MARKETING MIX MODELLING. LEARNED BY DOING.

Know what works.
Know where
to spend.

Stop taking the model’s word for it. Learn to build it, challenge it and turn the results into a marketing budget you can defend.

Start learning MMM

Try a Python lesson

Realistic marketing scenarios Code, not just concepts

Explore the full curriculum

Blender model of weekly sales split into baseline, TV, paid search and paid social, with a fitted sales line

↗Challenge the resultCalibrate the model with experiments.

BaselineTVPaid searchPaid social

THE MMM LEARNING PATH 18 guided weeks Excel · R · Python Models → Decisions

A better question than “what’s our ROAS?”

What would you do
with the next £100k?

That is the skill you are here to build. Connect what the data says to what the business should do next. Start with a small decision, right here.

YOUR FIRST BUDGET DECISIONInteractive preview

Same budget.
Different possibilities.

Move spend between two channels. Watch diminishing returns change the combined outcome.

Paid social£50k

Paid search£50k

£100k fixed budget. Invented response curves for learning, not a business forecast.

Response to spendIllustrative revenue

Combined illustrative revenue£112.7k

● Paid social● Paid search

Learn how to build and challenge these curves

Inside the learning experience

Build your model.
Build your judgement.

Go beyond knowing the terminology. Work through the mechanics, see the trade-offs and learn which questions to ask before acting.

01 / MMM SKILLS

Understand carryover

An ad ends. Its effect might not.

Explore how advertising can carry into later weeks. Build an adstock transformation and understand the assumption inside it.

Explore adstock

02 / MMM SKILLS

Read the response curve

More spend. Smaller returns.

See why average ROAS is not the next pound’s return. Work with saturation curves and compare marginal opportunities.

Explore marginal returns

03 / MMM SKILLS

Make the budget decision

Put the next pound to work.

Move from a fitted model to a practical allocation. Weigh channel response, constraints and uncertainty before recommending a change.

Explore allocation

04 / MMM SKILLS

Challenge the answer

A good fit is only the beginning.

Bring experiments into the picture. Check whether the model’s story survives holdouts, diagnostics and outside evidence.

Explore calibration

A spreadsheet foundation. Then R, Python and advanced methods. See the full curriculum

A little less watching. A lot more doing.

Your first experiment
starts right here.

Edit real Python, run it in your browser and see what changes. These three exercises preview the coding labs in our new learning app.

Build your first carryover transformation

A campaign can affect later weeks. This unnormalised transformation starts at the full week-one spend. The decay argument is the fraction of the previous state retained each week. It is a model input, not measured sales lift or extra media spend.

your_example.py Editable

Complete the TODO, then run the code. Your code stays in this browser.

Terminal · PythonReady when you are

Run the example to see your output here.

What should the example produce?

[100.0, 50.0, 25.0, 12.5]

This is an unnormalised exposure state: week 1 is 100, not the normalised concept lab’s 50 at 50% retention. The state is not an observed sales effect or extra spend. Run again at 80% retention and compare week 4.

The first run downloads Python (about 13 MB). The numbers are invented teaching examples, not estimates of your business performance.

Want the reasoning behind the code? Explore carryover, compare marginal returns or learn to validate a model.

Your practice ground for real decisions

The point is not
to finish a video.

It is to leave knowing what you would do next. Vexpower brings the explanation, the experiment and the decision together.

Find your learning plan

Step into the brief

Start with a realistic marketing problem. Know the business question before opening the model.

Do the work

Change an assumption. Run the code. Inspect the result. See what survives your own testing.

Make the recommendation

Turn an output into a decision, with the limitations and trade-offs made clear.

Learn with Mike Taylor & James Phoenix

Marketing judgement.
Technical depth.

Two complementary perspectives on making marketing more practical, analytical and testable.

Mike Taylor Marketing measurement & experimentation

James Phoenix Python, data & applied AI

Before you start

A few useful answers.

Frequently asked questions

Do I need to know Python before starting?

The MMM learning path starts with spreadsheets before progressing to R, Python and more advanced methods. You can try the browser exercises here without installing Python, and load a worked solution when you need one.

What will I practise?

You will work through modelling and measurement decisions: preparing data, transforming media, understanding contribution, checking a model and evaluating a budget recommendation. The scenarios use invented data for learning.

How do I get access?

Choose a learning plan on the Vexpower pricing page. You can explore the curriculum, free guides and homepage Python preview before choosing a plan.

Explore before you enrol

Follow the question that brought you here.

Start in the marketing mix modelling learning hub. Learn how carryover works, then use the residual diagnostics guide to see why a convincing fit still needs checking.

Marketing mix modelling Incrementality and experiments How marketing mix models work Marketing budget decisions MMM tools, code and platforms

Also explore prompt engineering →