What is the AI Marketing Scientist?

The AI Marketing Scientist lets you connect any AI agent through MCP/API to a read-only context layer over your Shako Stats workspace. Ask about geo incrementality tests, user-level holdouts, platform lift evidence, MMM calibration context, marginal CAC, test design, processed results, and whether a test will have enough power to detect the iCAC or iROAS you believe in. Answers are grounded in your own tests, model outputs, and diagnostics, with confidence intervals shown clearly.

Is the statistical model deterministic - will I get the same answer twice?

Yes. The statistical engine is deterministic and repeatable: the same data, design, and model return the same result every time, so your causal numbers are auditable and do not drift between sessions. Some models are sampled via MCMC, but they are seeded and converge to the same posterior. The engine is the authority; the AI explains it.

What measurement tools does Shako Stats offer?

Shako Stats starts with a deep geo incrementality testing workflow for designing powered market experiments and reading lift with multiple counterfactual models. The platform also organizes user-level holdouts and platform lift evidence, exposes results through MCP/API, and is expanding into MMM incrementality, calibration, forecasting, and budget-planning workflows.

Which AI tools does it work with?

Claude and ChatGPT connect with one-click OAuth. Cursor, Claude Code, Gemini, Codex, and other MCP clients can connect with read-only access keys. The same context is also available through a read-only REST API.

What is geo incrementality testing?

Geo incrementality testing measures the true causal impact of marketing by using geographic regions as test and control groups, comparing regions exposed to ads against a statistically built baseline of regions that were not.

Shako Stats builds a counterfactual from your historical data, a combination of control markets that mimics your test markets, then compares actual performance against it during the campaign to isolate the incremental lift.

Why use incrementality testing instead of last-click attribution?

Last-click attribution ignores views, brand awareness, and organic behavior, over-crediting branded search and under-crediting upper-funnel channels. Incrementality testing measures the lift you would not have earned otherwise, the ground truth of marketing performance.

Do I need a data science team to use Shako Stats?

No. Shako Stats is a no-code platform that automates the statistics, including synthetic controls, power analysis, and Bayesian models, so anyone can run rigorous tests by uploading a CSV.

A CSV with three columns: date, location (city, state, or DMA), and a metric (sales, revenue, or conversions). About 12 months of history gives the best accuracy, though the models can work with less.

Your AI Marketing Scientist on your real data

Start with our flagship geo incrementality workflow, then connect Claude, ChatGPT, Cursor, or any MCP client to your Shako workspace. Ask about geo experiments, user-level holdouts, platform lift evidence, power, lift, iCAC, marginal CAC, MMM calibration context, forecasting, optimization, and model agreement from your own measurement data, definitions, and statistical context.

We built the statistical engine. The AI you already use is the interface.

Geo incrementality today. User-level and platform evidence in the same workspace. MMM, forecasting, and optimization next. Data dictionary and metric definitions travel with every result, so your AI knows what the numbers mean.

Book a call Tell us what you need

AI connected to Shako Stats

Read-only MCP/API

Ask your AI

Plan the geo test before spend

Will we have enough power for a Meta geo test if true iCAC is about $45?

Yes. The best design reaches 82% power at that iCAC threshold. A faster 4-week design only reaches 61%, so I would use the 6-week plan.

Power analysisDMA balanceSpend history

Evidence

From the Shako workspace

Geo power at target iCAC

82%

6-week matched-market design

Recommended split

8 / 8

Balanced test and control DMAs

Model agreement

Passed

Multiple models agree within a point

One platform

Geo incrementality first. A broader marketing science platform next.

Shako Stats starts with transparent incrementality testing and turns every design, result, interval, diagnostic, data dictionary entry, metric definition, and planning context into evidence your AI can read.

Live nowGeo incrementality User-level holdouts Platform lift evidence AI Marketing Scientist Data API access

Coming soonMarketing Mix Modeling Forecasting & optimization

See it answer

Watch your marketing scientist work

Every thread below is a question your team already asks. Click any conversation — or let it cycle — and watch the answer come back from your real data: the number, its uncertainty, and the models behind it. Anyone on the team can ask.

Your AI, connected to Shako Stats

MCP/API read-only

12 conversations

Asked by a Growth leadGeo

Will I have enough power for a Meta geo test if the true iCAC is about $45?

Evidence from your workspace

Power

82%

Detectable iCAC

≤ $58

Split

8 / 8

Duration

6 weeks

Bring the AI you already use. Read-only, and every number is traceable.

See the AI Marketing Scientist

Why this is different

We built the engine. The AI is how you talk to it.

An LLM can describe a geo test or an MMM. It can't compute one on your data — it doesn't have the statistical engine, and it's never seen your numbers. So we built the engine: the deterministic source of truth. Then we put the AI on top as the tireless interpreter. Deterministic math you can reproduce, explained by an AI that never sleeps.

Your data

Your tests, designs, power analyses, lift, iCAC/iROAS, and diagnostics.

The Shako engine

Proven measurement models. Deterministic and repeatable — the same answer every time.

The AI you already use

Reads the engine's output and explains it. Teaches, interprets, never sleeps.

You decide

A number, its uncertainty, and what to do — grounded in your real data.

Two ways in. Nothing hidden.

The AI is an additional interface — not the only one. The dashboard and your AI read the exact same complete data.

The dashboard

Charts, tables, and graphs that explain every result. Every dataset we compute, every statistic over time — day by day — laid out plainly. We hide none of it.

