Two questions sit at the heart of every commercial decision. What is likely to happen? And what will change because of what we do? Most analytics answers only the first. adBrain builds both, because a prediction you can’t attribute to a cause is a decision you can’t trust.
Prediction: estimating likely outcomes
- Propensity modelling: the probability that a customer converts, churns, responds to an offer, or takes an action, used to rank and prioritise. This is what turns a customer list into a next-best-action queue.
- Forecasting: demand, revenue, and load forecasting that a team can plan and allocate against.
- Behavioural scoring: turning raw event history into features that models can act on.
Prediction is where most vendors stop. It tells you who and how likely, but not whether acting actually made a difference.
Causality: proving what changed
Correlation is cheap; causation is what you get paid for. adBrain applies causal methods to separate genuine impact from things that would have happened anyway.
- Incrementality & lift: measuring the additional impact of an intervention beyond the organic baseline, so budget follows real effect rather than last-click credit.
- Counterfactual reasoning: estimating what would have happened without the decision, so you can compare it against what did.
- Causal inference where classic A/B isn’t enough: geo-based tests and quasi-experimental methods for situations (channel-level spend, market-level effects) where a clean randomised split is impractical.
This is the same shift the strongest performance-AI vendors have made: away from attribution and vanity metrics, toward incrementality and proven lift. It is grounded in well-established methodology from the experimentation and causal-inference literature, including public research on causal impact and incremental return on spend.
On the numbers we publish. We frame outcomes as what a system is designed to move, and we substantiate any figure we quote against a client’s own measured results. Objective performance claims on this site follow UK consumer-protection rules. See the compliance checklist.
Prediction + causality = trustworthy decisions
Put together, the two disciplines close the loop: predict the likely outcome, act on it, and then measure the causal lift of having acted. That measured lift feeds straight back into the next model and the next decision, which is exactly where experimentation & optimisation and machine learning decisioning take over.