Measurement that ends in a decision.
The software we're building to run the practice: the five studies, the triangulation engine, and the honesty rules, turned into a working console. Verdicts first, evidence one click deeper, uncertainty never hidden.
01Overview
Verdict first
The overview opens with a sentence: what marketing contributed this period, whether it's improving, and the single biggest opportunity. Alerts for anomalies, stale sources, and method disagreements sit right under it, because those should change how you read everything below them.
02Optimizer
Budget optimizer
Reallocation as a direct interaction: drag a channel's spend and watch projected incremental revenue respond, or let the optimizer propose the best plan at the same total spend. Every channel is labeled with its evidence tier, recommendations stay within the spend ranges the evidence actually covers, and scenarios can be saved and compared against what really happened.
03Triangulation
Triangulation, made visible
The methodology's core loop as a product surface: per channel, the model's read next to the experiment's read. Agreement is stated plainly; disagreement is flagged with 'trust the test' guidance and a one-click path to design the next experiment. Thin, modeled-only estimates prompt a verification test instead of pretending to be proof.
04Channels
Response curves per channel
Each channel gets a plain-language verdict backed by its response curve: where current spend sits, where saturation begins, and what the next dollar returns. The analyst depth adds intervals and methods; every claim links to the methodology explainer that justifies it.
05Results · analyst depth
Every measurement, with its uncertainty
The results ledger shows every published measurement as it actually is: point estimate, interval, method, evidence tier, and publish status. Interval bars are drawn against the break-even line so 'confidently paying back' is visible at a glance. It all exports to CSV, because your numbers are yours.
06Data health · analyst depth
Data health & settlement
Measurement is only as good as its inputs, so the platform shows theirs: per-source freshness against expected cadence, anomaly-gate results with thresholds, and the settlement boundary that separates final data from numbers a platform may still restate. A result never publishes on a window the source is still restating, and stale or anomalous sources are flagged loudly instead of quietly averaged in.
07Studies
Study runs & QA gates
The study lifecycle is public inside the product: each run carries its QA gates (convergence, effective sample size, posterior predictive checks, out-of-sample coverage), and a result only becomes publishable when all of them pass. Exploratory work is labeled 'not decision-grade' so it can never masquerade as evidence.
08Methodology · analyst depth
The deep end, one click away
Progressive disclosure ends at the methodology view: empirical posterior densities per channel, credible intervals, model-health diagnostics, and plain-language explainers for every concept the console uses. Marketers get the verdict; analysts get the distribution it came from.
09Analyst · ask anything
Ask the evidence
“Which channel improved most since last quarter, and how sure are we?” Ask it in words and the AI answers from your isolated, governed data. It reads only the figures the deterministic pipeline already published, carries their uncertainty into every answer, and cites the methodology when you ask why. The measurement itself is never generated.
Built for the practice, shaped by the methodology.
The console renders the same evidence tiers, QA gates, and reconciliation rules described in our methodology. Client engagements run on it today, and it's evolving with every study we run.
Your data, treated like it's yours.
Isolation enforced in the warehouse
Every client reads through their own warehouse identity, and row-level isolation is enforced on every table beneath any application code. No query can cross clients, ours included.
Aggregate by design
The methodology runs on spend, revenue, and market-level outcomes. Geo experiments exist precisely so incrementality can be measured without user-level tracking, and every output stays aggregate.
Yours to take
Every published measurement exports: the numbers, the intervals, the evidence tiers. Your measurement history is an asset you keep, whatever happens to the subscription.
Want to see it live?
We'll walk you through the console on your kind of marketing mix, and what it would take to stand the system up for you.