Skip to content
MVP NINJA

MVP Ninja Labs · 05Built in-house

Keel

One screen that tells an SME owner how the business is really doing — and explains anything unusual in plain English.

Category
Internal BI · SME operations
Build shape
14-Day Build
Platforms
Web app, Weekly email digest, Integrations

What Keel does

Keel connects the tools a growing company already uses — e-commerce, accounting, CRM and tasks — and turns them into a weekly picture of revenue, customers, operations and anomalies, with AI explanations instead of chart-reading.

Key decisions

  • What changed comes before any chart.

    An owner has minutes, not hours. The first screen says what moved and why; the charts are one tap away.

  • Twelve metrics, and no report builder.

    A fixed set, computed reliably from the source systems, proves value before anyone needs to build their own reports.

  • Every sentence shows its numbers.

    Each explanation links to the rows and filters behind it, so the owner can check the AI before acting on it.

The numbers exist. They just live in six tools and one person's head.

A 30-person business has revenue in Shopify and Xero, customers in a CRM, work in Asana and support in a helpdesk. The owner pieces together the picture from exports on a Sunday night — and notices problems weeks late.

  1. 01Weekly reporting is a manual export-and-paste job
  2. 02Nobody agrees which number is right
  3. 03Problems — a drop in repeat orders, a late supplier — surface too late
  4. 04Generic BI tools need an analyst the company doesn't have

The SME owner-operator

Founder or MD of a 15–80 person company

Commercially sharp, time-poor, reads numbers but doesn't build spreadsheets for fun.

What they need to get done

  • See the health of the business in two minutes
  • Understand why a number moved, not just that it did
  • Know which operational issues need attention this week
  • Share one version of the truth with the team

The assumption to prove

Owners will act on dashboards if anomalies are explained in plain language with a suggested next step.

Pick a small set of metrics that matter to an SME, compute them reliably from source systems, and lead with 'what changed and why' rather than a wall of charts.

In the first version

  • Connectors for e-commerce, accounting, CRM and tasks
  • Revenue, customers, orders and operations views
  • Anomaly detection with AI explanations
  • Weekly digest email
  • Team access with roles

Deliberately left out

  • Custom report builder
  • Forecasting models
  • Data warehouse exports
  • Embedded analytics for customers

The core journey

  1. Step 01

    Connect tools

    Shopify, Xero, HubSpot and Asana in a few clicks

  2. Step 02

    Confirm metrics

    Keel proposes definitions; owner confirms

  3. Step 03

    Morning view

    What changed since yesterday, in one screen

  4. Step 04

    Investigate

    Open an anomaly and read the explanation

  5. Step 05

    Act

    Assign a task or share with the team

  6. Step 06

    Weekly digest

    Monday email with the week's story

Design exploration

Early layouts copied the classic BI grid of twelve charts. In walkthroughs the eye had nowhere to land. We rebuilt the home screen around three questions: are we healthy, what changed, what needs me.

