Analytics patterns
When you have a long list of business questions, it helps to recognise which kind of analytics each one needs. Patterns are reusable approaches that map to types of questions. Using them lets you estimate effort more accurately, reuse technical approaches across use cases, and batch similar work in delivery.
This part defines the six patterns, when to use each, what you need to build them, and then shows how they map to real use cases in two worked examples (B2B SaaS and non-bank lending) — the same structure used in TrueState 360 to turn decisions into a prioritised roadmap.
The six patterns
1. Descriptive / Monitoring — "What is happening?"
The pattern: Define a metric, aggregate it, visualise it.
Dashboards, KPI tracking, and reporting fall into this category. You're not explaining causes or predicting the future; you're giving a clear, timely view of the current state.
| When to use | What you need | Example approach |
|---|---|---|
| Primary need is visibility and alignment on the numbers. | Defined metrics, a data source, a refresh schedule, and a visualisation tool. | Funnel by step; cohort table; time series of a KPI; breakdown by segment or region. |
Examples: Funnel conversion dashboard, portfolio composition report, arrears ageing, monthly revenue by segment.
2. Diagnostic / Root Cause — "Why is it happening?"
The pattern: Segment the data, compare groups, isolate drivers.
Drill-down analysis, segmentation, and cohort comparison help explain why a metric moved or why one segment behaves differently from another.
| When to use | What you need | Example approach |
|---|---|---|
| You need to explain a trend or a gap between segments so that people can act on the cause. | Same as descriptive, plus a clear hypothesis or dimension to segment by (e.g. channel, product, cohort). | Compare segments side by side; cohort retention curves; contribution to change (which segment drove the move?). |
Examples: Why do broker-originated loans default at higher rates? Which document types cause the longest delays? Which channel has the highest CAC and why?
3. Predictive / Forecasting — "What will happen?"
The pattern: Define features, train a model, score or forecast.
Statistical models, machine learning, and scenario analysis produce estimates of future outcomes or likelihoods.
| When to use | What you need | Example approach |
|---|---|---|
| Decisions today depend on what happens next, and you have enough history to train or calibrate. | Historical data with the outcome you want to predict; a definition of the target variable and the horizon (e.g. churn in next 90 days). | Regression or classification (e.g. gradient boosting, logistic regression); time series forecast; survival analysis. |
Examples: Probability-of-default model, demand forecasting, customer churn prediction, lead scoring, hardship propensity.
4. Prescriptive / Optimisation — "What should we do?"
The pattern: Define an objective, set constraints, optimise.
Decision models, optimisation, and recommendation engines tell you what action to take, not just what will happen.
| When to use | What you need | Example approach |
|---|---|---|
| You have a clear objective (e.g. maximise margin, minimise loss) and levers you can control. | A quantified objective; constraints (capacity, rules, risk limits); and data linking actions to outcomes. | Linear or nonlinear optimisation; recommendation engine (e.g. next-best-action); allocation or scheduling solver. |
Examples: Pricing optimisation, collections contact strategy, marketing spend allocation, inventory replenishment, retention offer selection.
5. Detection / Alerting — "What needs attention now?"
The pattern: Establish a baseline, monitor for deviations, alert.
Anomaly detection, threshold monitoring, and automated triggers surface exceptions so that people can act before a small issue becomes a big one.
| When to use | What you need | Example approach |
|---|---|---|
| The cost of missing something is high and you can define "normal" vs "not normal." | A stream or regular snapshot of the metric(s); a definition of "normal" (e.g. historical range, rules) and escalation path. | Threshold rules; statistical control limits; anomaly detection (e.g. deviation from expected pattern); early-warning score. |
Examples: Fraud detection, early warning on borrower stress, covenant breach alerts, sudden drop in conversion.
6. Measurement / Attribution — "Did it work?"
The pattern: Define an intervention, establish a control, measure the difference.
A/B testing, causal inference, and attribution modelling answer whether a change (policy, campaign, process) actually caused the observed outcome.
| When to use | What you need | Example approach |
|---|---|---|
| You need to prove or disprove that a specific change drove results. | Clear definition of the intervention and the outcome; a way to compare treated vs control (randomised or quasi-experimental). | A/B test; difference-in-differences; attribution model (e.g. last-touch, multi-touch, or model-based). |
Examples: Credit policy effectiveness, marketing attribution, broker incentive ROI, impact of a new onboarding flow.
