Lauren, you walked me through four brands on four separate budgets, and an MMM that left you doing the analysis on your own.
This page covers how Fospha would run an MMM with you as a partner, starting with Learning Resources, and how awareness gets measured against Amazon Sponsored Ads. Come back to it whenever the new headcount is in place.
Three things worth taking from this
- 1You need an MMM partner who can stand next to you when the CFO asks detailed questions. "I felt like I wasn't set up for success." Fospha puts an account manager, a technical account manager, a customer intelligence team and a measurement science team behind the model, and they read out the results with you.
- 2About half of your spend sits in Amazon Sponsored Ads, where no platform can tell you what awareness is worth. Fospha models impressions alongside clicks, so upper-funnel channels get credit for the Amazon sales they drive.
- 3The MMM covers the quarter. The daily model covers the weeks in between, with saturation curves and confidence ranges so each budget move has a number behind it.
An MMM you do not run alone
Fospha can set up each brand as a separate instance, so budgets and models never blend. Learning Resources goes first, and the other three follow when you are ready.
- Our measurement science team builds the MMM and presents the results with you, including to senior leadership.
- An account manager and a technical account manager sit alongside them, and the team helps you agree a realistic pilot goal against your 2.0 Profit ROI target.
- The team helps with data collection, and your ad platforms connect by API.
- Forecasts and saturation curves come with confidence ranges, so you can see how far to trust each one before you move budget.
Proving awareness beyond Amazon ads
About half of Learning Resources spend sits in Amazon Sponsored Ads, and no platform can tell you what the awareness spend around them is worth.
- Fospha models impressions and views alongside clicks, so channels higher in the funnel get credit for the sales they drive.
- Where those sales land on Amazon, total-funnel measurement traces them back to the campaigns that drove them, with Target and Walmart coming early 2027.
- Saturation curves with confidence intervals show where a channel still has headroom before returns flatten.
- Fospha's measurement science team can design and run Geo Lift tests, and the results can calibrate forecasting and the mix model.
One measurement system, three ways to answer your questions
Lauren, you need an MMM, a daily view for the weeks in between, and a way to prove what awareness is worth against Amazon Sponsored Ads. Fospha runs all three on one set of reconciled data, with no pixel to install, so each result sharpens the others and your team plans on a single set of numbers.
The daily model every Fospha customer runs on. Impression-led credit across every channel and campaign, down to the individual ad, reconciled to your real backend revenue. Connects by API with no pixel.
A whole-business read built on Google's open-source Meridian framework and delivered by our measurement science team. It separates baseline demand from the incremental contribution of each channel.
Geo Lift experiments designed and run by our marketing scientists, measured on your own orders, with the results fed back into the numbers you already plan on.
MMA is the daily anchor. The MMM and Incrementality Testing attach to it and calibrate against it, so you plan on one set of numbers.
How each layer works
Pick a layer to see how it is built, what goes in and what comes out, then open the full explainer to share with your team.
Click-based measurement over-credits the bottom of the funnel. Brand search, direct and email collect the credit for demand that YouTube, TikTok, Meta and influencers created earlier. Fospha corrects that bias in six steps.
- Baseline. Raw GA4 last non-direct click conversions, every day.
- Validation. Checked against your ecommerce platform: orders deduplicated, excluded transactions removed, UTMs re-bucketed.
- Click measurement. Last click, data-driven attribution and machine learning models, averaged so no single method decides the answer.
- Post-purchase. Surveys and discount codes credit influencer activity that clicks miss.
- Reconciliation. Any shortfall is redistributed against your ecommerce orders, so 100% of your sales are modelled.
- Impression measurement. A daily media mix model moves credit from brand search and direct to the channels that created the demand.
YouTube, TikTok and Snapchat rarely get credit when the sale lands in Direct, Organic or brand search. The daily model measures what their impressions actually drove.
Captures the delayed impact of impressions. Conversion activity peaks on the day, awareness activity keeps working for days and weeks.
