Big Data LDNMissed us at Big Data LDN? Book a demo

We tell

what to do next with confidence

Meet Canopy, the business context layer that makes your AI deterministic.

Business context Claude can't get anywhere else.

Canopy records how your metrics drive each other, who owns them and what actually worked.

Book a demoSee how it works
fct_ordersamountdim_customerssegmentfct_revenueamountdim_skupricefct_websourcedim_adsspendfct_invoicesmrrfct_supportcsatdim_regionsmarket
RevenueSUM(amount)CACspend / new_custsAOVrevenue / ordersMRRSUM(mrr)Churnlost / startLTVARPU × lifeNPSpromoters − detractorsARPUrevenue / usersConversionpurchases / visits
FinanceMarketingProductSarahJamesSofiaLauraRevenueMRRCACLTVChurnDavidEmmaNPSARPUAOV
Claude
Why did revenue fall last month?
KPI Treesearch_metrics

Revenue didn't fall last month. August was your best month of the quarter

August came in at £378,621.99, up 3.9% on July and up 2.8% on August 2025. Two figures on the dashboard do look bad, and both are artefacts.

The −7.2% "month to date" is 1–9 September against 1–9 August, and that window is not like-for-like. August 1–9 contained four weekend days, September 1–9 contained two. Split like-for-like, every day type is up:

Aug 1–9 avgSep 1–9 avg
Weekend days£15,434.33£15,774.14+2.2%
Weekday days£10,557.45£10,677.48+1.1%

The recorded why points at margin, not revenue

KPI Tree's Canopy business context layer returned ten items. None describes a revenue fall. They are almost all cost and margin: Deliveroo moved you to a 32% commission tier from 1 Aug, and JJ Foodservice put beef up 11% on 28 July.

What to do

1.

Rebase the £2.4M key result to a reachable number. Oliver Blackwood owns both the objective and the task.

2.

Decide on the Guest contact data programme. Laura Fitzgerald to either start it or cancel it this week.

3.

Fix the duplicate Revenue metric. The Snowflake one has no RACI at all, and it raises the alarming −7.2%. James Harrington should either assign it or retire it.

Reply to Claude...
Query usingMCP
GeminiOpenAIClaudeMicrosoft CopilotSnowflake CortexDatabricks Genie
0:00

Metric moves. Owner notified. Action tracked. Impact verified.

The whole platform rests on one causal model of your business, built from your warehouse, your semantic layer, your business apps and your org chart. Every claim on it is earned: correlations are re-tested nightly, hypotheses survive a cascade of statistical tests, and action plans prove their impact in the real world. Canopy records the context that explains the numbers, and agentic workflows act on that evidence, so agents move the business, not just describe it.

It works inside the tools your team already uses, and it does two things a dashboard never could. It grounds the AI they already ask, from Copilot to Claude, in business context it can't get anywhere else, so every answer reflects how the business actually works. And it shows your team the whole system and their place in it, so they understand the business far better and act with confidence.

Grounded in proven causality. Every driver tested, every owner named, every action verified.

  1. One causal model, not a folder of dashboards

    Every relationship is a directed driver edge with confidence attached. AI drafts it, your team corrects it, your data decides. It is the ground truth your AI agents inherit.

    Metric trees

    Revenue

    £1.2M15%

    Active Customers

    2,8478.5%

    Trial Signups

    3125.2%

    Support Tickets

    1,0966.4%
    CausalCausalCausalPruned by Emma

  2. Every relationship found for you, not drawn by hand

    The nightly sweep re-tests every pair of metrics you have, whether or not anyone has drawn a relationship between them, and surfaces the ones that move together, ranked by strength. That is where a hypothesis starts, and never where it ends: a candidate waiting to be tested, not a finding.

    How driver signals are found

    Out of Correlation

    Nightly sweep22,791 pairs re-tested
    NPS ~ Churnr 0.71
    Ad spend ~ Trialsr 0.62
    Page speed ~ Checkoutr 0.58

  3. Causality proven statistically, not asserted in a narrative

    Correlation alone is not causation, so every edge is run nightly through our own proprietary ML models and an array of statistical tests, and labelled causal only when it passes them all. Then the people who know the business prune what the statistics cannot see.

    Root cause analysis

    1,204 candidates

    ƒStatistical testsEvery one must pass
    Proprietary ML modelsConfidence it is the cause97%
    87 proven driversof 1,204 candidates

    Back into Causation

  4. A personalised action plan, for every employee

    Each surviving hypothesis becomes a plan for the person who owns the lever, with the driver attached and a window to measure it in. The Accountable owner is notified the moment the metric moves, so everyone sees the few things that fall to them.

