Connect any AI agent through MCP or the read-only API to your Shako Stats measurement context: processed data, data dictionary, metric definitions, tests, power analyses, lift readouts, diagnostics, model outputs, and marginal CAC. Ask in plain language and get answers grounded in your own data and Shako's statistical outputs.
The AI explains the statistical outputs. It does not replace them, edit them, or launch anything on your behalf. Everyone on the team can ask, from analyst to CMO.
Your AI · connected to Shako Stats (read-only)
What's my marginal CAC on Meta right now — and should I shift budget?
Your Meta prospecting iCAC is $41(95% interval $33–$52) from your April geo test. At today's spend, the calibrated MMM puts marginal CAC at $63 — past your $55 target. Shifting ~15% of Meta budget toward Search keeps blended CAC on target with the smallest revenue risk.
Was that geo test clean enough to trust?
Yes. The holdout was balanced on pre-period sales, the counterfactual fit was tight, and both models agree within 4%. The lift cleared your minimum detectable effect — I'd treat it as decision-grade.
Illustrative conversation — every number is computed from your own tests and models, not generic advice.
Shako Stats runs the measurement models and deterministic calculators. Your AI reads the approved context, including the data dictionary and metric definitions, then explains what the evidence says, what is uncertain, and what decision it supports.
A textbook explains the method and leaves the work to you. Your AI Marketing Scientist connects to your actual tests, processed datasets, metric definitions, designs, results, diagnostics, and statistical outputs, then answers the question you asked with your numbers.
“Will I have enough power if the true iCAC is $45?” “Which tests proved out?” “Is this result clean?” “Do the models agree?” It answers from the designs, power analyses, diagnostics, and readouts in your workspace.
iCAC, iROAS, marginal CAC at your current spend, and budget-shift tradeoffs, grounded in your incrementality tests and calibrated statistical outputs, with the uncertainty shown clearly.
Walls down. A junior analyst, a growth marketer, a data scientist, a VP, or the CMO can ask about the test that just ran and get the same rigorous, plain-language answer.
Connect Claude and ChatGPT with one click. Cursor, Claude Code, Gemini CLI, Codex, or any MCP client connects with a read-only access key in minutes. The REST API is available for code agents and pipelines too.
Your agent can read tests, results, metric definitions, data dictionary context, and approved statistical context. It can never create, launch, edit, or change anything. Portfolio keys follow organization permissions and are revocable anytime.
Connect once, then just ask. The reasoning happens in your AI; Shako Stats serves the data, definitions, precomputed model outputs, calculators, and coaching.
Add Shako Stats as a connector in Claude or ChatGPT with one-click OAuth, or generate a read-only access key for Cursor, Claude Code, Gemini, Codex, or any MCP-compatible AI agent.
“If Meta's true iCAC is $45, will my test have the power to detect it?” “What's my marginal CAC?” “Read last month's holdout. Is it clean?” No SQL, no exports, no ticket queue.
Your agent reads the same designs, power analyses, lift results, metric definitions, diagnostics, coaching playbooks, deterministic calculators, and precomputed statistical outputs your dashboard uses, so answers are grounded in real measurement context instead of guessed.
Numbers with uncertainty, what they mean, and what to do next, in language you can forward to the team or take into the budget meeting.
We don't make you switch chat apps. Shako Stats plugs into your AI through MCP/API: one-click OAuth for Claude and ChatGPT, a read-only key for everything else.
Keys and one-click OAuth connections both live under AI MCP / API Access. Both are read-only and revocable anytime.
Other providers hand you a single blended number. We show the model evidence because you can't defend a decision you can't see into.
Every test result is read by multiple counterfactual models, not one. When they agree, you know the lift is real. When they don't, we show you the disagreement, not a cherry-picked winner.
65% and 95% confidence intervals, everywhere a number appears. You see the most-likely range and the cautious one. Uncertainty is part of the answer, never hidden inside it.
Ad platforms make money selling more ads. That's their incentive, and it's fine, but it isn't ours. We aren't partnered with them, and nobody gets to sway the results we show you.
You kick off the design or results readout. When the models finish, the numbers go straight to your workspace. No human review gate, no editorial pass.
If nothing agrees, we say so. That might be embarrassing for a provider selling certainty. We're doing statistics with the best techniques we know and showing you what they found.
Measurement knowledge shouldn't live behind one analyst's dashboard. When anyone from a junior analyst to the CMO can ask about the test that just ran and why the result is what it is, the whole org gets smarter.
Your agent connects to a proper machine interface over your measurement workspace: the same numbers your dashboard shows, served in a form AI can reason over.
Your AI Marketing Scientist reads your geo, user-level, and platform incrementality context together: processed data, metric definitions, designs, power analyses, lift readouts, diagnostics, model agreement, and calculators. The answer to 'what should I do with my budget?' reflects the evidence your workspace actually contains.
Inform. Attribution gives the MMM a fast, directional read on what's working between tests.
Calibrate. Incrementality tests anchor the MMM to causal ground truth, not just correlation.
Validate. Holdout tests check that attribution and the model agree with what really happened.
The models and tests matter, but the workflow around them matters too. Shako Stats is designed to become the operating system around experiment planning, metadata, documentation, and cross-test learning.
Centralize datasets, mappings, and historical records so experiments and models always start from the same source of truth.
Organize tests by audience, creative strategy, bidding logic, or business objective so learnings remain searchable and reusable.
See what tests are planned, in-flight, or completed so overlapping interventions and measurement conflicts are easier to manage.
Turn methodology, definitions, and experiment design guidance into an internal operating system instead of leaving them scattered across decks.
Use setup guides and starter prompts to get your AI connected and answering questions on day one.
Paid B2B plans built for brands & agencies.