Credit is counted twice
Each platform counts the same order as its own, especially retargeting and brand search.
Marketing Mix Modeling · Philippines
Stop guessing which ads actually sell.
Marketing mix modeling built for Philippine ecommerce: your website and every marketplace in one view, with promos, holidays and every other driver of your sales accounted for, explained in plain language.
- Total revenue, not platform clicks
- Every number with an honest range
- Analyst-reviewed before release
Synthetic demo brand. Figures are illustrative.
The problem
Every platform claims the sale.
Meta, Google, TikTok and the marketplaces each report their own conversions. Added together they claim more revenue than you made, and none of them can tell you what would have happened without the ad.
Sales sit in silos
An ad on one platform can sell on another: a TikTok ad can sell on Shopee or on your website.
Sale days blur the picture
9.9, 11.11, paydays and holidays lift demand on their own, and get mistaken for ad results.
The product
From total revenue to next month's budget.
Revenue breakdown
Modeled revenue split into baseline demand, promotions and sale events, holidays and seasonality, and each paid channel, in pesos and percent.
- Effects that reduced revenue, such as slow seasons, are shown as negative, not hidden.
- Modeled revenue sits next to actual revenue, with the unexplained gap shown openly.
Channel performance
Spend, incremental revenue, ROAS and the return on the next peso for every channel, beside impressions and clicks as each platform reports them.
- Marginal ROAS below 1.0 means the next peso returns less than it costs.
- Open a channel to see the tactics the data can separate reliably.
Where the next peso works
Response curves show how revenue grows with spend for each channel, where you are today, and where spend has never been tested.
Budget planner
Enter next month's budget. We forecast revenue if you keep your current split and with the split the model recommends, channel by channel, for the same budget and dates.
- Changes stay within what your data supports.
- Channels the model cannot measure reliably stay as they are.
- Save plans, download them or share a one-page PDF summary.
Screens show a synthetic demo brand. Figures are illustrative.
How it works
From your exports to your first plan.
- 01
Share data
18–24 months of daily data: ad exports, store and marketplace orders, promotions. CSV or our sheet template.
- 02
Set up
We map channels, campaigns and stores, run data quality checks and resolve issues with you.
- 03
First model
Trained, checked and reviewed by an analyst, then released to your dashboard.
- 04
Keep improving
Refreshed on your plan's schedule; experiments sharpen results over time.
Methodology
Honest by design.
Daily revenue is modeled as a base level, the yearly season, the sales calendar and promotions, and each paid channel's response to its spend. Carry-over and diminishing returns are built in.
Ranges, not guesses
Every estimate is a probability distribution, so every number comes with an honest range.
Held back, not guessed
Channels your data cannot separate are flagged and kept out of results and plans.
Sharpened by experiments
Lift-test results calibrate the model. Until experiments confirm them, results are labelled directional.
Checked before you see it
- Every model is checked for fit, forecasting and whether each channel can be measured.
- An analyst reviews and approves every release.
- Model results are not, on their own, proof of causation, and we say so.
Your data, protected
- Private workspace per brand; access by invitation only.
- Two-step sign-in for analysts and admins.
- Customer names and addresses are never stored; order IDs are anonymised on upload.
Pricing
Plans that grow with your channel mix.
Every plan includes the full Bayesian model of your total revenue, analyst review before release, and the budget planner. Plans differ in how many channels and stores we model, how often we refresh, and how much support you get.
Starter
Monthly ad spend up to ₱500K
₱9,900 /mo
₱2,500 one-time onboarding
- 3 ad channels · 2 stores
- Channel-level results
- Monthly model refresh
- 2 budget scenarios / month
- 2 users · Email support
Growth
Monthly ad spend ₱500K–₱2M
₱24,900 /mo
₱5,000 one-time onboarding
- 6 ad channels · 4 stores
- 12 modeled tactics
- Every 2 weeks model refresh
- 10 budget scenarios / month
- Experiment calibration add-on
- 5 users · Priority email support
Advanced
Monthly ad spend ₱2M–₱10M
₱49,900 /mo
₱12,500 one-time onboarding
- 12 ad channels · 8 stores
- 21 modeled tactics
- Weekly model refresh
- 30 budget scenarios / month
- 2 calibration studies / quarter
- 10 users · Monthly analyst review support
Enterprise(for multiple brands)
Monthly ad spend ₱10M+
from ₱100,000 /mo
from ₱30,000 one-time onboarding
- Custom channels and stores
- Custom tactic detail
- Custom refresh schedule
- Custom budget scenarios
- Custom experiment calibration
- Custom users · SLA/dedicated support
Prices in PHP per month, before VAT.
Start with a 30-day pilot
We set up your model, review the results with you, and plan your next budget together.
FAQ
Questions brands ask us.
How is this different from the ROAS in my ad accounts?
Platform ROAS counts the conversions each platform can see and credits itself. MMM PH models your total revenue across every store at once, so the same order is never counted twice and sales that happen on a different platform from the ad are included.
What data do you need?
18–24 months of daily data: campaign exports with spend, impressions and clicks from Meta, Google, TikTok, Shopee, Lazada and TikTok Shop ads; order exports from your website and each marketplace; and your sale calendar. Stock-outs, site outages and delivery disruptions help if you have them.
Do you need our customers' personal data?
No. Customer names and addresses are never stored, and order IDs are anonymised on upload.
Does the model prove what caused my sales?
Not on its own. A marketing mix model estimates effects from historical data, with ranges. Lift tests are the strongest evidence; their results calibrate the model, and results are labelled directional until experiments confirm them.
What if a channel can't be measured?
If your data cannot separate a channel's effect, we flag it and keep it out of results and budget recommendations rather than guess.
Book a demo
See it on a demo brand.
We'll walk you through the platform on a synthetic brand, look at what data you have, and tell you honestly whether a model will work for you.