KeelOverviewRevenueCustomersOperationsAnomalies17 – 13 Sepvs previous weekFHFerngrove Home Goods
Monday 14 September · Good morning, HannahRevenue is up 10% on last week, led by Saturday's garden sale. One thing needs you: repeat customers fell 18%, mostly in Kitchen & Dining.
Summary by Keel AIFrom 12 metrics · numbers computed from 4 sources
Revenue£52.9kUp 10.0%£48.1k last week
Orders1,184Up 6.2%1,115 last week
Repeat customersAnomaly254Down 18.1%310 last week
Overdue tasks7Up 3 more4 last week
Revenue by day£52.9kUp 10.0%vs £48.1k
This week, 7 – 13 SepLast week
£12k£9k£6k£3k
Sat £10.6kUp 31% on last Saturday · garden sale
Mon 7Tue 8Wed 9Thu 10Fri 11Sat 12Sun 13
Orders to ship todayShopify · 07:104331 packed · 4 returns to checkOn track for the 15:00 courier collection
Tasks dueAsana · 07:08Chase Oakmoor PO-22142 days overdueSMReorder linen tea towelsTodayPRApprove September trade price listWedHW
Needs attention4By impact
AnomalyDetected Thu 10 SepRepeat customers fell 18%Below the expected range since Wed. 70% of the drop is Kitchen & Dining.Explained · 3 sourcesRead why
Supplier delivery 4 days lateOakmoor Ceramics · 38 orders · Asana3 invoices over 30 days£4,180 from trade accounts · XeroTea towels: 6 days of stock left126 left, selling 21 a day · Shopify
ShopifyXeroHubSpotAsana07:12
Wireframe · layout and priorities
KeelOverviewRevenueCustomersOperationsAnomalies17 – 13 Sepvs previous weekFHFerngrove Home Goods
Monday 14 September · Good morning, HannahRevenue is up 10% on last week, led by Saturday's garden sale. One thing needs you: repeat customers fell 18%, mostly in Kitchen & Dining.
Summary by Keel AIFrom 12 metrics · numbers computed from 4 sources
Revenue£52.9kUp 10.0%£48.1k last week
Orders1,184Up 6.2%1,115 last week
Repeat customersAnomaly254Down 18.1%310 last week
Overdue tasks7Up 3 more4 last week
Revenue by day£52.9kUp 10.0%vs £48.1k
This week, 7 – 13 SepLast week
£12k£9k£6k£3k
Sat £10.6kUp 31% on last Saturday · garden sale
Mon 7Tue 8Wed 9Thu 10Fri 11Sat 12Sun 13
Orders to ship todayShopify · 07:104331 packed · 4 returns to checkOn track for the 15:00 courier collection
Tasks dueAsana · 07:08Chase Oakmoor PO-22142 days overdueSMReorder linen tea towelsTodayPRApprove September trade price listWedHW
Needs attention4By impact
AnomalyDetected Thu 10 SepRepeat customers fell 18%Below the expected range since Wed. 70% of the drop is Kitchen & Dining.Explained · 3 sourcesRead why
Supplier delivery 4 days lateOakmoor Ceramics · 38 orders · Asana3 invoices over 30 days£4,180 from trade accounts · XeroTea towels: 6 days of stock left126 left, selling 21 a day · Shopify
ShopifyXeroHubSpotAsana07:12
Final · the same screen, designed

  • Question

    Charts first or narrative first?

    Decision

    Narrative first, charts on demand

    A sentence like 'Repeat orders fell 18% — mostly in one product line' is faster to act on than a line chart.

  • Question

    How many metrics?

    Decision

    Twelve, fixed for the MVP

    Choice is the enemy of adoption. A strong default set proves value before customisation.

  • Question

    How to earn trust in AI explanations?

    Decision

    Show the numbers behind every sentence

    Each explanation links to the rows and filters it was drawn from.

Keel’s design system

Quiet blue for data, coral only for anomalies, and a type scale that puts sentences before numbers.

  • Keel Blue

    #1D6FB8

    Primary data series

  • Harbour

    #7FB3E0

    Comparison series

  • Coral

    #E5484D

    Anomalies

  • Kelp

    #2F9E6E

    Positive change

  • Fog

    #F5F7FA

    Background

Typography

Sentence-case headlines for insights; tabular numerals everywhere; small caps labels for metric names.

Components

  • Insight card
  • KPI tile with delta
  • Anomaly timeline
  • Explainer drawer
  • Connector status
  • Digest email

Principles

  1. 01Answer before chart
  2. 02Every claim has a source
  3. 03Twelve metrics, not two hundred

