Predictive Analytics

Analytics

Also: Predictive modelling · Forecasting analytics · Predictive intelligence

What it isUsing historical data to forecast future outcomes

How it worksStatistical models trained on past behaviour

Marketing useChurn prediction, lead scoring, LTV forecasting

Quick definition

Predictive analytics uses statistical models and machine learning to forecast future outcomes based on historical data. In marketing, it is used to predict which leads will convert, which customers are likely to churn, and what a customer's lifetime value will be.

Where it shows up in the data

Lead scoring

Assigning a probability score to leads based on their characteristics and behaviour. Leads that match your ideal customer profile and show high engagement get higher scores and are prioritised for sales follow-up.

Churn prediction

Models that identify customers who are likely to cancel or stop buying before they do. Common signals include reduced login frequency, decreased purchase cadence and support ticket volume.

Lifetime value prediction

Forecasting how much a customer will spend over their entire relationship with your business. Used to decide how much to spend on acquisition and which customers to prioritise for retention investment.

Propensity modelling

Calculating the probability that a specific customer will take a specific action, such as upgrading, buying a particular product or responding to a campaign. Drives personalised targeting.

What it actually means

Predictive models are trained on historical data to find patterns that precede outcomes. If 80% of churned customers reduced their login frequency by 50% in the 30 days before cancelling, a predictive model will flag current customers showing that same pattern. The model does not know the future, it identifies customers who look like past churners and gives you time to intervene.

Predictive analytics does not tell you what will happen. It tells you what is more likely to happen based on what has happened before.

How it shows up

Predictive analytics shows up in CRM lead scores, email platform churn risk flags and ad platform lookalike audiences. In practice, the impact is measured by the downstream outcomes: did high-scoring leads convert at a higher rate? Did intervening with churn-risk customers reduce churn?

The Australian context

Australian businesses in financial services, retail and telecommunications are the most advanced users of predictive analytics, driven by large customer data sets and regulatory pressure on retention. For most other sectors, the practical application is simpler: using GA4 predictive audiences or Klaviyo's built-in churn risk scores.

Where people get this wrong

Building predictive models on too little dataModels trained on fewer than a few hundred outcomes are unreliable. The predictions will look confident but will not generalise. You need enough historical examples of the outcome you are predicting.

Treating a predictive score as a certaintyA 90% churn probability means the customer looks like 90% of past churners, not that they will definitely churn. Always use predictive scores to prioritise action, not to replace judgement.

Common questions

Do I need a data scientist to use predictive analytics?

Not for the common marketing use cases. Tools like Klaviyo, HubSpot and GA4 have built-in predictive features that require no modelling expertise. You need a data scientist when you are building custom models on proprietary data at scale.

How much historical data do I need before predictive analytics becomes useful?

As a rough guide, you need at least 500 to 1,000 examples of the outcome you are predicting. If you are predicting churn, you need records of at least 500 customers who churned and the behaviour that preceded it.

Keep exploring

About New Rebellion

New Rebellion is a marketing intelligence consultancy. We build tools, score Australian businesses on how their marketing actually performs, and publish Debrief every day. This dictionary is part of how we work in the open.

How we think →