    Personalised action plans

    Out of Personalised action plans

    Growth hypothesisA proven driver, off target
    Sarah ChenAccountable · Conversion

    Fix checkout payment errors, worth £31.5k/day

    Measured for four weeks against the forecast

  5. Impact verified, then the model learns

    When the window closes, the outcome is measured against the counterfactual forecast. What held becomes a verified driver edge, the objective it serves updates on the strategy map, and the next plan starts from what actually worked.

    Verified impact

    Four weeks

    Fix checkout flowVerified · +£32k

    Linked to Revenue · measured by the actuals pipeline

    The model learnsA verified driver edge

    Back into Verified proof

See cause and effect across your entire business

Decompose your North Star into every lever your teams control. Every relationship has a direction and a measurable strength. The Five Whys, pre-answered.

0:00
Explore the whole platform

Ask once. Act with confidence. That's deterministic AI.

Why is revenue down at Liverpool Street?

1 What moved

Revenue

£384k3.1%

Liverpool Street

AOV

£7.255.8%

Orders

52,9981.2%

Warehouse, semantic layer, org chart

ClaudeCanopy

Proof, not probability.

On its own, AI predicts the likeliest answer. On Canopy, it answers from proof.

Correlation

9,053,835

relationships tested on average per customer per day

Causation

Re-proven every night.

Last year

Last month

Last night

Causal proof is only the start. Canopy grounds Claude in business context it can't get anywhere else. Yes, even from your semantic layer.

Your semantic layer tells AI how a metric is calculated. Canopy adds what a definition cannot: how the business fits together, what was true when, who owns each number, where the plan is heading, and what people did about it and whether it worked. Every claim is checked against the numbers and carries its provenance.

Canopy is your business context layer. Five kinds of context, each checked against the causal model before it counts. It remembers what moved, who acted and whether it worked.

One story, eight steps

Revenue dipped at Liverpool Street. Follow it from the first sign to the fix.

  1. 1 · What moved, and what drives it

    Structural context · How the business fits together

    Liverpool Street
    CausalModerate

    Revenue

    £384k

    3.1%

    AOV

    £7.25

    5.8%

    Orders

    52,998

    1.2%

    Revenue at Liverpool Street is down 3.1%. The tree shows both of its drivers, AOV and Orders, are down too.

    Both drivers fell. Which one actually caused it?

  2. 2 · Which driver is the cause

    Causal proof · Why AOV, not Orders

    Which driver moved Revenue?

    Liverpool Street · last night

    AOV ↓ 5.8%

    Orders ↓ 1.2%

    Statistical testsEvery one must pass
    Proprietary ML modelsConfidence it is the cause
    97%
    22%
    Verdict
    Causal
    Moderate

    Both correlate with Revenue, but only AOV passes every test, and the models back it. AOV is the driver to chase.

    AOV is the cause. So when did it start?

  3. 3 · When it started

    Temporal context · What was true when

    AOV

    £7.25

    5.8%
    Current periodComparison period
    1. TodayAOV £7.25
    2. 1 week agoAOV £7.41
    3. 2 weeks agoAOV £7.58
    4. 3 weeks agoAOV £7.69Weekend shifts cut
    5. 4 weeks agoAOV £7.66

    Step back through AOV week by week and the slide starts three weeks ago, the week weekend shifts were cut.

    The cause is found. Now, who needs to act, and what did they do?

  4. 4 · Who needs to act

    Operational context · RACI, workflows, approvals

    AOV

    OverviewDataRACITasksContext

    Who owns this metric, who acts on it, and who is kept in the loop.

    Suggest

    Responsible

    David Mitchell+1 more

    David Mitchell

    Area Manager, Operations

    Emma Thompson

    Growth Lead, Marketing

    Laura Fitzgerald

    Head of Sales, Revenue

    Accountable

    Sarah Chen

    Consulted

    Emma Thompson+2 more

    Informed

    Nobody

    CloseSave

    David Mitchell is Responsible for AOV at Liverpool Street, so the fix is routed to him.

    The fix is his. So why were the shifts cut?

  5. 5 · What they are working towards

    Strategic context · Objectives, key results, initiatives

    Reduce labour cost to below 28%

    At riskCompany 1 Jul to 30 Sep 2026Sarah Chen

    The proof

    Numbers that prove the objective.

    Outcome62%

    Key results

    Labour cost ≤ 28%

    29.4% → 28% target

    62%

    Overtime hours below 4%

    awaiting first read

    The work

    What the team does.

    Execution45%

    Initiatives

    Rota optimisation

    On track

    David Mitchell

    70%

    Cross-training

    Planned
    20%

    The shifts were cut under his own Rota optimisation initiative, to bring labour cost below 28%. Any fix has to hold that line.

