Signals —
Core —
Scroll for the detail.
Business intelligence · Data modelling · Analytics
Data. Forecast.
Decision.
Keo Collective builds purpose-built software that captures operational data at the point of work, models it into probabilities and forecasts with Keo, our own purpose-trained AI, and delivers the result to the person who makes the decision. Professional services, engineering and manufacturing.
01 — Purpose
Why Keo Collective exists
Every product and every engagement is one step in the same chain. Nothing is built that does not serve it.
STEP 01
Trustworthy data
Purpose-built SaaS
Data captured once, at the point of work, by software designed for that work. No re-keying, no spreadsheets, no reconstruction after the fact. The record is the operation.
STEP 02
A probability or a forecast
Modelling and analytics
Keo, our own AI, models the captured data — descriptive measures first, then predictive: the likelihood of each outcome, the expected value, the range around it and the confidence in it.
STEP 03
Someone who can act
Decision delivery
The forecast is placed in the tool, the screen and the moment where the decision is made — by a person with the authority to change what happens next.
Analysis that stops at a report has not finished. The work is complete when a decision has been made on it.
02 — Keo
Keo, the intelligence behind the team
Keo is the name of the group and the name of the model. Keep Everything Organised.
Keo is our own AI — trained specifically on the data behind our products and the decisions made on it, and still learning from every dataset it works on.
Keo does the modelling. Our analysts check it. Your people make the call.
- Keo
- Models the data. Writes the reasoning.
- Our team
- Builds the products. Verifies the output.
- Your people
- Decide, and act.
03 — Practice
Three layers of analytics
All three run on the same governed data model. Each layer answers a different question.
Business Intelligence
Descriptive
What happened. How.
Reporting, dashboards and the measures beneath them. The business as it is recorded — by function, period, customer, job, machine, campaign.
- Governed KPI definitions
- Operational and financial reporting
- Dimensional data modelling
- Variance to target and to plan
- Cohort and segment analysis
Business Analytics
Predictive
What is likely. How likely.
Statistical and machine-learning models applied to the same data to weigh each outcome — demand, throughput, conversion, margin, risk — as a probability with a range.
- Time-series forecasting
- Regression and classification
- Probability and confidence intervals
- Anomaly and drift detection
- Survival and churn modelling
Decision Analytics
Prescriptive
What to do. What it costs.
The forecast turned into a choice: scenarios compared, constraints applied, the expected value of each option stated so the decision can be made and defended.
- Scenario and sensitivity analysis
- Monte Carlo simulation
- Capacity and schedule optimisation
- Resource allocation
- Threshold and alert design
04 — Methods
Modelling and analytical methods
Chosen per problem. The method is selected for the decision it has to support, not the other way round.
Dimensional modelling
Facts and dimensions structured so every metric has one definition and one source.
Metrics layer
KPIs defined once, in code, and reused by every report and model.
Time-series forecasting
Trend, seasonality and cycle decomposed; ARIMA, exponential smoothing, gradient boosting.
Regression
Linear, logistic and regularised models for drivers, elasticities and expected values.
Classification
Probability that a lead converts, a job runs late, a part fails inspection.
Probabilistic forecasting
Prediction intervals and quantiles, not point estimates alone.
Monte Carlo simulation
Thousands of runs across uncertain inputs to produce a distribution of outcomes.
Scenario analysis
Base, upside and downside cases with the assumptions that separate them.
Anomaly detection
Statistical control limits and model residuals to flag what does not fit.
Optimisation
Linear and constraint programming for schedules, capacity and allocation.
Survival analysis
Time-to-event models for churn, equipment life and job duration.
Cohort analysis
Behaviour tracked by the period a customer, job or campaign started.
Statistical process control
Control charts on cycle time, yield and throughput.
Attribution modelling
Marketing spend and channels weighed against pipeline and revenue.
Data quality scoring
Completeness, consistency and timeliness measured before anything is modelled.
Model monitoring
Forecast accuracy tracked against actuals; models retrained when drift appears.
05 — Products
Purpose-built software
Each product captures trustworthy data for one kind of work and carries its own intelligence layer. All designed, engineered, hosted and supported in-house.
06 — Data
The data foundation
A forecast is only as good as the data under it. Keo Collective owns the capture, the storage and the model, so the chain from record to decision is unbroken.
Every product writes to a governed data model with defined entities, keys and metrics. The same definition of throughput, margin or conversion is used in the dashboard, the forecast and the alert.
Data quality is measured, not assumed. Completeness, consistency and timeliness are scored on every dataset before it is modelled, and model accuracy is tracked against actuals after it is deployed.
- Capture
- Purpose-built applications at the point of work
- Storage
- Governed relational and key-value stores, one model per domain
- Metrics
- Defined once, versioned, reused everywhere
- Modelling
- Keo — descriptive, predictive and prescriptive on the same dataset
- Delivery
- Inside the application, on the screen where the decision is made
- Monitoring
- Forecast accuracy and data quality tracked continuously
07 — Contact
Start with the decision
Select the product or practice area it concerns. Enquiries are answered within one business day.