Using patterns in practice
Map each of your business questions to one or more of these patterns. That tells you what kind of analytics capability you need and lets you group similar work (e.g. all "descriptive" dashboards, or all "predictive" models) for delivery. In TrueState 360, we do this in Step 4: every use case gets a stakeholder and a pattern. The two worked examples below show the same idea applied end to end — journey stage, stakeholder, question, pattern, and example algorithm — so you can see how patterns look in context.
Example — B2B SaaS
Strategy context: grow to $500M ARR by moving upmarket and expanding into the US. The tables below show how questions at each stage of the customer journey map to patterns and example algorithms.
| Stage | Stakeholder | Question | Pattern | Example algorithm |
|---|---|---|---|---|
| Marketing | CMO | Which campaigns are driving signups, and what is the cost per trial? | Descriptive | Campaign performance dashboard; cost-per-trial by channel |
| Marketing | CMO | Which campaigns are most likely to bring in accounts that convert and stay? Where should we shift budget? | Predictive + Prescriptive | Logistic regression on trial-to-paid conversion by source; budget allocation model |
| Sales | Head of Sales | What is our win rate by deal size and segment? Where are we losing? | Descriptive + Diagnostic | Win rate dashboard; cohort analysis by deal size and segment |
| Sales | Head of Sales | Which prospects are most likely to close, and which should the team prioritise this week? | Predictive | Lead scoring model (gradient boosting on CRM activity and firmographic features) |
| Signup | Head of Product | Where are people dropping off in the signup flow? | Diagnostic | Funnel drop-off analysis by step and traffic source |
| Signup | Head of Growth | Which visitors are most likely to complete signup if we intervene? | Predictive | Propensity model on visitor behaviour signals |
| Onboarding | Head of Customer Success | What proportion of new accounts complete onboarding, and how long does it take? | Descriptive | Onboarding completion rate and time-to-completion dashboard |
| Onboarding | Head of Customer Success | Which new accounts are struggling and need intervention before they disengage? | Detection / Alerting | Early warning model on onboarding activity; threshold triggers on login and setup completion |
| Activation | Head of Product | What is our time to first value, and how does it vary by segment? | Descriptive + Diagnostic | Time-to-activation metrics; segment comparison |
| Activation | Head of Customer Success | Which users are at risk of never activating, and what nudge is most likely to help? | Predictive + Prescriptive | Classification model on first-week behaviour; recommendation model for intervention type |
| Ongoing usage | Head of Product | Which features are being used and which are being ignored? | Descriptive | Feature usage dashboard; engagement analysis by user and segment |
| Ongoing usage | CPO, Head of Customer Success | Which product investments would most improve retention? Which disengaged users are about to churn? | Diagnostic + Predictive | Feature importance analysis on retention cohorts; churn prediction model |
| Expansion | Head of Sales | Which accounts have grown usage without upgrading their plan? | Descriptive | Usage-vs-plan gap analysis |
| Expansion | Head of Sales | Which accounts are ready to expand, and what is the right offer and timing? | Predictive + Prescriptive | Expansion propensity model on usage signals; offer recommendation engine |
| Renewal | CFO, Head of Customer Success | What is our renewal rate by segment and cohort? | Descriptive | Renewal rate cohort dashboard |
| Renewal | Head of Customer Success | Which accounts are at risk of not renewing, and what should we do about each one? | Predictive + Prescriptive | Churn prediction model; next-best-action recommendation |
| Churn | CEO, Head of Customer Success | Why are customers leaving, and does the reason vary by segment? | Diagnostic | Exit reason analysis; churn driver segmentation |
| Churn | Head of Customer Success | Which active accounts are most likely to churn in the next 90 days? | Predictive | Time-to-churn model; survival analysis; binary classification |
| Win-back | Head of Sales | What is our win-back rate, and which types of former customers return? | Descriptive + Diagnostic | Win-back rate analysis; segment comparison of returning vs non-returning accounts |
| Win-back | Head of Sales | Which churned accounts are worth pursuing, and what offer has the best chance of working? | Predictive + Prescriptive | Win-back propensity model; offer optimisation |
Example — Non-bank lender
Strategy context: grow the loan book by lending to a broader range of borrowers, winning more customers directly, and lowering the cost of funding. The tables below show how questions at each stage map to patterns and example algorithms.