    The cut served the plan, and cost AOV. So what did he do about it?

  6. 6 · What actually happened

    Behavioural context · What people did, and whether it worked

    Company Snapshot

    • Task
    • Proposed
    • Accepted from Canopy
    • Decision
    • Blocker
    Operations3 people

    Sun

    6

    Mon

    7

    Today

    8

    Wed

    9

    OperationsLiverpool Street

    Sarah Chen

    Head of Operations

    Checkout fix+£32k verified
    Reset the Q3 target

    David Mitchell

    Area Manager

    Backfill shifts
    Supplier price rise

    Emma Thompson

    Growth Lead

    Pause weak ad setsCAC £36.10
    DSO check-in

    Canopy proposed backfilling the weekend shifts only, which keeps labour under target. Once David accepts, it is measured against AOV, like Sarah's verified checkout fix.

    Not everything is in the numbers. They showed when AOV began to slide. Only David knew why.

  7. 7 · Where the fact came from

    From your team and your channels · Typed in, or read from Slack

    New context item

    Context

    Weekend floor staff at Liverpool Street go from six to four from Saturday, under rota optimisation.

    State the fact in plain language. The kind, dates and attachments are worked out automatically.

    Metric (optional)

    Labour cost · Liverpool Street

    Valid until (optional)

    CancelCreate
    SourceLearntWatching

    Slack

    4 channels

    24 items
    ops
    12 items · learnt 2h ago
    revenue
    7 items · learnt yesterday
    sales
    5 items · 1 held as a claim
    random
    Not read

    David typed the cut in the day he made it, against Labour cost at Liverpool Street, and it is kept with his name on it.

    A note, with his name on it. But does it count as a fact?

  8. 8 · Why it counts as a fact

    Learnt and curated · Fact kept apart from opinion

    What Canopy learnt

    Weekend staffing explains the AOV slide at Liverpool Street. Spend per order fell from the first weekend on four staff, while orders barely moved. The sales channel's twenty August contracts are eighteen in the warehouse, so that stays a claim.

    Held up

    14

    Confirmed by your own completed work

    Retracted

    3

    Dropped when later results disagreed

    Still watching

    9

    Noticed, but not claimed yet

    Learning, nightly

    Drag the timeline to look at any period.

    38

    Last 10 days

    Weekend floor staff cut from six to four

    Event · typed in by David Mitchell · holds against Labour hours

    Active3 wks ago

    Closed 20 contracts in August

    Note · #sales · warehouse shows 18 closed-won

    Quarantined6 wks ago

    Canary Wharf stays out of the labour cost target

    Decision · typed in by James Hart

    Active7 wks ago

    Weekend labour hours fell by a third that week, so his note holds against the numbers. Claims they contradict, like the twenty contracts, are held back.

    That is the context behind every answer.

That is the context behind every answer. Here is what it does to one.

We benchmarked Canopy using Opus 5, one warehouse, three ways in: Snowflake over MCP Snowflake Cortex AnalystCanopy with Snowflake Semantic Views

Pointed straight at the warehouse, or at Cortex Analyst's semantic model, Claude returns a number and works out the why from scratch on every question. On Canopy it reads a causal model that is already statistically tested, owned by name, and checked against what happened next. So you get the insight first time, not a number and then a week of follow-ups.

What happened in the business in August that explains the numbers?
CanopyHad some of the context
Claude
The short version

August was a good month. The alarming numbers on the dashboard right now are September-to-date, not August — the health report's "EBITDA -21.8% MoM" is 1–9 Sep vs 1–9 Aug. Read on the full month, August looks like this:

Aug 2026Jul 2026Aug 2025
Revenue£378,621.99£364,535.26(+3.9%)£368,140.79(+2.8%)
EBITDA£111,351.63£99,888.10(+11.5%)£114,655.61(-2.9%)
Gross margin %60.259.768.5
COGS %39.840.331.5
Labour cost %22.123.514.5

The finding: August sold more than last August and earned less. Revenue was up £10,481.20 year on year, yet EBITDA was down £3,303.98. That is not a demand problem — it is entirely a cost-structure problem. COGS is running 8.3 points higher than last August and labour 7.6 points higher.

I'm quoting those as percentage points rather than converting to a cash bridge: the margin metrics aggregate daily rates and don't reconcile cleanly against the P&L lines, so a pounds figure would be spurious.

What Canopy records as the why

KPI Tree's Canopy business context layer explains both halves of that.