Your offline spend calibrates the model, so digital channels do not take more credit than they earned.
An ensemble handles non-linear effects and overlapping channels, and Shapley values keep credit assignment stable and fair over time.
On days when sales spike for reasons the model cannot see, such as Black Friday or an influencer mention, capping limits how much credit view-through channels can claim.
Every day the model retrains and is tested on unseen data. A version that misses our accuracy thresholds never reaches your dashboard.
Model Spotlight shows each step of how credit moved, so your team and your finance team can see exactly where every number came from.
Meridian splits observed revenue into a baseline, what you would earn with no media at all, and the incremental contribution of each channel, digital and offline. It sets direction for planning and budgets, where MMA distributes budget day to day.
A peer reviewed Bayesian framework your own analysts can inspect.
Every digital channel already flows through Fospha, so the MMM starts from a stronger base.
We do not rebuild the framework, so effort goes into data checks, quality assurance and testing against our own models.
- Prepare data. Transform and quality review.
- Model fit. Assemble, shape and fit the model.
- Validate. Convergence and fit checks, with channel specific adstock windows rather than one carry-over window for all paid media.
- Quality assurance. Our measurement science team checks the result.
- Present insights. ROI and budget optimisation, walked through with you.
The best fitting model is not automatically the best causal one, so it also has to estimate the unobserved baseline correctly. Where a brand has run lift tests or geo experiments, we calibrate against them. Inputs are weekly channel spend and KPIs, offline activity, and promotion, pricing and calendar context. This is a managed engagement run by our measurement science team.
Incrementality Testing answers the question every finance team asks: would these sales have happened anyway?
Divide the country into test and control markets, chosen so the synthetic control genuinely predicts what the test markets would have done.
Change spend in the test markets only. Hold out, scale up or cut back. Everywhere else carries on as normal.
Measure the gap in real orders between what happened and what the control predicted. That gap is the incremental effect.
- Power analysis first. If the data cannot support a valid read, we say so before you spend, with a go or no-go gate at no cost.
- Lift, iROAS and confidence. Incremental revenue, lift percentage and cost per incremental acquisition, each with its confidence range.
- A readout that ends in a decision. Run by a dedicated Marketing Scientist, not a dashboard left for you to interpret.
- Calibration. Results feed Incremental Forecasting and the marketing mix model, so the numbers you plan on get sharper.
| The decision | Use | Why |
|---|---|---|
| Annual and mid-year budget split, including offline | MMM | Whole business view with the baseline separated and confidence intervals shown. |
| Daily bids, creative and campaign moves inside a platform | MMA | Daily, ad level credit reconciled to backend revenue. |
| Proving a channel is truly incremental before moving real budget | Testing | A causal read on your own orders, not platform reported conversions. |
| Where the next dollar still has room to grow | Incremental Forecasting | Saturation curves on MMA output with confidence ranges, sharpened by test results. |
For overall, budget level decisions, trust the MMM, especially where its confidence intervals are tight. For granular, campaign level decisions inside a platform, trust MMA. Use the MMM to set the headline split and MMA to execute within it.
The full explainer on how MMA, the MMM and Incrementality Testing work together, and the science behind each one. Twelve slides you can share with your finance team.
Open the explainer →Brands seeing the same results
Topicals could see GMV Max results only on TikTok Shop. Fospha tracked the impact across TikTok Shop, Amazon and DTC in one view, and the brand raised GMV Max investment 460% in February.
Abyat's team struggled to defend budget calls to finance and leadership. Model Spotlight made credit visible and explainable, so the numbers held up when management challenged them.
Next steps
- Will will send the follow-up email with materials and offer details this week.
- Will will check with our MMM team on shipped versus POS sales.
- Lauren to review the materials, then pick a demo time, Q1 or sooner.
Platform walkthrough
Full walkthrough of the platform, reporting views, and daily workflow.
Watch videos →