Your AI — an additional interface

The same complete data, in plain language. List it, chart it, question it — interrogate your entire workspace however you want, with the AI you already use.

Generic advice

Can explain the method, but not compute your answer.

Without your workspace and the measurement engine behind it, you still do not have your answer.

Your AI + the Shako engine

Hands you the number — computed on your data.

The lift, its uncertainty, and the model agreement behind it. Advice is easy; the number is the product.

Deterministic & repeatable

Same data, same design, same model — the same answer, every time. Your causal numbers don't drift between sessions.

A PhD with your workspace in context

The AI has already read all of your data before you ask. Not generic advice — your scientist, on your numbers, at any hour.

Public models, proprietary design

We show every result model and name each one. The only thing we keep private is how we design balanced, high-power tests.

Explore the full platform

Built on better science

Every model. Every interval. Nothing hidden.

The AI explains Shako's statistical outputs. It does not replace them. When the models agree, your team can move with confidence. When they do not, everyone sees the uncertainty instead of a cherry-picked story.

Every model, side by side

Other providers may hand you one blended number. Shako shows every counterfactual model so agreement and disagreement are visible.

Two intervals everywhere

65% and 95% intervals appear anywhere uncertainty matters, so decisions do not depend on false precision.

Independent incentives

We are not partnered with ad platforms. Their incentive is selling more ads; ours is getting the measurement right.

No editorial pass

You kick off a design or readout, the statistical pipeline runs, and the result goes to your workspace without a human review gate.

"We do not know yet" is a valid statistical answer. That honesty is the point.

Read our builder's pledge

Private B2B plans

Build the measurement program your brands need

We’ll scope Incrementality Testing, Marketing Mix Modeling, the right tier, and pricing per brand. Seats are unlimited; products and brand access are configured for your billing account.

Book a call Send us your requirements

Priced per brand Private offer and secure checkout

Unified Marketing Measurement

The measurement engine
your AI can actually read

Shako Stats runs the measurement workflows and stores the evidence: test designs, power analyses, published lift results, data dictionary entries, metric definitions, diagnostics, model outputs, and budget context. The read-only MCP/API layer makes that evidence available to the AI your team already uses, so every feature becomes part of the same statistics-backed source of truth.

See how every method connects ->

AI Marketing Scientist

New!

Ask Claude, ChatGPT, Cursor, Codex, or any MCP client about your geo tests, user-level holdouts, platform lift evidence, metric definitions, data dictionary context, diagnostics, and marginal CAC.

Geo Incrementality

Early Access!

Our flagship workflow: design high-power geographic holdouts, compare multiple counterfactual models, and read lift, iCAC/iROAS, and intervals transparently.

Data API Access

New!

Pull published measurement evidence, metric definitions, lift, iCAC, iROAS, intervals, and model agreement into your warehouse, BI, reporting, and AI agents.

Coming into the broader marketing science stack

Geo incrementality is the flagship workflow today. These modules keep the same measurement philosophy as the platform expands into MMM, forecasting, attribution, user-level tests, and audience science.

Your AI Marketing Scientist on your real data

Plan the geo test before spend

From the Shako workspace

Geo incrementality first. A broader marketing science platform next.

Watch your marketing scientist work

We built the engine. The AI is how you talk to it.

Your data

The Shako engine

The AI you already use

You decide

Two ways in. Nothing hidden.

The dashboard

Your AI — an additional interface

Deterministic & repeatable

A PhD with your workspace in context

Public models, proprietary design

Every model. Every interval. Nothing hidden.

Every model, side by side

Two intervals everywhere

Independent incentives

No editorial pass

Build the measurement program your brands need

The measurement engine your AI can actually read

AI Marketing Scientist

Geo Incrementality

Data API Access

Coming into the broader marketing science stack

Marketing Mix Modeling

Timeseries Lift

Business Forecasting & Optimization

Platform Lift Evidence

User-Level Incrementality

Multi-Touch Attribution

User Level A/B Creative Testing

Lifetime Value Modeling

Average Revenue Per User

Prospect User List

Lookalike Modeling

Housefile Modeling

Persona Segmentation

And More...

Transparent where others are a black box

Frequently Asked Questions

What is the AI Marketing Scientist?

How is this different from just asking ChatGPT?

Is the statistical model deterministic, will I get the same answer twice?

What measurement tools does Shako Stats offer?

Which AI tools does it work with?

Can the AI change tests or launch campaigns?

What is Geo Incrementality Testing?

How does it work?

Why use this instead of Last-Click Attribution?

How do pricing and access work?

Do I need a Data Science team?

Is my data safe?

What data do I need?

Who is Shako Stats for?

What is your philosophy? (The Builder's Pledge)

Is Shako Stats Right for You?

Join the Newsletter

Master Marketing Science

Your AI Marketing Scientist on your real data

Plan the geo test before spend

From the Shako workspace

Geo incrementality first. A broader marketing science platform next.

Watch your marketing scientist work

We built the engine. The AI is how you talk to it.

Your data

The Shako engine

The AI you already use

You decide

Two ways in. Nothing hidden.

The dashboard

Your AI — an additional interface

Deterministic & repeatable

A PhD with your workspace in context

Public models, proprietary design

Every model. Every interval. Nothing hidden.

Every model, side by side

Two intervals everywhere

The measurement engine
your AI can actually read

The measurement engine
your AI can actually read