Skip to content

Stockertim Dictionary · Artificial intelligence · Business

Predictive analytics

Also called predictive analysis, predictive marketing analytics

In short

Using historical data with statistical or machine-learning models to estimate future outcomes, such as which customers are likely to buy, leave or respond to an offer.

  • Intermediate
  • Process
  • Checked

In detail

Predictive analytics is the practice of looking at what has already happened in order to estimate what is likely to happen next. In marketing it usually means scoring each customer or prospect with a probability: how likely they are to buy, to open an email, to respond to a discount or to stop buying altogether.

The work starts with past records where the outcome is known. A model, which may be a simple statistical formula or a machine-learning method, looks for patterns that came before each outcome, such as how recently someone bought, how often they visit and which products they viewed. Once the model has been tested on records it did not learn from, it is applied to current customers to produce a score for each one.

It differs from rules-based marketing automation, where a person writes the rule (send a reminder three days after a basket is left), and from generative AI, which produces new text or images. Predictive analytics produces estimates, and every estimate carries uncertainty. A score of 30% means that among many similar customers about three in ten bought in the past, not that a particular person is undecided.

The value lies in where effort goes: a team can spend a limited budget on the customers where action is most likely to change the result. The risks are stale data, patterns that no longer hold and scores that quietly treat some groups unfairly, so models are monitored and retested.

Worked example

Harbour Foods scores 2,000 lapsed customers with a model that estimates the chance each one returns after a voucher. One group of 400 has an average score of 30%; the rest average 8%.

  1. Expected returns in the high group: 400 x 30% = 120
  2. Expected returns in the rest: 1,600 x 8% = 128
  3. Vouchers sent if only the high group is targeted: 400, against 2,000 for everyone
  4. Share of expected returns reached with the high group: 120 / 248 = 48.4%

Targeting a fifth of the list reaches nearly half the expected returns, if the scores hold up when tested against real results.

In practice

Marketing teams use predictive scores to decide who receives an offer, which leads sales should call first, how much stock to plan for a promotion and which customers may need a retention message. Good teams compare predictions with what actually happened each month and retire models that drift.

Watch out

  • A prediction is not a cause: a model may find that coupon users buy more without showing that coupons make them buy.
  • Scores built on old data can fail quietly when prices, channels or customer habits change.

Learn it properly

Official sources

Facts last checked on 3 November 2026.

Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.

English