Why August beat July:

  • The King's Head kitchen refit finished ahead of schedule, so the venue reopened a full week early and caught the bank holiday trade that had been written off.
  • The Sunday roast menu relaunched across all London venues with sharing boards and a fixed-price family option.
  • At Shoreditch, moving quiz night from Tuesday to Thursday lifted revenue per labour hour — Thursdays were dead and the kitchen was already staffed.

Why the year-on-year margin is gone:

  • Menu re-costing came back showing produce inflation at about 9%, steeper than budgeted, with beef and dairy the worst offenders.
  • The draught supplier raised keg prices 6%, affecting every venue on the same contract.
  • Leadership chose to re-engineer the menu rather than raise prices — protecting the headline price going into a soft quarter, but absorbing the inflation into margin. August is what that decision costs.

Two other recorded items shape the read: the area manager is on leave through the bank holiday fortnight (cover split between two senior venue managers, until 22 Sep), and the new till system is blocked on payment provider approval, so the faster table-ordering flow is still switched off at the pilot sites.

Which venues

All four venues beat July. Against last August, only one is ahead:

  • Canary Wharf £24,832.15 — +3.9% (+£934.15) — the only venue up
  • Liverpool Street £26,658.84 — -3.9% (-£1,085.55)
  • Paddington £26,977.50 — -4.4% (-£1,241.53)
  • Shoreditch £32,883.14 — -5.5% (-£1,911.05)

Canary Wharf being the outlier lines up with the recorded decision to hold its staffing flat through the quarter despite the labour cost target, on the grounds that the lunch trade is too volatile to cut cover. On these numbers that call is paying.

What's already in flight

Both funded initiatives are aimed squarely at this and both are working month-on-month:

  • Menu engineering programme (Rachel Morrison, £45,000 budget, £18,500 spent, 33% progress) targets COGS % down — COGS fell 40.3 → 39.8 July to August.
  • Rota optimisation rollout (Michael Patel, £30,000 budget, £12,000 spent, 33% progress) targets labour cost % down — labour fell 23.5 → 22.1. Canopy confirms it went live at the pilot venue with demand-based templates.

So the levers are moving in the right direction; they just haven't clawed back the year-on-year gap yet. Two departmental objectives sit on this — "Reduce labour cost percentage to below 28%" (Oliver Blackwood) is at risk, and "Improve gross margin to 68% across all venues" (Thomas Williams) is nominally on track at 34% outcome progress but only 16.7% execution progress.

One thing August does not explain. The quarterly EBITDA goal of £950,000 (Thomas Williams and Oliver Blackwood, due 30 Sep) stands at £239,950.11 — that is July £99,888.10 + August £111,351.63 + Sep 1–9 £28,710.38, and it reconciles exactly. With 21 days left and August's best-in-quarter month at £111,351.63, that target cannot be reached. Q3-to-date is 44.3% below the same span of Q2 (£430,770.18), so Q2 ran at roughly £143k/month against Q3's ~£105k. That gap is seasonal or structural and predates August; it is not caused by anything in the August numbers.

What to do
  • Take the keg and produce increases to the P&L, not just the menu. Rachel Morrison owns COGS % and the menu engineering programme. The decision to re-engineer rather than re-price was made before produce inflation came back at 9% — that's a materially different input than when the call was made, and it's Thomas Williams' and Oliver Blackwood's decision to revisit as EBITDA owners.
  • The one check that settles it: whether the draught contract is open to renegotiation or a second supplier. Every venue is on the same contract, so it's a single conversation with a group-wide 6% effect. Nobody in Canopy is recorded as owning that — Rachel Morrison is the right person to open it.
  • Unblock the till rollout. The payment provider approval is holding the faster table-ordering flow off the pilot sites. That is a phone call to the provider, not an analysis, and it's the one recorded blocker with no owner attached.
  • Reset the £950k quarterly target. It is arithmetically unreachable and it is currently making a company objective read as on-track at 29.3%. Thomas Williams and Oliver Blackwood are Responsible.
  • Look again on 1 October, when September closes and gives a clean full-month read on whether the two initiatives keep pulling COGS and labour down.

What would change this conclusion: if the August revenue lift was mostly the King's Head reopening early rather than the Sunday roast relaunch, then the underlying like-for-like trade is flatter than +2.8% and the cost story is worse than it looks. I can't separate those — there's no King's Head metric in the tree; the EBITDA search returned only Canary Wharf, Shoreditch, Liverpool Street and Paddington. That's a measurement gap worth closing.

Two things I can do on your say-so: Canopy has a pointer saying the full autumn menu push context — promo dates and first-wave venues — is in the #marketing channel, and I can pull that thread in. I can also record the "revenue up, EBITDA down year on year, cost-driven" finding against EBITDA so it's there next month. Both need your go-ahead — ingesting reads the whole channel and can adopt the Slack workspace as a context source.

