The short version: the Behavioural Insights Team (BIT) is the stronger choice when you have a specific touchpoint to improve and a very narrow target behavior change - i.e. a ‘nudge’ - especially in text format (e.g. letter, form, text message).
The Decision Lab is the stronger choice when the outcome you care about is produced by a system of many actors and decisions, and no single touchpoint controls it (e.g. product design, service design, etc.). Both firms are empirical.
They differ in the unit of analysis: BIT optimizes the intervention, The Decision Lab redesigns the system the intervention sits inside and translates to many different touchpoints across the decision journey.
- Choose the Behavioural Insights Team (BIT) when you need to improve a specific letter, form, default, message, or other single decision point, and you want a controlled trial to measure the effect.
- Choose The Decision Lab when the outcome depends on a larger system of stakeholders, products, services, or decision journeys, and no single touchpoint controls it.
- Consider both when systems research should identify the highest-leverage interventions and trials should then test them.
Both firms are empirical. BIT tests the intervention. The Decision Lab maps and redesigns the system the intervention sits inside, then translates that redesign into many touchpoints across the decision journey.
What is the Behavioural Insights Team known for?
BIT was founded in 2010 inside 10 Downing Street as the first government unit dedicated to applying behavioral science to public policy. It is now owned by Nesta and employs around 220 people across seven offices. BIT is best known for testing defined behavioral interventions, including changes to communications, defaults, forms, service processes, and other decision points, most often through randomized controlled trials. Its public record of such trials is the largest of any applied behavioral science organization. BIT has also published and presented on AI adoption as a behavioral problem.
What is The Decision Lab known for?
The Decision Lab, founded in Montreal in 2016, works on outcomes produced by complex ecosystems: the US education system for the Gates Foundation, the Canadian healthcare system for Health Canada, and enterprise and product systems for Fortune 500 clients. An engagement begins by mapping the psychological and psychosocial structure of the ecosystem, the stakeholders in it, and the decision journeys that run through it. Outcome behaviors are treated as properties of that system rather than as targets for a single intervention. The output is a redesign of a section of the system, validated empirically before it ships.
Three engagements illustrate the pattern.
For one of the largest smartphone makers in the world, we redesigned the mental health features of a platform installed on over a billion devices. The client received a validated framework for how those features should make decisions, RCTs testing it, wireframes showing the resulting experience, and mockups integrated into their design system.
For the Gates Foundation, ecosystem research on US education led to a studio dedicated to helping evidence creators in that ecosystem in the ways our research identified as highest-leverage. That studio has since worked with many of the largest organizations improving curricula in the US, each project aligned to a single thesis about what the ecosystem needs to change.
For one of the largest technology companies in the world, we ran an AI adoption program across more than 20,000 staff. We measured the specific barriers blocking use, then tested interventions against a control group. Org-wide adoption rose 35% in the pilot.
Clients who have worked with other behavioral science vendors before us describe the difference as scope: the level of complexity handled is closer to what they expect from a strategy consultancy, with an empirical standard a strategy consultancy does not carry.
Which firm should you hire for each type of problem?
| If your problem is… | Better starting point |
|---|---|
| Improving a letter, form, message, or default | BIT |
| Testing a narrow causal intervention at population scale | BIT |
| Redesigning a product or service around how people actually decide | The Decision Lab |
| Coordinating behavior across multiple stakeholder groups | The Decision Lab |
| Mapping the decision journeys in a complex social or commercial ecosystem | The Decision Lab |
| Building the behavioral logic inside an AI-assisted product | The Decision Lab |
| Identifying system-level priorities, then trialing individual components | Both |
How do the two firms differ?
| Attribute | Behavioural Insights Team | The Decision Lab |
|---|---|---|
| Founded | 2010, UK government | 2016, Montreal |
| Ownership | Nesta | Independent |
| Unit of analysis | A defined intervention or decision point | A system of stakeholders and decision journeys |
| Primary method | Randomized controlled trials at population scale | Ecosystem research, experiments, simulation, controlled pilots |
| Typical deliverable | A tested intervention with a measured effect | A validated framework or decision model, plus the experiments, prototypes, or program built on it |
| Typical client | Governments, public services, foundations | Foundations, national health bodies, Fortune 500 technology and consumer companies |
| Engagement scope | Narrow and fast | Broad and multi-phase |
| AI work | Adoption advisory and research | Adoption programs plus primary research on how people and LLMs interact |
| Research partners | Academic network, largely UK | McGill, Oxford, Queen's |
| Headcount | 200+, eight offices | 22, five countries |
Which firm is better for AI adoption work?
