Manifesto

Understanding changes behaviour. Dashboards don't.

Every company has dashboards, and most now have AI assistants too. Neither has changed how the decision gets made.Every company has dashboards, and most now have AI assistants too. Neither has changed how the decision gets made. We believe a business runs better when everyone in it, and every agent working for it, can see how it actually works: what drives each number, who owns it, and whether the last action moved it.

Dashboards stop at what happened. The decision gets made somewhere else.

Business intelligence answered the first question a long time ago. The number is on a screen within the hour, governed and correct. The two questions that follow were never its job, so they are still answered the way they always were: in a meeting, from memory, by whoever argues best.

Question one

What happened? Solved.

Two decades of BI made the number cheap. Revenue is down 8%, and everyone in the room can see it.

Question two

Why did it happen? Still a guess.

Is it pricing, churn, or a pipeline problem that started six months ago? A dashboard shows every metric at the same level, with no hierarchy and no direction, so the loudest theory wins and nobody tests it.

Question three

What did we do, and did it work? Never recorded.

Someone acts, or says they did. When the number moves again, nobody can say whether the action caused it. The lesson lives in one person's head, and it leaves when they do.

Your dashboards are honest. The lie is what they leave out.

Marketing reports campaigns accurately. Sales reports pipeline correctly. Operations tracks support volume faithfully. What none of them shows is how these things cause changes in each other: a launch in January fills the pipeline in February, floods support in April and surfaces as churn in June. By the time someone has traced that chain across three departments, the moment to act has passed.

“People change when they see the system, see their place in it, and see what moves when they do.”

What we believe

Stop building dashboards. Start building shared understanding.

Your business is not a set of departments each producing their own metrics. It is one system, and that system has logic that can be mapped, tested and owned. Here is the whole of it on one page: your data at the base, evidence climbing through the middle, and everything your team and your agents do resting on what has been proven.

Data warehouses

The numbers

Semantic layers

The definitions

Business apps over MCP

Launches, decisions, blockers

Org structure

Directory-synced

For humans

Answers, briefings and pushes

Read and write over MCP

Anthropic: 21% → ~95% with curated context

Canopy Agentic Workflows the metrics start them, humans gate them

Canopy Agents four out of the box, any model, your keys

Canopy

The business context layer

Structural context The causal tree, the org chart

Temporal context What was true when, plan against actual

Behavioural context What people did, what worked

Operational context RACI, workflows, approvals

Strategic context Objectives, key results, initiatives

Map & Measure

Metric Trees

Business Models

Prove & Act

Root Cause

Ownership

Verified Impact

Strategy

Grounded in proven causality

Verified proof

Meta spend → Revenue, +£32k vs forecast, q < 0.01

Personalised Action Plans 31 experiments running

hypothesis → a real person’s plan → measure the change

Causation the test cascade

ADF p 0.02 → Granger p 0.003 → q < 0.05, 87 of 1,204 survive

Correlation

22,791 pairs tested nightly, NPS – Churn r 0.71

Data warehouses

The numbers

Semantic layers

The definitions

Business apps

Launches, decisions, blockers

Org structure

Directory-synced

An answer is not an understanding. People act on what they can see for themselves.

Ask an assistant how many orders came in last week and you get a number. You do not get why it moved, what it affects or what to do about it. We still teach children mental arithmetic even though every one of them carries a calculator, because working through the logic builds an instinct no shortcut can replace: you learn to sense when an answer does not add up. A metric tree does that for a business. Trace cause and effect often enough and you stop waiting for the report and start anticipating the problem.

Metric trees

North star

Your ultimate goal

Lag indicators

Past performance

Lead indicators

Predictive metrics

CausalCausalStrong

Revenue

£1.2M

15%

Active Customers

2,847

8.5%

Trial Signups

312

5.2%

Demo Requests

184

3.7%

Cause and effect has to be proven, not asserted.

A line drawn between two boxes is an opinion. So is a paragraph from a language model that noticed two charts moving together. In KPI Tree every relationship between two metrics starts as a hypothesis: surfaced by a nightly sweep, tested over time by our own ML models and a cascade of statistical tests, and pruned by the people who know the business. Then it is confirmed the only way that counts, by acting on it and measuring what happened next. Most candidates fail. That is the point.

Root cause analysis

Verified impact

Meta spend → Revenue+£32k vs forecastq < 0.01

Personalised action plans

hypothesis: spend lifts revenueSarah's plan: spend +£10kmeasure: 4 weeks

Statistical causation

ADF p 0.02Granger p 0.003q < 0.0587 of 1,204 survive

Correlation

NPS ~ Churn · r 0.7122,791 pairs re-tested nightly

Every number needs a name. A metric nobody owns gets reported, not improved.