22tool calls0Snowflake queries
Judge: partial, 9 of 10 specifics · 520k tokens of context read

Each surface is judged on whether its answer drew on the business context recorded for this question. The judge saw the answers unlabelled, and saying plainly that the data is not there counts in a surface's favour. The judge's own term and specifics count sit under each answer.

One question, asked verbatim, one run per surface against the same dataset. Answers are unedited, and each panel carries the tool-call and warehouse-query counts of that run.

Context you can act on. Learning from day one. Confidence that compounds.

Every claim in Canopy has been checked against the numbers, so the first answer is the right one and your team can act on it. Then every action is checked against what happened next, so the longer you use it, the more each answer has behind it.

  1. Day1Connect itYour metrics, importedWho: Your data team
  2. Day2Ask itIn Claude or SlackWho: Everyone
  3. Week1Own itOwners on every metricWho: Managers
  4. Week2Switch it onDaily briefingsWho: Each owner
  5. Month1Prove itActions verifiedWho: Runs itself
  6. Month3Automate itAgents and workflowsWho: Process owners
  7. Month6Build on itA track recordWho: The whole team
Day1

Connect itWho: Your data team

Bring the metrics you already have. Canopy starts learning how they relate.

Your metrics come in from your semantic layer, BI tool or warehouse exactly as defined, and stay in sync. KPI Tree only ever reads from your warehouse. Then Canopy tests how every pair of them moves together, so you don't need to know what to measure before you start.

  • dbt
  • Snowflake
  • Databricks
  • BigQuery
  • Apache Ossie
  • Tableau

214 metrics imported

Day 1 · 09:12

Every definition kept exactly as written.

  • dbt Semantic Layer168
  • Snowflake Semantic Views34
  • Tableau, through Apache Ossie12
Learning how they relate22,791 pairs

Every pair tested, then re-tested every night.

Integrations
Day2

Ask itWho: Everyone

No new tool for your team. Canopy works inside the AI they already ask.

Your team asks in Claude, ChatGPT, Copilot or Slack, and every answer respects what the person asking is allowed to see.

  • Claude
  • ChatGPT
  • Gemini
  • Copilot
  • Databricks Genie
  • Snowflake Cortex
  • Slack
In the assistant they already use

Connect Canopy to Claude, ChatGPT or Copilot and nothing changes for your team: they ask as they do today, and the assistant calls Canopy for the answer. It works both ways too: they can act from the assistant, from filing a task to building a tree.

DesktopCLI / IDE

Claude

Connected

Paste the server into Claude. It signs in as you.

{
  "mcpServers": {
    "kpitree": {
      "url": "https://yourcompany.kpitree.co/mcp"
    }
  }
}
Copy configOpen in Claude

ChatGPT

Paste the server into ChatGPT. It signs in as you.

Settings > Connectors > Add
Name: KPI Tree
URL: https://yourcompany.kpitree.co/mcp
Copy config
AlsoGemini, Copilot, Cursor and any MCP client.
Canopy over MCP
Or in Slack

Mention @kpitree in any channel: what is driving the change, chart attached, answered from the same causal model. Nobody's day-to-day changes.

Emma #revenue
@kpitree what's driving the drop this week?

KPI Tree

KPI Tree app · 09:14

Conversion rate, down 23% (proven causal, lag 3d).

The Slack assistant

Then your first tree

Built with Claude, or on the canvas

Most teams start with one tree. Ask Claude to build it from the relationships Canopy found overnight, or draw it on the canvas in the web app. Either way it starts from the statistics, with the strength on every edge: no workshop, no blank canvas.

Metric trees

Sarah in Claude
Build a revenue tree from the drivers Canopy found overnight.

Claude used KPI Tree

Done: Revenue and its two strongest drivers, on your canvas. Shall I copy it for your other sites?

Liverpool Street
CausalModerate

Revenue

£384k

4.2%

AOV

£7.25

5.8%

Orders

52,998

2.1%
Metric trees
Week1

Own itWho: Managers

Nobody wants a goal they can't move. So every owner can see how to move theirs.

Map your team onto the tree, or let the RACI sweep propose an owner for each metric. Nothing is sent to anyone yet, and nobody is held to a number they don't understand.

Who owns it

Owners on the metrics they can move

Map people onto the tree and every owner sees their metric, what drives it and the levers in their hands. The RACI sweep proposes an owner for anything left unowned, with its reasoning.

Metric ownership

Revenue

OverviewDataRACITasksContext

Who owns this metric, who acts on it, and who is kept in the loop.

