Consumer Duty communications testing FAQ

These are direct answers to the questions heads of Consumer Duty, compliance, customer communications and marketing ask when they look at testing customer communications.

How do I test Consumer Duty communications?

You test a communication by putting it in front of people who look like the customers who will receive it and recording what they understood, what they missed and who struggled. Dutified does this with a panel of AI-modelled customers built on your customer profile and the wider UK public, and reports the result across the four Consumer Duty outcomes. Read the step-by-step guide.

What is customer understanding testing?

Customer understanding testing is a structured check that the people who receive a communication understand it well enough to make good decisions. It relates to the consumer understanding outcome in PRIN 2A.5 of the FCA Handbook, which is one of the four Consumer Duty outcomes. Read the explainer.

What does Dutified do?

Dutified tests a firm's customer communications on a panel of AI-modelled customers and shows what customers understood, what they missed and who struggled. It covers emails, letters, web pages, app screens, ads, statements and call scripts while keeping a full decision trail for audit.

Who is Dutified for?

Dutified is for heads of Consumer Duty, compliance, customer communications and marketing at UK banks, building societies, insurers, asset managers, wealth and advice firms. These teams use it to see how customers respond to a communication before it is sent and again after it has been changed.

How do I test letters for vulnerable customers?

You test a letter for vulnerable customers by including readers who reflect the vulnerability characteristics in FCA guidance FG21/1 and by reading the results for each group separately. Dutified's panel includes the vulnerable groups in FG21/1 and reports who struggled as well as how many. Read the guide.

Who are the modelled customers based on?

The modelled customers are built on your firm's customer profile and on the wider UK public, including the vulnerable groups described in FCA guidance FG21/1. This lets a communication be read by people who resemble your customers and by the population beyond them.

Does any customer data go into Dutified?

No customer data ever enters Dutified. The panel is modelled on a customer profile and the wider UK public, so a firm tests a communication without sharing the data of the people it serves.

What does a Dutified result show?

A result shows what customers understood, what they missed and who struggled, organised across the four Consumer Duty outcomes. Each result comes with a full decision trail so that a reviewer can see how it was reached.

Can we retest a communication after changing it?

Yes, a firm can retest a changed communication with the same panel to show whether the change worked. Using the same panel means any difference in the result comes from the change to the communication and not from a different set of readers.

Who decides whether a communication is good enough to send?

Your firm decides, because Dutified gives feedback and does not give legal or compliance advice. Dutified reports how modelled readers responded, and your teams use that feedback alongside the FCA's own materials and your internal sign-off.

How accurate are Dutified's AI customers?

Dutified's AI customers are most reliable at showing which version of a communication is clearer and where readers lose the thread, and every model version is scored against published UK research before any client test runs on it. Each version is checked on how close its scores sit to the published figures, whether it ranks documents in the same order, whether it moves the right way when a document is improved and how it does for vulnerable customers in particular.

What do you measure the AI customers against?

We measure them against a fixed baseline of published UK studies from a mix of regulators, government, charities and independent researchers, so no single organisation sets the bar. The baseline is locked before a new model version is built, which means a version cannot be tuned to pass the test it is marked on.

What are the AI customers built from?

The AI customers come from Dutified's own consumer model, which is informed by more than 100 published UK studies on literacy, numeracy, financial knowledge, digital skills, health and life events. No customer data is used and no single study or organisation sits at the core of the model.

How does a new study get into the model?

Every new study passes a gate before it can influence a result, where we check that we are allowed to use it, check its figures a second time, judge its quality and test whether it makes the model better at predicting studies it has not seen. A study that would swing a result sharply on its own is held back until more evidence points the same way.

Do AI customers just give the average answer?

Drifting towards the average answer is the best known weakness of AI respondents, and Dutified is designed against it by giving each AI customer their own reading level, numeracy, confidence and circumstances drawn from the spread found in UK studies. Results show the range across your customers and a separate run weighted towards vulnerable customers, so the people at the edges are not lost in an average.

Are the AI customers digital twins of real people?

The AI customers are modelled people built to match the mix of a population, and none of them is a copy of a real individual. That is why a firm can run a test without sharing any personal or customer data.

Will I get the same result if I run a test twice?

A retest uses the same standing set of AI customers, so a change in the result comes from the change you made to the communication rather than from a different group of readers. Each result also records which version of the model produced it, so it can be traced and repeated later.

What happens when the model is updated?

A new model version replaces the old one only after it has been scored on the same locked baseline and done at least as well, and every test records the version it used. Past results stay exactly as they were, and you can rerun a communication on the new version whenever you want a comparison.

Where is Dutified less reliable?

Dutified is strongest at comparing versions of a communication and pinpointing the sentences and groups that cause trouble, so its exact percentages are best read as a guide to the size of a problem rather than a forecast of a precise figure. It is also less certain for very new products where little has been published about how people understand them, and the report flags any finding that rests on thinner evidence.

Can we add our own research?

We are building a way for firms to bring their own research and outcome data, such as complaint themes or call reasons, so the AI customers reflect what you already know about your customers. Every client source will pass the same gate as a published study before it is used.

How is Dutified different from other synthetic research tools?

Dutified is built for one job, which is testing whether UK financial services customers understand what a firm sends them, measured against the Consumer Duty and the FCA rules for each product. General synthetic research tools are designed for product and marketing questions about what people prefer, while Dutified checks understanding, the decisions people would make and how vulnerable customers fare.

Can we see how a result was reached?

Every finding names the rule or evidence behind it, and each result keeps a decision trail of what was tested, which model version ran it and what changed between versions. A model card describing how the model is built and checked is shared with clients on request.

What kinds of communication can Dutified test?

Dutified tests emails, letters, web pages, app screens, ads, statements and call scripts. Firms use it for customer communications and financial promotions across banking, insurance, investment and advice.

What are the alternatives to customer panels for testing communications?

Firms also use focus groups, online surveys, readability scores and internal review, which each answer a different question. AI-modelled customer panels such as Dutified add a way to test more communications and to see which groups struggled. Compare the approaches.

What is the consumer understanding outcome?

The consumer understanding outcome is one of the four Consumer Duty outcomes, set out in PRIN 2A.5 of the FCA Handbook. It asks firms to support customers' understanding so they can make effective, timely and properly informed decisions. Read the explainer.

How do I test a financial promotion with customers?

Show the promotion to readers who resemble the people it will reach, record what they understood about the product and its risks, and see which groups got it wrong before you change it and test again. Read the guide.

What are synthetic customers?

Synthetic customers are AI-modelled people built to respond to material the way a described group of real people might. Dutified uses a panel of them, built on your customer profile and the wider UK public, to test communications. Read the explainer.

How do I evidence customer understanding for the board?

Keep a record of what readers understood and missed, which groups struggled, what you changed and what a retest showed, for each communication that matters. Read the guide.

Which FCA documents should I read?

Start with PRIN 2A in the FCA Handbook, which sets out the Consumer Duty rules, then read FG22/5 on the Consumer Duty and FG21/1 on the fair treatment of vulnerable customers. The FCA's own Consumer Duty pages link to all three.

Where to look at the FCA

Definitions of the terms used here are in the Consumer Duty communications glossary.

Dutified gives feedback and does not give legal or compliance advice. It reports how modelled readers responded, and every decision stays with your firm.

To test your own communications with Dutified, join the waitlist or read how it works.