Turn business data into a modelling problem
Define the target, time horizon, entity or analysis objective and let the platform handle supported preprocessing steps.
AutoML · ready-to-use models
No-code machine learning models built around real business questions.
Graphite Note AutoML automates repetitive model-building work and provides ready-to-use analytical templates for forecasting, classification, regression, segmentation, customer analytics, portfolio analysis and item basket analysis. The goal is not simply to train a model. It is to reach a reliable business answer faster and connect that answer to a decision.
From dataset to usable model
Automate the modelling work. Keep the business decision in view.
AutoML is useful when a business has a clear outcome to predict or pattern to understand but does not want to repeat the same technical workflow manually for every dataset, segment, store, product or operating unit.
Graphite Note supports model health checks, performance metrics and model comparison so users can judge whether a result is ready for use rather than treating model creation as a black box.
Test candidate algorithms consistently
Reduce manual trial-and-error by comparing suitable modelling approaches and performance metrics.
Review model health before acting
Inspect diagnostics, data quality, model behaviour and potential dataset improvements before operational use.
Understand the drivers
Move beyond a prediction by exposing feature importance, model context and business-readable interpretation.
Connect model output to a decision
Use forecasts, scores, segments and affinities inside targeting, planning, intervention and optimisation workflows.
Graphite Note models
What machine learning models does Graphite Note support?
Graphite Note includes ready-to-use model types for common predictive and advanced analytics problems. Each model starts from the business question, not from an algorithm menu.
Choose a model to see the decision it can support.
Time Series Forecasting
Know what demand will look like before you have to act.
Forecast future values from historical time-based patterns, seasonality and supported business drivers.
Typical business uses
Revenue forecasting, demand planning, inventory requirements, customer volumes and promotion-aware forecasts.
EXAMPLE DECISION
How much stock should each store carry next month?
Signal
Historical sales by store or SKU, seasonality, holidays, promotions and supported external business drivers.
Model output
A time-based forecast at the operating level, supported by forecast-error metrics, diagnostics and model comparison.
Business move
Adjust replenishment, staffing, purchasing or commercial plans before the expected demand change reaches the P&L.
Binary Classification
Turn a yes-or-no business risk into a prioritised action list.
Predict one of two outcomes for each row, customer, transaction or asset.
Typical business uses
Churn risk, conversion propensity, campaign response, fraud flags and equipment failure risk.
EXAMPLE DECISION
Which high-value customers are most likely to churn?
Signal
Customer behaviour, tenure, usage, purchases, service interactions and other variables known before the outcome.
Model output
A probability or class for each customer, transaction or asset, with performance metrics and explanatory drivers.
Business move
Prioritise the cases where the predicted risk or opportunity is large enough to justify an intervention.
Multiclass Classification
Route each case to the most likely category instead of forcing a binary choice.
Predict which of several possible classes an observation belongs to.
Typical business uses
Customer category, product class, next-best content type, service tier or issue classification.
EXAMPLE DECISION
Which service path should this customer enter?
Signal
Attributes and historical examples where the final category, issue type, product class or service route is known.
Model output
A predicted class across multiple possible outcomes, with confidence or probability by class where supported.
Business move
Route, personalise or prioritise the case according to the predicted category and the cost of being wrong.
Regression
Estimate the number the business needs, not only the class it belongs to.
Predict a continuous numeric outcome and expose the variables most associated with that prediction.
Typical business uses
Sales value, demand, price, order value, energy usage and website traffic.
EXAMPLE DECISION
What order value should we expect from this customer?
Signal
Historical observations with a known numeric outcome plus customer, product, channel, operational or market variables.
Model output
A predicted continuous value with model-performance metrics and explanatory feature information.
Business move
Use the estimate to allocate budget, capacity, inventory, sales effort or thresholds where expected value matters.
Customer Lifetime Value
Know which customers are worth protecting before they disappear.
Estimate future customer value and purchasing behaviour using repeat-purchase patterns.
Typical business uses
Prioritise valuable customers, retention spend, loyalty actions and customer acquisition economics.
EXAMPLE DECISION
Which customers deserve retention spend this quarter?
Signal
Purchase history, transaction frequency, recency, spend and repeat-purchase behaviour over time.
Model output
An expected future customer value or purchasing profile that can be compared across the customer base.
Business move
Prioritise retention, loyalty and acquisition spend around customers whose expected future value justifies the cost.
General Segmentation
Find the groups already hiding in the data before deciding how to treat them.
Discover natural groups in customer, product or operational data without predefined labels.
Typical business uses
Behavioural segments, product groups, operating profiles, anomaly discovery and targeted strategies.
EXAMPLE DECISION
Which naturally different customer groups need different treatment?
Signal
Customer, product or operational features selected to represent meaningful differences in behaviour or profile.
Model output
Distinct clusters with segment membership and descriptive characteristics that explain how the groups differ.
Business move
Design targeting, service, assortment or operating strategies around groups that behave differently in practice.
RFM Customer Segmentation
Separate champions, loyal buyers and at-risk customers with a transparent commercial framework.
Segment customers using Recency, Frequency and Monetary value.
Typical business uses
Identify champions, at-risk customers, frequent buyers, high spenders and reactivation audiences.