Thomas Gilbert, widely regarded as the founder of human performance technology, found that the environment people work in (the information, tools and incentives they are given) accounts for roughly three quarters of how they perform. Most companies invest in talent, then starve it of the thing that matters most: a clear line of sight from the work to the outcome. So every metric carries a full RACI. When it moves, the Accountable owner hears first, with the driver that moved it attached, and if nobody acts it escalates up the org chart.

KPI Tree app · 09:14

Revenue is 15% below target. Conversion rate is the primary driver (Granger-causal at lag 3d). @Sarah Chen you are Accountable.

View driversCreate task

Metric ownership

An action is not finished until the metric says so. Impact is verified, never self-reported.

Joe says his project made an impact. Did it? Most tools cannot say, so the claim stands and the lesson is lost. Every action in KPI Tree is tracked against the metric it was meant to move. When the window closes, the outcome is measured against the counterfactual forecast. What held becomes a verified driver. What did not is just as useful, because the next plan starts from what actually worked, not from memory.

Fix checkout flow

Verified · +£32k

Linked to Revenue · measured by the actuals pipeline

Pause underperforming ad sets

Measuring…

Verified impact

The same is true of AI agents. An agent without context hallucinates. A person without context guesses.

Foundation models will keep getting smarter. They will never know your org chart, how your metrics drive each other, or what your team tried last quarter. People and agents need the same thing: a tested model of how the business works. So we give both the same one.

A definition is not context. Agents need the layer above the semantic layer.

Your semantic layer tells an agent how a metric is calculated, and it should stay the source of that truth. It cannot tell the agent what drives the metric, who owns it, what has already been tried or what the plan expects. Pointed straight at a warehouse, a model returns a number and has to work out the why from scratch on every question. Canopy is the business context layer that holds the rest: the same causal model your team works from, open to the assistants you already use.

Canopy, the business context layer

Context is earned, not entered. Fact and opinion never sit side by side.

Most tools that promise a company brain index documents and chat so an AI can search them, and search treats every message as true. Canopy checks each claim against the numbers before it counts. A sales lead says twenty contracts closed, the warehouse says eighteen, so the claim is held back and says so. Every night the claims that made it in are re-read against the results that followed, and what no longer holds is retracted. Every fact keeps its trail.

What Canopy learnt

Keg prices, not footfall, explain the September food cost rise. Three of the four sites moved together in the week the supplier repriced. The sales channel's twenty August contracts are eighteen in the warehouse, so that stays a claim.

Held up

Confirmed by your own completed work

Retracted

Dropped when later results disagreed

Still watching

Noticed, but not claimed yet

Learning, nightly

Drag the timeline to look at any period.

Last 10 days

Keg prices up 8% from 1 Sep, all sites

Event · #ops · holds against Food cost %

Active2 Sep

Closed 20 contracts in August

Note · #sales · warehouse shows 18 closed-won

Quarantined1 Sep

Canary Wharf stays out of the labour cost target

Decision · typed in by James Hart

Active28 Aug

Semantic layer versus business context layer

Automate a decision only once it can be trusted. With a person where a person belongs.

An answer reliable enough to act on every time is one an agent can take the step on, and a workflow can run end to end. So agents work from evidence the platform has already tested, with the permissions of the person who runs them, a named approver where judgement is needed, and an escalation up the org chart when nobody acts. The agent drafts and proposes. The Accountable owner commits.

Trigger

Fired

On target missedMetrics

Completed

Generate Action Plan

Opus 5.1Agents

Waiting

Wait for Sarah's approvalUtilities

Escalate to manager

WorkdayOwnership

Canopy Agentic Workflows

Two disciplines, one product. Data intelligence and human performance.

BI tools have never thought about human behaviour. Performance tools have never had causal data. We call the practice of joining them Data Engagement: connecting data intelligence to behaviour change, and closing the loop from metric to action to verified impact.

Read the guide to Data Engagement

Insight without action is just noise.

See the causal model of your own business, with a name on every number and proof behind every action.

Book a demo⌘B

About KPI Tree

What should we do differently tomorrow? It is the question every business runs on, and the one that started KPI Tree.

Read the story behind it

Experience That Matters

Built by a team that's been in your shoes

Our team brings deep experience from leading Data, Growth and People teams at some of the fastest growing scaleups in Europe through to IPO and beyond. We've faced the same challenges you're facing now.