Marketing Mix Model
From Data to Action
Every Signal in One Place
Measurement numbers alone tell you what happened. To decide what to do next, you need to know why — and the why is almost never in a single data source.
Maxma maintains data pipelines across your full marketing stack and keeps them reconciled in one place:
- Ad platform delivery and configuration — spend, impressions, reach, frequency, CPM, and campaign settings from every connected platform
- Attribution sources — platform-reported conversions, Google Analytics, and UTM-based tracking
- Causal measurement — ongoing MMM incrementality and incrementality test results
Each source answers a question the others can't. Platform attribution is fast but self-graded; GA shows the click path but misses view-driven impact; MMM and incrementality tests measure true lift but don't explain the mechanics on their own. Side by side, they check each other — when platform CPA looks great but incremental CPA is climbing, that gap is the insight. Creatify's team hit exactly this: platform, UTM, and KPI trends each told a different story, and the unified view revealed pMax conversions concentrating in mobile inventory that drove no real lift — leading them to exclude it, scale spend 4× in 30 days, and cut incremental CPA ~40%.
From "What Changed" to "Why It Changed"
When the MMM shows cost per incremental conversion jumping after a budget increase, the number itself doesn't tell you what to fix. Answering that means joining the MMM trend with delivery data, campaign settings, and audience signals — work that traditionally takes an analyst days, and that most teams never get to at all.
Maxma's built-in analytics agent runs that investigation in minutes, on demand or on its own. The two examples below are condensed from real deep dives (names and numbers adjusted).
Example: a video budget scale-up that got expensive
A team scaled video spend roughly 3.5× in a few weeks. MMM cost per incremental conversion rose ~60%. The agent joined the MMM trendline with platform reach, frequency, CPM, and campaign configuration, and decomposed the efficiency loss into causes — sized, so the fixes could be prioritized. Here's what the report looks like, condensed:
- Video spend scaled ~3.5×; MMM cost per incremental conversion rose ~60%.
- Three causes, unequally responsible: retargeting saturation (~50%), diminishing returns at scale (~30%), auction cost inflation (~20%).
- Fixing retargeting alone closes about half of the efficiency gap.
| Driver | Share | Evidence |
|---|---|---|
| Retargeting saturation | ~50% | A quarter of spend on very small retargeting audiences served many times per week — already in-funnel, near-zero incremental value |
| Diminishing returns | ~30% | Spend scaled ~3.5×, incremental conversions ~2× — response-curve concavity at higher spend |
| Auction cost inflation | ~20% | Prospecting CPMs nearly doubled as the spend surge heated up the auction |
- 1Pause the sub-scale retargeting pools; add weekly frequency caps to the rest≈½ the gap
- 2Cap retargeting as a share of channel spend; redirect the budget to prospectingincluded above
- 3Re-scale in modest weekly steps toward the response curve’s efficient range≈⅓ the gap
The full report backs each row with the underlying charts and campaign-level tables — every claim traceable to the data it came from.
Example: where did a 3× budget increase actually go?
Another team ramped brand spend 3× in a few weeks and watched cost per incremental conversion multiply. The agent broke the spend increase down with a simple identity — spend = reach × frequency × CPM:
- 1.6× more people reached — the only part that can generate incremental conversions
- 1.4× higher frequency — repeat impressions on people already reached
- 1.35× higher CPM — auction inflation from the spend surge
Well under half the increase bought new people. Reach had plateaued and 28-day frequency hit 4× — the audience pool was saturated, and more budget was buying repetition, not coverage. The prescription followed directly: pull spend back to the efficient range, cap frequency, and expand audiences horizontally before scaling vertically again.
Execute Directly From Maxma
A recommendation you still have to implement by hand across three ad platforms isn't much of a shortcut. When a change is worth making, Maxma packages it as an action proposal: the exact changes, listed campaign by campaign, waiting on your approval — nothing touches your accounts until you approve.
Video is past its saturation point while TikTok prospecting sits on the steep part of its response curve. This shifts the budget across five campaigns — net weekly spend unchanged.
Video_Prospecting_NAcampaign-25%- Video_Prospecting_EUcampaign-25%
- Video_Retargeting_NAcampaign-50%
Prospecting_Broad_NAcampaign+40%- Prospecting_Interest_EUcampaign+40%
- Net weekly budget change$0
One click applies the whole change set across platforms — here, pulling budget out of saturated video and moving it into under-spent prospecting in the same proposal. Every executed change is recorded with its before and after state, and the next weekly measurement shows what the move actually did.
An AI Teammate in Slack
You don't have to be the one watching. Maxma's AI analyst monitors your measurement as it refreshes and messages your team in Slack when something needs attention — cost per incremental conversion breaking the guardrail you set, reach flattening while spend keeps growing, a channel's contribution shifting faster than usual.
This week's MMM refresh shows Video's cost per incremental conversion up ~35% over three weeks while spend roughly doubled. The response curve puts current spend well past the efficient range — the added budget is mostly buying frequency on people already reached.
- • Weekly reach is flat while frequency keeps climbing — spend is buying repetition, not new people
- • Cost per incremental conversion has been above your guardrail two weeks running
- • Pull weekly Video spend back toward the response curve's efficient range
- • Expand audiences before scaling vertically again
Want the full deep dive? Reply here and I'll put it together.
It communicates like a teammate, not an alerting system: what changed, why it likely changed, and what it recommends — and you can ask follow-up questions or request a deeper dive right in the thread. Proactive monitoring can be tailored to what you care about, so the first time you hear about a problem isn't in next month's reporting.