Suggest

Responsible

David Mitchell+1 more

David Mitchell

Area Manager, Operations

Emma Thompson

Growth Lead, Marketing

Laura Fitzgerald

Head of Sales, Revenue

Accountable

Sarah Chen

Consulted

Emma Thompson+2 more

Informed

Nobody

CloseSave
Metric ownership

How it moves

See how any number moves

In the web app, any change traces to its cause: the path from the outcome to its driver, the evidence on every link and who owns each number. Nobody has to start there, but every owner can.

Root cause analysis
Root cause analysis
Week2

Switch it onWho: Each owner

Nothing goes out until you say so. Then every owner starts the day with a plan.

Once every metric has an owner, each owner switches their briefing on, by Slack or email. It covers only what they own: what moved, what is driving it and the actions worth taking.

KPI Tree

KPI Tree app · 08:00

Good morning, Sarah. Tuesday briefing: revenue is tracking £31.5k/day behind target, down 13.1% MTD. Two things need you.

Conversion is the driver, not traffic. The £31.5k/day gap traces to checkout conversion. Sessions are flat.
Liverpool Street down 40.9% MTD. £26.9k against £45.5k in February, your single biggest drag. You are Accountable.
Orders up 2.1%. On track, so nothing needs you there.
Open briefingView my metrics
Personalised action plans
Month1

Prove itWho: Runs itself

Nobody marks their own homework. The numbers do.

When an action moves its metric the way the driver predicted, the result is verified by the same pipeline as your actuals, and the next plan ranks first what has already worked.

Your first month

  1. Week 1

    Mon 1

    Metrics imported

    Learning how they relate

    Tue 2

    First tree, built in Claude

    From what Canopy found overnight

    Thu 4

    Owners mapped

    The RACI sweep proposed the rest

  2. Week 2

    Daily briefings switched on, 08:00
    Mon 8

    Proposed: fix checkout payment errors

    First item in Sarah's plan

    Tue 9

    Checkout fix shipped

    The task, done

  3. Week 3

    Daily briefings, 08:00
    Measuring the fix against the actuals

  4. Week 4

    Daily briefings, 08:00
    Tue 23

    Verified: +£32k

    The checkout fix, 14 days on

    Wed 24

    Ranked first: scale it to the EU store

    In the next plan, because it worked

Verified impact
Month3

Automate itWho: Process owners

Ready-made agents, or build your own. For any job in the business.

  • Claude
  • OpenAI
  • Gemini
The agents already running

The action plan, the briefings, the RACI sweep and the Slack assistant have been agents all along. Each is a tile in the gallery: what it does, when it runs, how long it takes, and every run it has ever made with its cost.

Personalised Action Plan

Agent

Declining metrics, their drivers, and the actions that fall to each owner.

Weekdays 07:00 ~2 min
Run

Update RACI Assignments

Agent

Finds unowned metrics and proposes owners, with reasoning.

Mondays ~1 min
Run
Run history
Succeeded07:02action_planScheduled12k in / 1.8k out£0.04
Succeeded06:41update_raciScheduled8.3k in / 900 out£0.02
SucceededYesterdayaction_planWorkflow14k in / 2.1k out£0.05
Canopy Agents
Or build your own, for any job

Describe the job in plain English, choose the model it runs on and the tools it may call, then give it a schedule in a workflow. A Monday labour cost check, a supplier price watch, a Friday pipeline sweep: whatever the business needs. Building your own agents is part of the Enterprise plan.

Name

Labour cost watch

Description

Monday check of every site's labour cost against the 28% target

Model

Claude Sonnet 5

System prompt

Every Monday, compare labour cost % for each site against the 28% target. For any site over target, name the Responsible manager and file one task with the gap in pounds and the driver behind it. If every site is under target, say so in one line and stop.

Tools (4)
get_metric_calculationscompare_dimension_metricsfetch_metric_raci_metricscreate_task

Use this agent in a workflow

Run it on a schedule, in response to a metric or task event, or as one step in a longer flow. Send the result to Slack, email, or another step.

Open in Workflows
Building an agent

Then chain them together

Workflows bespoke to your business processes.

A metric trigger, an agent step, a named approver, an escalation when nobody acts.

Revenue target missed, find the driver and act on itPublished
Search nodes

Triggers

On a schedule
Manually run
Inbound webhook
Metric threshold

+6 more

Steps

Run agent
Query metric
Condition

+2 more

Actions

Send Slack
Send email
Create task
Assign RACI

+2 more

Trigger
On target missedRevenue, at period closeMetrics
Root-cause the missOpus 5.1Agents
Wait for Sarah’s approvalAccountable. Escalates in 24hUtilities
Approved
CRM

Re-score stalled opportunities

Finance

Reforecast deferred revenue

Slack

Brief the account owners

Wait 14 daysUtilities
Verify the impactAgainst the counterfactualMetrics
Month6

Build on itWho: The whole team

Every answer has a track record. And every decision builds on the last.