| Stage | Stakeholder | Question | Pattern | Example algorithm |
|---|---|---|---|---|
| Marketing | CMO | Which campaigns are generating applications that go on to settle? What is the cost per settled loan? | Descriptive | Campaign-to-settlement attribution dashboard; cost per settled loan by channel |
| Marketing | CMO | Where should we shift budget to get more of the right applications? | Prescriptive | Channel ROI model; budget allocation optimisation |
| Referral | Head of Distribution | Which brokers are sending the most volume and the best quality applications? | Descriptive + Diagnostic | Broker performance dashboard; quality scoring by settlement rate and default rate |
| Referral | Head of Distribution | Which brokers should we invest more in, and which are costing more than they are worth? | Diagnostic + Prescriptive | Broker value model; commission cost-benefit analysis |
| Application | COO | What is application volume by channel and how is the mix trending? | Descriptive | Application mix dashboard; channel trend analysis |
| Application | CEO, COO | Are there sources of applications we are underinvesting in? | Diagnostic + Prescriptive | Channel opportunity analysis; segment gap analysis |
| Pre-approval | CCO | What proportion proceed to full approval, and where do they fall over? | Descriptive + Diagnostic | Pre-approval conversion funnel; decline reason analysis |
| Pre-approval | CCO | Which applications are likely to fail at approval so we can flag them early? | Predictive | Early approval failure prediction model; classification on application features |
| Document collection | COO | How long is this taking, and where are the delays? | Descriptive + Diagnostic | Time-in-stage analysis; delay driver analysis by document type and borrower segment |
| Document collection | COO | Which applications are going to miss settlement targets unless we intervene? | Predictive + Detection | Time-to-completion prediction; escalation trigger model |
| Approval | CCO | What is our approval rate, and how does credit performance compare to what we expected? | Descriptive + Diagnostic | Approval rate dashboard; predicted vs actual default performance monitoring |
| Approval | CCO, COO | Are there applications in the queue we should prioritise or escalate? | Prescriptive | Application prioritisation scoring; queue optimisation model |
| Settlement and funding | COO | What proportion of settlements are on time, and where do delays come from? | Descriptive + Diagnostic | Settlement time dashboard; delay reason analysis |
| Settlement and funding | COO | Which loans in the pipeline are at risk of not settling on time? | Predictive | Settlement delay prediction model; risk scoring |
| Servicing | CFO | How is the book performing by cohort, product, and channel? | Descriptive | Portfolio performance dashboard; cohort analysis |
| Servicing | CCO, CFO | Which borrowers are showing early signs of stress before they miss a payment? | Detection / Alerting | Early warning model on repayment behaviour; anomaly detection on payment patterns |
| Hardship | CCO | What proportion of hardship arrangements result in the borrower returning to normal repayments? | Descriptive + Diagnostic | Hardship outcome tracking; cohort comparison by arrangement type |
| Hardship | CCO | Which borrowers are at risk of entering hardship in the next three months? | Predictive | Hardship propensity model; early intervention scoring using repayment and behavioural features |
| Arrears | Head of Collections | Which contact approaches are most effective at resolving arrears? | Measurement / Attribution | A/B analysis of contact strategies; intervention effectiveness measurement |
| Arrears | Head of Collections | Which borrowers in arrears are most likely to resolve without escalation, and which need immediate action? | Predictive + Prescriptive | Arrears resolution prediction model; next-best-action for collections team |
| Discharge or refinance | CFO | How many customers are leaving at refinance and what is the typical timing? | Descriptive + Diagnostic | Discharge reason analysis; prepayment speed by cohort and product |
| Discharge or refinance | CFO, Head of Retention | Which borrowers are most likely to refinance away in the next six months, and is there an offer that would retain them? | Predictive + Prescriptive | Prepayment prediction model; retention offer optimisation |
Next steps
Once you've mapped your questions to patterns (and, if you use it, to stakeholders and journey stages as in the examples above), you can score them for business impact and feasibility, then build a roadmap. For that full process — from business context to prioritised use cases and a phased plan — see TrueState 360. The technical essentials behind the patterns (regression, classification, data cleaning, feature engineering) are covered in the earlier parts of this guide.