Decision science

The questions it answers

Did the ads cause the sales, or would they have happened anyway? How should a fixed budget be split between channels next quarter? Which of five projects goes first, and how do you defend the choice? Why do customers stall at the last step of a form? These need methods, not opinions.

The methods

Quantitative:

• Controlled experiments where they are possible
• Bayesian analysis, so results come with honest uncertainty
• Marketing mix modelling where experiments are not practical
• Multi-criteria decision analysis and structured foresight when there is more than one objective

Behavioural and social science:

• Interviews, surveys and workshops to learn why people act as they do
• Thematic analysis, so qualitative evidence is handled with the same care as numbers
• Framing, defaults and choice design for offers, ads, forms and pages

How people actually decide

Daniel Kahneman’s Thinking, Fast and Slow describes two modes of thought: one fast and intuitive, one slow and deliberate. Most buying decisions lean heavily on the first. So do more board decisions than anyone admits. People respond to framing, to what they see first and to a story that hangs together.

We use this in two places. In the work, we design offers, ads and pages around how customers really choose, not how a spreadsheet assumes they choose. In the readout, we present findings in a way people can take in and act on.

Nobody buys from one number

A correct model that nobody acts on has changed nothing. So every piece of analysis ends as a narrative: what we think is happening, why, how sure we are, and what it means for the next decision. Uncertainty is stated in words as well as intervals. The story is built from the evidence, never in place of it.

Where it shows up

Decision science is not a separate product we sell once. It runs through every account we work on. The analysis tells us what to do next, what to produce, and which systems and pipelines to build. When the question is bigger than the next campaign, it also gives clients strategy support: scenarios, foresight and what to plan for.

The “Is this change real?” tool on our maths page runs on the same Bayesian method. Put your own numbers in and see how much evidence a result really carries.

What you get

A short written readout: the decision, the evidence, the assumptions and the story that connects them. It can be revisited when things change. Training for your team, if they want to run the method themselves next time.

Where we usually start

1. An incrementality test, designed, run and read out
2. A budget allocation model with scenarios
3. A decision workshop that ends in an agreed, documented choice