Key screens

KeelOverviewRevenueCustomersOperationsAnomalies17 – 13 Sepvs previous weekFHFerngrove Home Goods
Monday 14 September · Good morning, HannahRevenue is up 10% on last week, led by Saturday's garden sale. One thing needs you: repeat customers fell 18%, mostly in Kitchen & Dining.
Summary by Keel AIFrom 12 metrics · numbers computed from 4 sources
Revenue£52.9kUp 10.0%£48.1k last week
Orders1,184Up 6.2%1,115 last week
Repeat customersAnomaly254Down 18.1%310 last week
Overdue tasks7Up 3 more4 last week
Revenue by day£52.9kUp 10.0%vs £48.1k
This week, 7 – 13 SepLast week
£12k£9k£6k£3k
Sat £10.6kUp 31% on last Saturday · garden sale
Mon 7Tue 8Wed 9Thu 10Fri 11Sat 12Sun 13
Orders to ship todayShopify · 07:104331 packed · 4 returns to checkOn track for the 15:00 courier collection
Tasks dueAsana · 07:08Chase Oakmoor PO-22142 days overdueSMReorder linen tea towelsTodayPRApprove September trade price listWedHW
Needs attention4By impact
AnomalyDetected Thu 10 SepRepeat customers fell 18%Below the expected range since Wed. 70% of the drop is Kitchen & Dining.Explained · 3 sourcesRead why
Supplier delivery 4 days lateOakmoor Ceramics · 38 orders · Asana3 invoices over 30 days£4,180 from trade accounts · XeroTea towels: 6 days of stock left126 left, selling 21 a day · Shopify
ShopifyXeroHubSpotAsana07:12
Screen 01Morning overviewHealth, changes and what needs attention — before the first coffee is finished.
KeelOverviewRevenueCustomersOperationsAnomalies17 – 13 Sepvs previous weekFHFerngrove Home Goods
Anomalies /Repeat customersOpen · detected Thu 10 Sep, 06:00Share with team
Repeat customers fell 18% last week, from 310 to 254. Most of the drop came from Kitchen & Dining buyers who stopped getting the 30-day reorder email on 8 September.Explanation by Keel AINumbers are computed from your data first; the AI only describes them.Sources123Is this right?YesNot quite
Repeat customers per dayExpected range learned from the previous 8 weeks
ActualExpected rangeAnomaly
60504030
From Wed 9
Week of 17 AugWeek of 24 AugWeek of 31 AugWeek of 7 Sep
Last week254Down 18.1%Expected281 – 345Week before310
What changed7 – 13 Sep vs previous week · 56 fewer repeat customersKitchen & Dining: 70% of drop
By product line
Kitchen & Dining−39142 → 103Bedding & Linen−988 → 79Garden−446 → 42Lighting & Decor−434 → 30
By channel
Email link−47118 → 71Direct & search−6124 → 118Social−242 → 40Paid search−126 → 25
Contributing factors3 found · 1 ruled out
  1. 1The 30-day reorder email stopped sending0 sends since 8 Sep, against 212 the week before2
  2. 2Kitchen & Dining relies most on that email58% of its August repeat orders came via email1
  3. 3Email templates were migrated the same day“Move templates to new editor” closed 8 Sep3
Ruled out: prices, stock and delivery times
Data behind this explanationView rows
  1. 1Shopify · Orders5,020 orders, 17 Aug – 13 Sep07:10
  2. 2HubSpot · Email sendsWorkflow “30-day reorder”07:12
  3. 3Asana · Marketing1 matching task07:08
Suggested next steps
Turn the reorder email back onAssign to SamEmail 212 customers who missed itReview list
Screen 02Anomaly explainedWhy repeat orders dropped, with the data behind the sentence.
Screen 03Weekly digestThe week's story, delivered on Monday morning.

How it works underneath

Scheduled syncs pull from source systems into a small analytics store. Metrics are computed in SQL; anomalies are detected statistically and only then explained by an LLM, so the AI never invents the numbers.

