Quant-grade data & analytics consultancy
We build quant-grade forecasting, real-time analytics, and regulated reporting for teams where the numbers have to be right, from hedge funds to commerce brands. An on-demand team of senior specialists, with 30+ years of combined experience on a modern data stack.
Anyone can plug in a connector. We do the modelling, the real-time engineering, and the reconciliation that most teams cannot.
Forecasting and statistical models that hold up out of sample: demand, price and margin, risk and exposure. Robust statistics, time-series, and walk-forward validation, so the numbers are defensible, not curve-fit to the past.
Low-latency pipelines and a modeled layer on Snowflake or BigQuery with dbt, feeding dashboards that update as the data moves. Built for teams where minutes, or milliseconds, decide the outcome.
Audit-grade, reconciled statements and filings for finance-critical and regulated reporting: US financial reporting and SOX-style controls, hedge-fund regulatory reporting, and producer-responsibility (RPRA) programs. Every figure ties to the source, to the cent.
True contribution margin, blended CAC, MER, and LTV:CAC across your storefront, ad platforms, and email, plus marketing-mix modeling to guide spend. One trusted definition of profit by product and channel.
Anonymized to respect client confidentiality. The figures are real, aggregated across the work.
Context. A multi-brand e-commerce operator ran a near-identical stack for every brand, each with its own storefront, Meta and Google ads, and email. Standing up analytics for a new brand meant rebuilding 100+ dashboard components by hand, and no two brands defined margin or CAC the same way.
What we built. A modeled BigQuery data layer per brand, a Python engine that clones and remaps a reference dashboard to any new brand automatically, and one shared metric layer for the numbers that drive spend: contribution margin, blended CAC, MER, ROAS, and LTV:CAC, plus marketing-mix modeling.
Result. Across the portfolio the platform models nearly $480M in revenue across 4M+ orders and 2M+ customers, and unifies $17M+ in ad spend into one trusted source. Onboarding a new brand went from days of manual work to an automated clone, and every brand now reports margin and CAC the same way.
2023, the starting point. The label was doing about $38.9M a year on roughly 75K orders, split across separate domestic and international Shopify storefronts, each with its own Meta and Google ads and email. The data lived in silos, with no single view of customers, cohorts, or which channel and product actually drove profitable orders.
What we did. We unified both storefronts and every ad platform into one modeled data layer, joining orders, customers, and line items into cohort, CLV, contribution-margin, and blended-CAC views by product and channel, plus demand forecasting to guide buying and spend.
2024, momentum. Revenue rose 43% to $55.5M as orders jumped from 75K to 120K and new-customer acquisition climbed to 71K for the year.
2025, compounding. Revenue reached $62.4M, the international storefront doubled from $6M to $13M, and orders nearly doubled from the 2023 base to 135K. The active base grew to 89K customers running a $577 lifetime value and a 28% repeat rate, all finally visible in one place.
2022, the starting point. The retailer was doing about $2.6M a year on roughly 13K orders, with storefront, ad, and email data sitting in separate silos and no clear read on which channels and products actually paid back.
What we did. We unified the storefront, Meta and Google ads, and email into one modeled data layer, with cohort, CLV, contribution-margin, and blended-CAC views by product and channel, plus demand forecasting to guide buying.
2023 to 2024, acceleration. Revenue climbed to $3.6M, then $5.1M, as order volume nearly tripled and acquisition scaled profitably.
2025, more than doubled. Revenue reached $5.7M, up 119% from the 2022 base, on 42.6K orders (3.4x) across 71K customers at a $347 lifetime value and a 32% repeat rate.
A sample of the brands and programs behind the platform. Real figures, anonymized.
Stood up analytics from scratch and scaled from near zero to eight figures at a premium price point.
Grew from under $0.5M to $1.4M a year as ads, email, and storefront came together in one margin view.
Revenue climbed 40% while average order value rose 35% as merchandising and channel mix got clearer.
Stabilized revenue and lifted average order value 31%, shifting the mix toward more profitable products.
Unified a very high-volume storefront into one trusted model of customers, orders, and repeat behavior.
Consolidated years of orders, customers, and returns into a single source for merchandising and retention.
Real-time positions, exposure, and P&L from a low-latency pipeline, plus regulatory reporting reconciled to the book.
Automated regulatory statements that tie to the source and to BI, plus capacity-aware volume forecasting.
Robust statistics, eigenvector weighting, and time-decay forecasting on messy real-world signals, the toolkit behind the client work.
The stack top data teams use, plus the quant and real-time engineering that sets the work apart.
A simple path from messy data to forecasts, real-time views, and reports you can run the business on.
We map your data and the decisions it should drive, then find where numbers disagree, where money leaks, and what a clean model would unlock.
A tested, documented data layer plus the models, real-time views, or reports you need, validated so the numbers hold up out of sample and reconcile to the source.
Models retrain and reports refresh on a schedule, so decisions stay current without anyone pulling spreadsheets. Hand-off or ongoing, your call.
Dedolytics is a boutique, on-demand data team. For every engagement we assemble the senior specialists it actually needs, so you get exactly the right expertise without the overhead of a large agency or the key-person risk of a lone freelancer. A typical project team brings 30+ years of combined experience building production data, ML, and analytics systems across finance, retail, e-commerce, and regulated reporting, on a modern stack: Python, SQL, dbt, Snowflake, BigQuery, and Power BI.
To pressure-test the forecasting and robust-statistics toolkit behind our client work, we designed and built a production analytics system from scratch: roughly 2,100 lines of Python, a real-time event pipeline, and interactive dashboards. The point was not the domain. It was proving the pipeline holds up on messy, real-world signals.
This is the same toolkit our client work runs on. Forecasting, handling messy data robustly, and weighting signals by their real importance are the exact methods behind the engagements above. The domain changes. The math does not.
An in-house research and engineering build that demonstrates capability, not a product Dedolytics sells. Read the technical write-up.
Tell us what decision you are trying to make with more confidence. If we can help, we will say how. If we are not the right fit, we will say that too. We work with US teams remotely, with business-hours overlap.
Or reach out directly
hello@dedolytics.org