EXAMPLE DECISION
Which dormant high-value customers should we reactivate first?
Signal
How recently each customer purchased, how frequently they buy and how much monetary value they generated.
Model output
RFM scores and customer groups that distinguish high-value, active, emerging and at-risk customer behaviour.
Business move
Match reactivation, loyalty, suppression and upsell treatment to the commercial value and recency of each group.
ABC / Pareto Analysis
Concentrate attention where a small part of the portfolio drives most of the outcome.
Classify products or entities into A, B and C groups based on their contribution to a chosen business metric.
Typical business uses
Inventory prioritisation, assortment management, supplier focus and resource allocation.
EXAMPLE DECISION
Which SKUs deserve the tightest inventory attention?
Signal
Product, customer, supplier or other entity-level contribution to revenue, margin, volume, demand or another chosen measure.
Model output
A, B and C groups showing which entities contribute most, moderately or least to the selected business outcome.
Business move
Apply tighter service levels, replenishment rules, supplier focus or management attention to the highest-contribution group.
Customer Cohort Analysis
See whether customers acquired at different times or through different channels behave differently later.
Group customers by a shared starting point, commonly first purchase, and compare behaviour over time.
Typical business uses
Retention curves, revenue by acquisition cohort, channel quality and repeat-purchase behaviour.
EXAMPLE DECISION
Which acquisition channel produces customers who keep buying?
Signal
Customer start date or first purchase, acquisition source and subsequent transactions, retention or revenue over time.
Model output
Comparable cohort curves and measures showing how retention, repeat purchase or revenue develops after acquisition.
Business move
Shift acquisition budget and onboarding effort toward cohorts that create stronger downstream customer economics.
New vs Returning Customers
Know whether growth is coming from acquisition or from customers choosing to return.
Separate first-time and repeat customers over time to understand acquisition and retention dynamics.
Typical business uses
New-customer growth, returning-customer share, loyalty trends and retention effectiveness.
EXAMPLE DECISION
Is this month’s growth driven by new customers or repeat behaviour?
Signal
Customer identifiers and transaction history sufficient to distinguish a first purchase from subsequent purchases.
Model output
New-versus-returning customer counts, revenue, shares and trends over time.
Business move
Change acquisition, CRM or loyalty investment depending on whether the business needs more new customers or stronger repeat behaviour.
Item Basket Analysis
Find what customers buy together, then turn affinity into a commercial action.
Find products that are bought together and quantify useful product affinities using basket-level transaction data.
Typical business uses
Often-bought-together recommendations, bundle design, cross-sell placement, merchandising and promotion combinations.
EXAMPLE DECISION
What should we recommend with a coffee machine?
Signal
Basket-level transactions showing which products occur together in real customer orders.
Model output
Product affinities ranked with measures such as support, confidence and lift rather than raw co-occurrence alone.
Business move
Test an often-bought-together bundle, recommendation or merchandising change and measure attach rate and average basket value.
AutoML vs Decision Intelligence
AutoML builds models. Decision Intelligence decides what to do with them.
Read the full Decision Intelligence vs AutoML comparison
Model workflow
- Predictive model development
- Model comparison and selection
- Performance evaluation and diagnostics
- Prediction, segmentation and affinity analysis
Business decision workflow
- Prediction plus business context
- Causal analysis where relevant
- Intervention and constraint logic
- Recommended next action and expected impact
Direct answers
AutoML questions buyers and AI search systems ask.
What machine learning models are available in Graphite Note?
Graphite Note supports time series forecasting, binary classification, multiclass classification, regression, customer lifetime value, general segmentation, RFM customer segmentation, ABC or Pareto analysis, customer cohort analysis, new-versus-returning customer analysis and item basket analysis.
What is Item Basket Analysis?
Item Basket Analysis, also known as market basket analysis, identifies products that are frequently purchased together and measures the strength of those relationships. Retail and eCommerce teams can use the result for bundles, recommendations, merchandising and cross-sell decisions.
Does AutoML choose the best model automatically?
Graphite Note automates supported parts of model preparation, training, comparison and selection, while still exposing performance metrics and model-health information so a result can be checked before use.
Do we need a data science team to use Graphite Note?
Not for a normal managed engagement. Graphite Note handles model development, validation and interpretation while the client team supplies business context and owns execution.
Is Graphite Note only an AutoML product?
No. AutoML is part of the stack. Graphite Note also supports causal analysis, decision logic, optimisation and managed delivery around the business outcome.
Decision IntelligenceFrom signal and prediction to a governed business action.AutoMLAutomate repetitive model-building work without losing business context.Causal Machine LearningSeparate drivers from correlations and estimate intervention effects.Predictive AnalyticsForecast demand, score customers and anticipate business outcomes.Decision OptimisationCompare feasible actions and rank them by expected business impact.Decision APIDeliver recommendations, expected impact and evidence into workflows.Embedded Decision IntelligenceAdd a decision layer to ERP, CRM, BI, workflow and vertical software.Build vs BuyA practical framework for deciding whether to build, buy or embed.
Start with the model. Finish with the decision.
Use Sandbox to explore predictive models, or bring us a real business outcome for a managed decision-intelligence engagement.