You can act on the first answer, because every claim is checked against the numbers. Six months on, every action has been checked against what happened next too, so decisions stop waiting for the next analysis, or for the loudest voice in the room.

A week, six months in

  1. Every morning
    Weekdays · 08:00

    Daily briefing

    Each owner's plan for the day

    Mondays · 08:00

    Weekly briefing

    The week just gone, reviewed

  2. Monday
    Slack · 09:30

    @kpitree what's driving the drop?

    Conversion, proven causal. David is Responsible.

  3. Tuesday
    Copilot · 10:45

    Why is Liverpool Street down?

    Weekend staffing: spend per order down 5.8%.

  4. Wednesday
    Claude · 12:10

    Accepted: restore Saturday staffing

    Now a task, linked to the finding that raised it.

  5. Thursday
    Workflow · 14:00

    Pipeline below target

    An agent drafted the plan. Laura approved it.

  6. Friday
    Verified · 15:00

    Weekend rota verified: +6.1%

    Spend per order at Shoreditch, 14 days on.

  7. Every night
    Every night, every driver re-tested
Canopy, the business context layer

Enterprise compliance, from day one. Read-only on your warehouse. Governed through your directory.

SOC 2 Type II, continuously monitored

Controls independently examined for Security, Availability and Confidentiality and monitored every day, with CREST penetration testing and the Type II report in the Trust Center.

AICPA SOC 2 badge
Security
Availability
Confidentiality
Trust Center
Continuous Monitoring
Controls Verified
CREST Penetration Testing

We do not store your raw data, calculations run in memory, in real time

KPI Tree reads your warehouse with scoped, read-only credentials and computes every aggregation, comparison and correlation in its own encrypted in-memory engine. Nothing is copied into a lake: only the aggregated daily series each metric produces is kept, encrypted at rest with keys unique to your workspace, or with your own key.

Warehouse connection

Connected
Hostanalytics-prod.••••••.cloud
Service credential••••••••••••
TLS verifiedScoped service credentialsSecrets encrypted

Read-only access. Credentials never leave the encrypted store.

Governed through your directory, for people and agents

SSO with SAML or OpenID Connect, users provisioned and removed from Entra ID, Okta or Google Workspace, audit logs streamed to your SIEM, and IP allowlisting on Enterprise. Agents act only with the permissions of the person who runs them.

Identity
Microsoft Entra IDOktaGoogle Workspace
Audit logs
DatadogSplunk and any SIEM
Network
ZscalerApproved IP ranges only

The company brain for how your business makes decisions.

Understanding changes behaviour. Dashboards don't. Understand and automate every decision, with a name on every number.

Book a demo

Common questions

What is KPI Tree?

KPI Tree is the company brain for how your business makes decisions. It maps your metrics into causal trees that connect every outcome to the drivers beneath it, proves which relationships hold with statistical tests re-run nightly, gives every metric a named owner with a clear RACI, and tracks every action against the metric it was meant to move until the impact is verified. Canopy, its business context layer, makes all of that available to people and to AI agents, and Canopy Agentic Workflows automate the processes that follow.

What is a company brain?

A company brain is a system that remembers how a business makes decisions: what moved, why, who acted, and whether it worked. Most tools that use the name index documents and chat so an AI can search them. KPI Tree's company brain starts from a statistically proven causal model of your governed metrics and layers recorded context on top, and every claim is checked against the numbers before it counts, so fact and opinion never sit side by side.

What is a business context layer?

A business context layer sits between your data and your AI systems and carries the organisational knowledge raw tables cannot. Canopy is KPI Tree's business context layer. It holds five kinds of context: structural (the metric tree and the org chart), temporal (what was true at any date), behavioural (what people did and whether it worked), operational (RACI, workflows and approvals) and strategic (objectives, key results and initiatives). Every item keeps its provenance.

How is a business context layer different from a semantic layer?

A semantic layer defines how a metric is calculated and returns a governed number. It gives AI the right number. It does not say whether that number is good. A business context layer adds what a definition cannot: what drives the metric and how strongly, who owns it, what has already been tried and what the plan expects. Canopy does not replace your semantic layer. Keep dbt, Looker or Snowflake Semantic Views as the source of calculation truth, and Canopy syncs those governed definitions and builds the causal, ownership and outcome context above them.

