Marketing mix modeling software for ecommerce: the options
Two free open-source options, Google's Meridian and Meta's Robyn, plus paid vendors on quotes or tiers. Before choosing, build the weekly input file the software needs: Google says a national-level model wants at least three years of it.
By Joris van Huët, Founder & CEOPublished 5 min read
Run the numbers for your store: the free multi-channel budget calculator.
The free options are Google's Meridian and Meta's Robyn, both open source; the paid options are listed below with what each vendor publishes about price. The software is the easy part. Meridian requires Python 3.11 or 3.12 and a complete dataset without missing values, recommends weekly data, and Google's rule of thumb is at least three years of it for a national-level model, the kind you run without regional data.
What MMM software can an ecommerce brand use?
Open source. Google describes Meridian as an open-source MMM that is free for anyone to use. It requires Python 3.11 or 3.12, and Google recommends a minimum of 1 GPU. Meta's Robyn is an open-source package with R and Python (Beta) versions, and its documentation says it was updated in December 2024, so check the project's current state before you build on it.
Paid, from the comparison table. Each price is the vendor's listed pricing as read on the date shown, and it changes without notice:
- Recast: Bayesian marketing mix modeling, sold on a quote; the table lists no public pricing page (homepage).
- Prescient AI: media mix modeling in three tiers (SMB, Growth and Enterprise) by business size, adjusted for the number of sales channels, on annual contracts and quoted via a demo. Its pricing page says the platform is built for brands spending $100K+ a month on advertising (pricing, as read on 2026-09-26).
- Haus: geo-lift experiments, with Causal MMM built on them on its Plus and Enterprise plans. Each of its four plans starts with a demo and no price is listed (pricing, as read on 2026-09-26).
- Fospha: an impression-based MMM and multi-touch hybrid. Lite is $1,500/mo (€1,300/mo) for $100k to $500k in monthly media spend, Pro is $2,000/mo plus a share of media spend, and Enterprise is through sales (pricing, as read on 2026-09-08).
Northbeam, Rockerbox and Funnel.io list MMM inside larger attribution or data products; the comparison table has their entries, and Northbeam's prices have their own post.
What does the software need from you?
Meridian recommends weekly rows with a KPI such as revenue, media data by channel (spend plus an exposure measure such as impressions or clicks) and control variables. Its rule of thumb is a minimum of two years of weekly data for geo-level models and three years for national-level ones. Robyn's analyst's guide says an MMM needs a minimum of two years of historical weekly data, with a recommended ratio of 1 independent variable to 10 observations.
History is not the only limit. Meridian's own example calls four data points per parameter "too low to estimate the model reliably", and the guide by data volume works through that arithmetic. No source cited here gives a minimum number of orders for a small store: the thresholds are the tools' own guidance.
Both tools also take experiment results as calibration input, and Robyn's guide strongly recommends using them, so list the holdouts you have run. Robyn's documentation cites a third-party whitepaper finding that uncalibrated models show a 25% average difference to the ground truth; it gives no sample or date, and the figure is published by Meta, not audited here.
Can you build the input file? A 30-minute check
Four checks on one spreadsheet before you install anything:
- Weeks. One row per week, from the first week with spend to the last full week. Pass: at least 156 rows, three years, for a national-level model. Fail: fewer, so Google's rule of thumb is not met.
- Columns. Revenue as the KPI; for each channel, spend and an exposure measure; and your controls. Meridian lets you use spend as the exposure measure when impressions or clicks are unavailable, but warns that periods of high ad cost might then be read as high ad volume. Pass: every channel has both columns, or you accept that proxy. Fail: a channel with neither.
- Summable columns only. Meridian says rates and ratios such as CTR, CPC and ROAS cannot be used, and that you must compute raw volumes from them. Pass: only spend, impressions, clicks, units and revenue. Fail: a platform export that gives only rates.
- No gaps. Meridian requires a complete dataset. For a channel that was off, fill the gap with 0; for revenue or controls, impute the value by interpolation or a similar method, because a 0 skews the estimates. Pass: every cell filled the documented way. Fail: blanks, or zeros standing in for missing revenue.
If the file passes, count weeks per parameter and look for spend that varied; the guide by data volume has both checks. If it fails, fix the file first: a holdout can answer one channel's question while the history builds (incrementality testing).
Where does a one-off causal read fit?
A causal attribution read like Causality Engine's is a different tool. It takes one GA4 export (12 months or more reads best) and no spend data, and it shows what each channel caused next to what last-click gave it, with a data-health score and a next step for each channel. It gives no response curves or budget optimizer. One read is €99, once, excluding VAT, refundable within 30 days; Pro is €299 a month.
Sources, 30 September 2026: An introduction to Meridian, Install Meridian, Collect and organize your data and Amount of data needed (Meridian documentation, Google for Developers, 2026); Robyn, Key Features and An Analyst's Guide to MMM (Meta Marketing Science, documentation updated December 2024); Recast, Prescient AI pricing, Haus pricing and Fospha pricing (vendor pages).
Related answers
Frequently asked questions
Is there free marketing mix modeling software?
Yes. Google's Meridian is open source and free for anyone to use, and Meta's Robyn is an open-source package. You prepare the weekly data and run the model yourself, and Meridian requires a complete dataset without missing values.How much data does marketing mix modeling need?
Google's Meridian documentation gives a rule of thumb of a minimum of two years of weekly data for geo-level models and three years for national-level models; Robyn's guide says a minimum of two years of weekly data. Meridian calls its guidance rough and directional.Can a small ecommerce brand use marketing mix modeling?
Only if the data fits: enough weeks, few enough channels and controls, spend that varies, and a test or two to calibrate against. Meridian's own example calls four data points per parameter too low to estimate reliably. If the checks fail, start with a holdout test.
Go deeper: Causal attribution, explained.
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
Keep reading
Terms in this article
- AttributionAttribution identifies user actions that contribute to a desired outcome and assigns value to each. It reveals which marketing touchpoints drive conversions.
- Causal AttributionCausal Attribution uses causal inference to determine which marketing touchpoints genuinely cause conversions, not just correlate with them.
- ExperimentsExperiments are scientific procedures that test hypotheses or demonstrate facts. In marketing, experiments like A/B tests determine the causal effect of campaign changes, enabling data-driven decisions.
- IncrementalityIncrementality measures the true causal impact of a marketing campaign. It quantifies the additional conversions or revenue directly from that activity.
- Incrementality TestingIncrementality Testing measures the additional impact of a marketing campaign. It compares exposed and control groups to determine causal effect.
- Marketing MixThe marketing mix is the set of actions a company uses to promote its brand or product. It traditionally includes product, price, place, and promotion.
- Marketing Mix ModelingMarketing Mix Modeling (MMM) is a statistical analysis that estimates the impact of marketing and advertising campaigns on sales. It quantifies each channel's contribution to sales.
- Media Mix ModelingMedia Mix Modeling is a statistical technique that measures the collective impact of marketing and advertising on sales. It uses historical data to inform budget allocation.