A metric tree
Your north-star measure broken into the input metrics that drive it, each with a written definition, owner, grain and the exact SQL that computes it.
Service · Data
Commission App Clone Labs to answer the questions your product and operations teams keep circling: where users drop off, which customers stay, and whether the last experiment actually worked. We agree the metrics, check the data behind them, and hand back analysis, definitions and readouts your team can repeat.
Metrics defined before charts are drawn
Tracking audited against real events
Every finding ships with its query
Deliverables
Your north-star measure broken into the input metrics that drive it, each with a written definition, owner, grain and the exact SQL that computes it.
A comparison of the events your product should send with what actually arrives, listing missing, duplicated and misnamed events with fixes for engineering.
Conversion through key journeys and retention by sign-up cohort, segmented by channel, plan or platform, with the points of largest loss identified.
For tests you have run or are about to run: sample size checks, sample ratio validation, effect estimates with intervals and a plain recommendation.
Analysis of fulfilment times, support volumes, refunds or other operational data, pointing to the steps where cost or delay concentrates.
Early decisions
We rank the open questions by the decision each one would change, and decline the ones whose answer would not alter anything.
Whether revenue comes from the billing system or the warehouse, active users from product events or the database, so numbers stop disagreeing.
Whether cohorts are defined by sign-up date, first purchase or activation, and how reactivated users are treated.
Fixed-horizon or sequential testing, the minimum detectable effect worth caring about and the guardrail metrics that can veto a winner.
Your existing warehouse, product analytics tool and notebooks where possible, so the work can be rerun without us.
Engagement
We meet the people who will act on the answers, write each question down with the decision it informs and agree what “good enough” evidence looks like.
Before analysing anything we test the inputs: event volumes over time, duplicate users, timezone handling, bot traffic and gaps after app releases.
Questions are worked in priority order, with interim findings shared early so stakeholders can redirect effort before the final readout.
A written report and live walkthrough per question, plus the queries, notebooks and metric definitions committed to your repository.
Fit
Teams quote different conversion or retention figures in the same meeting, and decisions stall while people argue about which is right.
Pricing, onboarding, a feature sunset or a market launch needs analysis before leadership will commit.
A/B tests are being shipped but results are read informally, and you suspect some winners were noise.
Alternatives
If the goal is self-serve reporting across the organisation with governed metrics, a BI platform implementation is the better scope.
Open registerWhen events, billing and CRM data are scattered and unmodelled, pipelines and a warehouse need to come first.
Open registerForecasting demand, scoring churn risk or estimating causal effects from observational data is data science work.
Open registerAnalysis without a decision attached tends to produce interesting charts and no change. Each question we take on is written alongside the choice it informs, such as whether to keep a free tier, which onboarding step to rebuild, or where to put support staff.
Ask three teams for monthly active users and you will often get three numbers, each defensible. Writing one definition down, including what happens to test accounts, merged users and timezone boundaries, ends a surprising number of arguments.
Those definitions go into a metric tree, so everyone can see how the measure they own contributes to the one leadership watches.
Silent tracking breaks after app releases, bot traffic inflating sign-ups and duplicated users across devices are common. A short health check on inputs saves weeks of confident analysis built on a broken event.
Stopping a test the moment it looks good, ignoring a sample ratio mismatch or reporting only the metric that moved all produce false winners. Our readouts state the pre-agreed success metric, the interval around the effect and any guardrail that moved, so a flat or negative result is reported as clearly as a positive one.
FAQ
Whatever you already run where possible: your warehouse with SQL, product analytics tools such as Amplitude, Mixpanel or GA4, and Python notebooks for heavier statistics. We avoid introducing tools you will not keep.
Yes, and the tracking audit is usually the first deliverable. We identify which events can be trusted, which need fixing, and which questions can be answered reliably today.
Each readout ends with a recommendation and the strength of evidence behind it. Where the data cannot support a confident answer, we say so and suggest how to get better evidence.
Yes. Queries, notebooks and metric definitions are committed to your repository with notes on inputs and assumptions, so the work can be refreshed without us.
Related services
Analytics instrumentation and attribution set up so acquisition data feeds the same metric definitions.
Open registerFunnel drop-offs found in analysis become concrete design changes to test next.
Open registerSubscription products where activation, expansion and churn analysis shape the roadmap.
Open registerHiring
Primary sources
Dated official documentation, standards, and research that support the factual claims on this page.
Reference on exploratory analysis, hypothesis testing and measurement.
Official guide to GA4 event collection and measurement setup.
How to read this page
Scope, cost and timeline figures are planning guidance until confirmed in a written proposal. Diagrams and screens are illustrations, not client deliverables, and source-code rights and handover follow the signed agreement. See the delivery process and the full disclaimer.
Next decision
Define outcomes, constraints, evidence, rights and handover before delivery begins.
Commercial rights, repositories, environments, documentation, acceptance and handover remain contract-defined.