Does Canopy make AI deterministic?

Deterministic AI, sometimes called causal AI, answers from relationships proven to hold. Generative AI on its own predicts the likeliest answer, working out the numbers, the drivers and the owners afresh on every question, so the same question can come back with a different answer, or an invented one, and someone has to check it. Canopy makes the facts behind every answer deterministic: metric values come from your governed definitions, driver strengths from statistical tests re-run nightly, owners from the RACI, and recorded context only after it has been checked against the numbers. The model's wording can vary, but the numbers, the drivers and the named owner behind the answer do not.

Which AI assistants and agents can use Canopy?

Anything that speaks MCP. One command adds Canopy to Claude Code, Codex or Cursor, and it connects to the Claude and ChatGPT desktop apps, Gemini, Microsoft Copilot, Databricks Genie and Snowflake Cortex. In Slack, mention @kpitree in any channel and the assistant answers from the same causal model. Every answer is scoped to the asking user's permissions.

Does our team need to learn another tool?

No. Connect Canopy to the AI assistant they already use, such as Copilot, Claude or ChatGPT, and they ask as they do today: the assistant calls Canopy for the answer. It works both ways too: they can act from the assistant, from assigning an action to filing a task. Or they mention @kpitree in Slack. The KPI Tree web app is there when someone wants to see how an answer was reached.

What does the first month look like?

On day one your data team imports your metrics from your semantic layer, BI tool or warehouse, and Canopy starts learning how they relate. On day two your team can ask in Claude, ChatGPT, Copilot or Slack, and builds its first tree with Claude or on the canvas: most teams start with one. In week one managers map owners onto it, with the RACI sweep proposing the rest, and nothing is sent to anyone yet. In week two each owner switches their daily briefing on, and by the end of the first month the first actions have been checked against what happened next. Agents and workflows usually follow by the third month.

Does KPI Tree prove causality?

Yes. Every pair of metrics is re-tested for correlation nightly, and every candidate driver runs through our own proprietary ML models and an array of statistical tests, not LLM pattern matching. The people who know the business prune what the statistics cannot see, and completed actions verify the surviving driver relationships against real outcomes.

What if we don't know who should own each metric?

That is where most accountability efforts stall, and why KPI Tree starts from the statistics. The nightly run shows what drives what across your business, so the first version of your metric tree is drawn for you. You then map people onto the metrics they can move, and each owner sees what drives their number and the levers in their hands, which is what makes people willing to own it.

How does Canopy learn, and how does it keep fact apart from opinion?

Canopy reads the Slack channels you allow, one channel at a time, and takes facts your team types in against a metric with an expiry. Nothing counts as a fact until it has been checked against the numbers. A claim the warehouse contradicts is held back and says so, and anything later results disagree with is retracted. Every night the model re-reads its own claims against the results that came in, and the ledger shows what held, what was retracted and what is still being watched.

Can we run our own AI agents in KPI Tree?

Yes. KPI Tree ships ready-made agents, including the personalised action plan, the daily and weekly briefings, RACI assignment and the Slack assistant. On the Enterprise plan you can also build your own: describe the job in plain English, choose the model it runs on from Claude, OpenAI or Gemini, choose the tools it may call, and give it a schedule in a workflow. Agents act only with the permissions of the person who runs them, under spend caps, and every run keeps its history and cost.

What can a Canopy Agentic Workflow automate?

Any business process that starts from a signal in the business. Ten triggers are available: a target missed at period close, a metric threshold, a metric outlier, a metric gone stale, a task event, a RACI assignment, a company briefing being ready, a schedule, an inbound webhook, or a manual run. Steps run agents, query metrics, branch, delay and wait for a named approver, and actions post to Slack, send email, create tasks, assign RACI, escalate up your org chart when nobody acts, and call any external system.

How is KPI Tree different from a BI tool like Tableau or Looker?

Tools like Tableau and Looker are built to visualise data on dashboards and charts, leaving you to interpret what a number means. KPI Tree is built to drive decisions. It connects each metric to its causal drivers, its named owner and the actions meant to move it, then verifies whether those actions worked. A dashboard tells you a number went down. KPI Tree tells you why it moved, who owns it, and what to do next.

Is KPI Tree SOC 2 compliant, and where does our data live?

Yes. The KPI Tree platform is SOC 2 Type II, independently examined for Security, Availability and Confidentiality and monitored continuously, with CREST penetration testing and the Type II report available in the Trust Center on Growth plans and above. Your data stays in your warehouse: KPI Tree connects with scoped, read-only credentials and runs one scheduled query per metric. Anything cached is encrypted at rest with keys unique to your workspace, or with your own key.