One clean BigQuery warehouse modelled around DTC reality, plus a private MCP server your team queries from Claude.ai in plain English. No SQL. No analyst queue. No model hallucinations — because the data is modelled before any LLM sees it.
We work with DTC brands doing $1M–50M in revenue who've outgrown spreadsheets and don't trust their current numbers.
One source of truth for revenue across Shopify, Amazon, marketplaces, subs and ERP — joined to product, customer, and channel. End the "the dashboards disagree" tax.
Multi-touch attribution, MMM and incrementality running on your own data — so paid decisions are made on real margin contribution, not platform-reported ROAS.
Cohort LTV, churn risk, demand and inventory forecasts. Plus alerts when something breaks — before your team notices.
Pull data from Shopify, Amazon, Meta, Google, TikTok, Klaviyo, Stripe, your ERP, your fulfilment provider — wherever you sell, market, or operate — into a single BigQuery warehouse you own. Clean, modelled, queryable, daily-refreshed.
Rule-based MTA, Markov chain, data-driven, MMM, geo-holdouts and incrementality testing — all running in your warehouse. Plus a first-party server-side pixel for post-iOS 14.5 identity recovery and conversion send-back to ad platforms.
Predictive LTV, churn prediction, cohort retention, RFM segments your team builds self-serve. Push audiences to Klaviyo, Meta, Google as reverse-ETL out of the warehouse — no separate CDP.
Revenue, demand, inventory, LTV, churn, and full financial forecasting — contribution margin, cash-flow projections, what-if and budget planning — built on the unified P&L that lives in your warehouse.
Creative-level dimensions, scoring (Hook / Watch / Click / Conversion), thumbnails, auto-tagging — joined to actual orders, LTV, and attribution in your warehouse, not in a separate creative-analytics tool.
BI-tool agnostic. Looker Studio, Tableau, Power BI, or whatever your team already uses. Sales, marketing, customer, product, financial, subscription dashboards — plus prompt-based visualisations for non-technical users.
Slack alerts on threshold breaches, outlier and anomaly detection on revenue, refund rate, ad spend, inventory. Issues surface before your team notices.
Custom API integrations, subscription analytics, data migrations, B2B+DTC blends, multi-brand groupings, weird business logic. If it involves your data and doesn't fit a packaged schema, we model it properly.
Your data infrastructure evolves as your business grows. New channels, new metrics, new questions. We maintain pipelines, add new sources, ship new dashboards — on demand, every month.
Every client gets a private MCP server connected to Claude.ai as a custom connector. Your team asks the warehouse questions in plain English and gets the answer in seconds — no SQL, no dashboard opening, no analyst ticket. Same wedge powering AI conversations across DTC right now, except ours doesn't hallucinate because the modelling is done first.
SELECT / WITH queries only. Nothing can be modified, dropped, or written back.
Data preloaded from BigQuery into in-memory DuckDB, refreshed every 30 min – 4 h. Queries return instantly.
Emails, names, addresses, customer numbers SHA-256 hashed before Claude sees them. Geographic, product, and financial fields stay intact.
One Cloud Run instance per client, in your own GCP project, behind Google OAuth and a per-client email allowlist. Your data never leaves your tenancy.
Your BigQuery, your tables, your SQL. If you ever leave us, you keep the warehouse and everything in it. Most packaged tools hold your unified data hostage in their platform.
Looker Studio, Tableau, Power BI — whatever your team already opens every Monday. No proprietary dashboard UI to learn or migrate to.
Every client gets a private MCP server. The warehouse is modelled to be queried by humans and by AI without hallucinations. Packaged tools build AI on their schema; we make your warehouse AI-ready.
Custom COGS tables, brand groupings, B2B+DTC blends, weird business rules — modelled properly in the warehouse, not crammed into a packaged schema. Single-tenant in your own GCP project.
Setup, maintenance, schema changes, new sources, schema evolution — we handle it. You don't staff this internally. One partner accountable for accuracy, freshness, and "why does this number look weird?".
Audit in a week. Build in 2–6 weeks. You see working output in the first sprint because we've shipped this shape for 15+ ecommerce brands and know the patterns.
Their Meta Ads manager and Google Ads were both claiming credit for the same conversions. We built a Markov chain attribution model that revealed 35% of their spend was going to channels that weren't incrementally driving sales. After reallocation, ROAS improved by 28% within two months.
Data lived in Shopify, Meta, Google Ads, Klaviyo, a subscription platform, and three spreadsheets that nobody trusted. We built a centralised BigQuery warehouse with automated pipelines, pulling from 14 sources. Their weekly reporting meeting went from "arguing about which numbers are right" to "deciding what to do next."
Chronic overstock on slow movers, frequent stockouts on bestsellers. We built a demand forecasting model using historical sales, seasonality, and marketing calendar data. Inventory turns improved, cash stopped being tied up in product sitting in a warehouse.
Sound like problems you have?
Book a data triageA live sample of the analytics dashboards we build. Real structure, real queries — not mockups.
Revenue, marketing attribution, cohort analysis, and operational metrics — the kind of single-pane view we build for every client. Interactive, auto-refreshed, built on BigQuery.
Reliable data, low maintenance, fast queries. We pick tools that scale with your business and don't lock you in.
No hourly billing. No scope creep surprises. You know what you're paying before we start.
No. We build the data infrastructure they use. Your media buyer still buys media. Your analyst still analyses. They just finally have numbers they can trust.
Read-only access to your ad platforms, Shopify, email tool, and whatever else we're connecting. We use least-privilege access — no customer PII is copied unless explicitly required and agreed.
Packaged tools give you their dashboard on their schema. We give you a clean, AI-ready warehouse you own, in the BI tool you already use, with natural-language access via MCP and prompt-based visualisations on top. They're not mutually exclusive — Northbeam attribution can pipe into our warehouse, Motion's creative scoring can join to your orders in BigQuery, Triple Whale's pixel can sit alongside our server-side one. We're often the layer underneath.
No anonymized peer benchmarks — we're single-tenant by design, each client's warehouse lives in their own GCP project. No mobile app — Slack alerts cover the daily-mobile use case. No sub-minute real-time streaming — hourly is our floor. If those are deal-breakers, Polar / Glew (benchmarks), Triple Whale (mobile), or Triple Whale's pixel-driven view (live ticker) are stronger fits there.
That's usually where we start. The audit identifies what's broken, and the build phase fixes it. Most clients come to us because they know their numbers are wrong but don't know why.
Yes. We integrate with whatever you're already using. If something needs replacing, we'll tell you — but we don't rip things out for the sake of it.
Usually within a week. Audits take about a week to complete. Build projects take 2–6 weeks depending on complexity. You'll see working output within the first sprint.
Tell us what's broken, we'll tell you what it takes to fix it. No sales pitch, no slide deck — just a straight conversation.
Book a data triage