MMM-Lite: Marketing Mix Thinking for D2C Brands Without a Data Science Team

What is MMM-lite, and who is it for?

Marketing mix modelling — MMM — answers the budget question statistically: given the history of spend by channel and revenue, what did each channel contribute, where are returns diminishing, and what allocation would have done better? The full discipline needs meaningful data volume, someone comfortable with regression diagnostics, and months of calibration. Most Indian D2C brands between ₹20 lakh and ₹1 crore a month in spend have none of those, and respond by defaulting back to platform attribution — the one measurement system everyone agrees is biased.

MMM-lite is the middle path we run on scaled accounts, including the brand budgets we currently manage between ₹25 lakh and ₹60 lakh a month: keep the mix modeller’s questions and discipline, drop the econometrics, and implement the whole thing in five spreadsheets any competent growth lead can own. It does not replace experiments — it is the connective tissue between them, the thing that makes your incrementality test results govern day-to-day budgets instead of dying in a slide.

Why is mix thinking back in 2026?

As of 2026, the direction of travel is documented on both sides of the industry. On the tooling side, the barrier collapsed: Meta has long maintained Robyn, its open-source MMM library, plus the GeoLift experiment framework, and Google released its Meridian mix model as open source to general availability in 2025, per the companies’ own announcements. On the demand side, trade press has spent the years since iOS App Tracking Transparency describing an MMM revival, as user-level attribution data thinned and privacy rules hardened. Treat the revival as directional industry reporting, not a measurement of adoption — but the strategic read is uncontroversial: the industry’s centre of gravity is moving from tracking individuals to modelling aggregates, and a scaled D2C brand that still budgets purely on platform ROAS is budgeting against the current.

The point of a mix model is not the model. It is forcing every channel to argue for its budget in the same currency.

How does MMM-lite compare to attribution and full MMM?

The honest position of MMM-lite in that table: it is not more accurate than a well-built Meridian or Robyn model. It is enormously more accurate than attribution-only budgeting, available immediately, and legible to a founder — which is what most brands in this band actually need this year.

What are the five sheets?

The whole system is five artefacts, reviewed on one weekly hour.

1. The baseline sheet. Weekly rows, closing discipline, one owner: total spend by channel, blended revenue (site, marketplace, quick commerce — separately, then together), MER, and contribution after ad spend computed on delivered orders, not placed ones. This is the single source of truth; every other sheet feeds on it. If you build nothing else, build this — it is also the pre-scale baseline that makes every future experiment readable.

2. The credit sheet. Each channel’s dashboard-claimed revenue, multiplied by a discount factor your own experiments set: if the retargeting holdout showed half the claimed lift, retargeting carries a 0.5 factor until the next test moves it. Where you have no experiment yet, start harvesting channels low and prospecting channels at par, write the assumption down, and let the quarterly test calendar replace guesses with results one line at a time.

3. The curve sheet. For each major channel, a scatter of weekly spend against blended revenue over trailing quarters. You are not fitting a model; you are looking for the bend — the spend band where each incremental lakh visibly buys less. Crude, honest, and better than nothing by a wide margin: most scaled accounts have never once plotted it.

4. The scenario sheet. Next quarter’s budget as three allocations — hold, shift 15% toward the best marginal read, shift 30% — each priced in expected contribution using the credit and curve sheets. The point is not precision; it is forcing the reallocation argument to happen in numbers before it happens in the account.

5. The decision log. One line per budget move: date, change, reasoning, expected effect, actual effect at review. This is the sheet that compounds — it is the difference between an account that learns and an account that re-argues the same reallocation every quarter with new anecdotes.

How do you read diminishing returns without a model?

Three practical reads substitute for the response curves a full MMM would estimate. The marginal week test: compare your five highest-spend weeks against your median weeks on MER and contribution — if the high-spend weeks bought revenue at meaningfully worse efficiency, you have found the bend without any statistics. The creative-supply check: in most Indian D2C accounts the apparent saturation of a channel is actually creative fatigue wearing a media costume; before concluding a channel is saturated, check whether fresh concepts still buy efficiently while tired ones decay. The seasonal overlay: mark festive CPM windows on every curve, because a Q3 efficiency dip is usually the auction, not your saturation point. Between them, these reads answer the question the scenario sheet needs: which channel deserves the next shifted rupee, and which has stopped arguing convincingly for its current one.

What does the weekly hour actually look like?

The system lives or dies on one recurring hour, so here is its anatomy. Minutes one to fifteen: close the baseline. Last week’s row lands — spend by channel, blended revenue by destination, MER, contribution on delivered orders. No discussion yet; discussion before the row closes is how numbers get negotiated instead of recorded. Fifteen to thirty: variance, not narrative. Only lines that moved beyond their normal range get airtime, and each gets one owner and one sentence in the log: what moved, best current explanation, what would confirm it. Resist the weekly temptation to re-explain the whole account — the sheet remembers so the meeting does not have to. Thirty to forty-five: the credit and curve check. Did anything this week contradict a discount factor or a curve read? A retargeting line suddenly outperforming its 0.5 factor is either noise, a creative change, or a reason to schedule the next holdout early — the meeting’s job is to pick which and log it. Forty-five to sixty: one decision. The best weekly meetings move exactly one thing — a budget shift, a test brought forward, a quota conversation with the creative partner — because a meeting that moves five things a week is reacting, not steering. Founders are welcome at this hour on one condition: decisions still route through the log and the decision-rights page, because the fastest way to corrupt a measurement system is a senior voice that overrides it casually. Run this hour for a quarter and the artefacts stop being sheets and become the account’s memory — which is the asset most scaled brands discover they never had.

Where does MMM-lite break?

Respect the limits, because they are real. MMM-lite cannot estimate adstock — the delayed effect of upper-funnel spend — so it will systematically undervalue brand video unless you correct it with geo experiments, which remain the honest way to price awareness work. It cannot separate correlated channels that move together, which is why the credit sheet leans on holdouts rather than curves for harvesting lines. And it degrades exactly when things get interesting — big launches, distribution jumps, festive quarters — because the baseline shifts under the sheets. Graduate to a real model when the account clears roughly ₹1 crore a month across four or more meaningful channels, when quick commerce and offline muddy blended revenue beyond what geo splits can untangle, or when a genuine analyst joins — and when you do, the five sheets become the calibration inputs and sanity checks that make the real model trustworthy. Until then, the published ranges in our D2C Spend Index are the external sanity check each sheet gets read against, and the credit sheet is the page a sceptical CFO should be shown first.