granularity    channel        campaign/ad
privacy-proof  yes            no

The Privacy Shift

Cookie deprecation, ATT (App Tracking Transparency), and consent rules have eroded the user-level signal MTA depends on.

Because MMM uses only aggregate data, it is largely immune to these changes. This is the main reason MMM has surged back into favor at large advertisers.

Correlation, Causation, Incrementality

Both methods risk confusing correlation with causation. Search often gets over-credited because it captures demand that other channels created.

True incrementality answers: would this sale have happened without the spend? Geo experiments and holdout tests are the gold standard that MMM and MTA should be calibrated against.

Naive:  Search ROAS = 8.0  (looks amazing)
Geo holdout reveals 40% would convert anyway
Incremental ROAS = 8.0 * 0.60 = 4.8

Time Horizons Differ

MTA looks within a conversion window of days to weeks. MMM can model carryover and long-term brand effects spanning months via adstock and trend terms.

If you judge a brand campaign only with MTA, you will systematically undervalue it because its payoff lands outside the window.

Data Requirements

MMM needs 2-3 years of weekly history: spend by channel, sales, price, promotions, distribution, and external factors like weather or macro indices.

MTA needs clean, deduplicated event logs with persistent user identifiers across touchpoints, which are increasingly hard to obtain.

MMM rows (weekly):
week | sales | tv_spend | search_spend | price | promo
2024-W01 | 412k | 90k | 30k | 19.99 | 0
2024-W02 | 488k | 90k | 35k | 17.99 | 1

Triangulation: Use Both

Mature measurement stacks triangulate: MMM for strategic budget allocation across channels, MTA for tactical in-channel optimization, and experiments to calibrate both.

When MMM and MTA disagree, the gap itself is informative, often pointing to untracked or over-credited paths.

Measurement triangle:
  MMM        -> how much per channel (strategy)
  MTA        -> which campaign/ad   (tactics)
  Experiment -> ground truth        (calibration)

Common Failure Modes

MMM fails when channels move together (collinearity), when history is too short, or when the model is over-fit to noise.

MTA fails when tracking is incomplete, when last-click bias dominates reporting, and when offline conversions never enter the dataset.

Choosing the Right Tool

Ask: is the question strategic or tactical? Does it span offline media? Is user-level data reliable here?

Annual budget planning across TV, digital, and retail leans MMM. Optimizing a paid-search account next week leans MTA. Validating either leans experiments.

if question == 'how split total budget?': use MMM
if question == 'which keyword/creative?':  use MTA
if question == 'is this truly incremental?': run experiment

Quick Check

Test your understanding of when MMM beats attribution.

Recap

MMM is top-down, aggregate, privacy-proof, and great for strategic allocation across all media including offline. MTA is bottom-up, granular, and best for tactical digital optimization but vulnerable to tracking loss.

Neither is complete alone. Triangulate MMM, MTA, and experiments, and let incrementality testing settle disputes between them.

Frequently asked questions

Is the “MMM vs Attribution” lesson free?

Yes — the full text of “MMM vs Attribution” is free to read here on the web, and the Digital Marketing Academy course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the Digital Marketing Academy course, upgrade to CoddyKit PRO.

What will I learn in “MMM vs Attribution”?

Top-down vs bottom-up. You practise Digital Marketing Academy with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.

Do I need any experience to start Digital Marketing Academy?

No prior experience is required. Digital Marketing Academy on CoddyKit is structured for beginners through advanced learners; this is — lesson 1 of 4, so you can start here or from the beginning and move at your own pace.

How long does the “MMM vs Attribution” lesson take?

Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.

Can I write and run code in this Digital Marketing Academy lesson?

Yes. Every Digital Marketing Academy lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.

All lessons in this course

← Back to Digital Marketing Academy