    • ExternalShopify, Xero, HubSpot, Asana
    • ServiceScheduled sync jobs
    • DataPostgres analytics schema
    • ServiceAnomaly detection (SQL/stats)
    • AIExplanation writer
    • InterfaceKeel web app
    • ExternalDigest email

Connections

  • Shopify, Xero, HubSpot, AsanatoScheduled sync jobsAPIs
  • Scheduled sync jobstoPostgres analytics schema
  • Postgres analytics schematoAnomaly detection (SQL/stats)
  • Anomaly detection (SQL/stats)toExplanation writerfacts only
  • Postgres analytics schematoKeel web app
  • Explanation writertoKeel web appinsights
  • Explanation writertoDigest email

Likely stack

  • Next.js + TypeScript
  • Postgres
  • Scheduled jobs
  • SQL metric layer
  • LLM with structured facts
  • Transactional email

A 14-day build plan

  1. Days 1–2

    Metric definitions

    Agree twelve metrics and exactly how each is calculated from source data.

  2. Days 2–4

    Narrative dashboard design

    Home screen, insight cards and explainer drawer; test comprehension with a walkthrough.

  3. Days 4–11

    Connectors and metrics

    Syncs for four tools, metric SQL, anomaly rules, AI explanations grounded in facts.

  4. Days 11–13

    Digest and roles

    Weekly email, team invites, connector health.

  5. Day 14

    Launch

    Backfill history, verify numbers against source systems, deploy.

Across devices

Owners check the business from their phone first. The mobile view keeps the narrative and the anomaly list; detailed charts stay on larger screens.

  • Insights stack as cards
  • KPI tiles become a swipeable row
  • Digest email is designed mobile-first

Internal software that behaves like a good analyst: it watches the numbers, notices what matters and explains it plainly.

KeelOverviewRevenueCustomersOperationsAnomalies17 – 13 Sepvs previous weekFHFerngrove Home Goods
Anomalies /Repeat customersOpen · detected Thu 10 Sep, 06:00Share with team
Repeat customers fell 18% last week, from 310 to 254. Most of the drop came from Kitchen & Dining buyers who stopped getting the 30-day reorder email on 8 September.Explanation by Keel AINumbers are computed from your data first; the AI only describes them.Sources123Is this right?YesNot quite
Repeat customers per dayExpected range learned from the previous 8 weeks
ActualExpected rangeAnomaly
60504030
From Wed 9
Week of 17 AugWeek of 24 AugWeek of 31 AugWeek of 7 Sep
Last week254Down 18.1%Expected281 – 345Week before310
What changed7 – 13 Sep vs previous week · 56 fewer repeat customersKitchen & Dining: 70% of drop
By product line
Kitchen & Dining−39142 → 103Bedding & Linen−988 → 79Garden−446 → 42Lighting & Decor−434 → 30
By channel
Email link−47118 → 71Direct & search−6124 → 118Social−242 → 40Paid search−126 → 25
Contributing factors3 found · 1 ruled out
  1. 1The 30-day reorder email stopped sending0 sends since 8 Sep, against 212 the week before2
  2. 2Kitchen & Dining relies most on that email58% of its August repeat orders came via email1
  3. 3Email templates were migrated the same day“Move templates to new editor” closed 8 Sep3
Ruled out: prices, stock and delivery times
Data behind this explanationView rows
  1. 1Shopify · Orders5,020 orders, 17 Aug – 13 Sep07:10
  2. 2HubSpot · Email sendsWorkflow “30-day reorder”07:12
  3. 3Asana · Marketing1 matching task07:08
Suggested next steps
Turn the reorder email back onAssign to SamEmail 212 customers who missed itReview list

Only seeing your numbers at month-end?

Tell us where your data lives today. We’ll scope a dashboard that explains the week in plain English.

Closest package

14-Day Build

$2,700Fixed · 14 days

A real product with several workflows, an admin side and the integrations it needs to work.

  • Up to 3 core workflows
  • Up to 12 screens
  • User + admin roles
Start an internal tool brief

Your idea gets its own written scope and fixed price before anything starts. Compare packages

In the first version

  • Four source connectors
  • Twelve core metrics with definitions
  • Narrative overview and KPI tiles
  • Anomaly detection with AI explanations
  • Weekly digest email
  • Team access and roles

Later, once it's proven

  • Custom metrics
  • Forecasting
  • More connectors
  • Goal tracking

7–21 daysFrom $1,500

Start my brief