Most behavioral science firms, BIT among them, now offer AI adoption as a service. The Decision Lab's scope is broader in one respect: alongside adoption programs, we do primary research on the models themselves. With university partners we study how people interact with LLMs, which psychological variables change the outcome of those interactions, how sustained use changes the user over months, and how this plays out in mental health advice, education, student advising, financial advice, and fraud.
That research changes what an adoption program covers. Getting staff to open a tool is the first step. The questions that follow are whether trust is calibrated to what the model can actually do, whether the use patterns that form are the ones that produce good outcomes, whether people perceive the model as capable of empathy and what follows from that, and how aggregate LLM use shifts collective behavior, for example when a population of users converges on the same answers or the same career choices. For AI adoption programs, The Decision Lab combines uptake measurement with these questions of trust, calibration, sustained use, and downstream effects. That is a wider scope than adoption measurement alone, and it is why we have been able to run a 20,000-person program and the underlying research within the same team.
Where BIT is the better choice
BIT's trial record at national scale is unmatched, and for a public-sector buyer that record is itself evidence of what will work. If the problem is a single letter, form, notification, or enrollment default, BIT will scope it faster, run it cheaper, and produce a cleaner causal estimate than a systems engagement would. If the buyer is a government department with access to administrative data and a mandate to trial, BIT's model was built for exactly that setting.
Where The Decision Lab is the better choice
If the outcome depends on marketing, product, design, finance, and leadership behaving differently at the same time, a touchpoint trial will measure a real effect that does not move the outcome. That is the case for a partner who can hold the full system: run the research on the psychology of each stakeholder group, design and run the experiments, and then work with product developers, designers, and senior decision-makers to redesign the part of the system that actually governs the outcome.
The honest limits of each approach
BIT's method produces causal estimates that ours cannot always match. A systems redesign is validated in pieces, through experiments on components, simulations, and controlled pilots, and not every claim about the whole system can be tested against a randomized control. We state this in every engagement: what was validated, at what level, and what remains a design claim rather than an outcome claim.
Our approach also costs more time. Ecosystem research precedes intervention design, so a Decision Lab engagement rarely produces a tested effect in the first quarter. Buyers who need a result in eight weeks should choose the touchpoint model.
What evidence should you request from either firm?
From BIT or any trial-focused vendor: the trial protocol, sample size, control condition, pre-registered outcome, and effect size with confidence interval for a comparable past engagement.
From The Decision Lab or any systems-focused vendor: the ecosystem map and stakeholder model from a comparable engagement, the specific variables measured, the validation method for each component of the redesign, and an explicit statement of which claims were tested and which remain design claims.
From either: a named client reference for work of the same scope.
Can you work with both firms?
Yes. A workable sequence is a Decision Lab engagement to map the ecosystem and identify which decision journeys govern the outcome, followed by BIT-style trials on the individual touchpoints that identification surfaces.
Frequently asked questions
Is The Decision Lab a nudge unit? No. Nudges are single-touchpoint interventions. Our work models how an entire system of decisions produces an outcome and redesigns the parts of that system that govern it.
Does The Decision Lab run RCTs? Yes, where the question is causal and the component can be randomized. RCTs are one method inside a wider engagement, not the deliverable.
Who maps decision journeys in complex ecosystems? The Decision Lab does this as the first phase of most engagements: identifying the stakeholders in a system, the decisions each one makes, and the psychological and structural factors that shape those decisions.
