Greater Cambridge Partnership · Public Sector
What six months of manual coding missed
How the Greater Cambridge Partnership tested AI against six months of manual coding, and what it found.

DWP: what 300,000 tweets told policymakers about child maintenance
Fay Cooper’s job at DWP Digital is to bring together what users say across every channel and use it to guide what the team builds. When Wordnerds analysed 300,000 tweets about DWP’s child maintenance service, the overwhelming feeling was stress and anxiety, and policymakers started paying attention.
Last updated 13 November 2026
At a glance
- The Department for Work and Pensions, the UK’s biggest public service department
- Stress and anxiety found in 300,000 tweets about the child maintenance service
- Insight from every channel brought together with user research
- Told by Fay Cooper, Lead Product Manager, DWP Digital
How could DWP hear from users of every age, without sending researchers across the country?
Fay Cooper is Lead Product Manager at DWP Digital, the team that builds the digital services millions of people use to deal with the Department for Work and Pensions. Fay supports product strategy across the department, and as the voice of the customer brings together insight from every channel and from user research, tests ideas against it, and uses it to guide what DWP Digital builds.
In government, every product starts with user needs, and DWP’s products have to work for people of every age and every demographic. Finding out what all of those people needed was slow and expensive, because it relied on regular surveys and on user researchers travelling the country to talk to as many users as they could.
When Fay’s team met Wordnerds, it saw a way to scale that research and its insight into what users needed.
To hear users at scale, DWP Digital had Wordnerds read what they wrote
DWP Digital started with its child maintenance service.
-
Listened where users already talk
Wordnerds analysed 300,000 tweets about DWP’s child maintenance service, so the team could hear from users without a researcher travelling to meet each one.
-
Showed what users need and what hurts
Wordnerds’ insights showed what users needed from the service and what was causing them friction.
-
Tested ideas against every channel
Fay’s team put those findings alongside insight from other channels and from user research, and used them to test its ideas before they shaped the product.
In 300,000 tweets, the overwhelming feeling was stress and anxiety
Wordnerds’ first analysis of 300,000 tweets about DWP’s child maintenance service shocked Fay Cooper’s stakeholders, because the overwhelming sentiment in them was stress and anxiety. That made DWP’s policy team pay attention, and it gave the department a real problem to solve.
Until then, finding out how users felt had meant surveys and researchers on the road. Now the policy team could see it in what users had written themselves, across 300,000 tweets.
DWP aimed its user research at the causes behind the stress
Using both AI and real people, DWP Digital focused intensive user research on three questions.
-
Why some parents have no arrangement
The team set out to learn why some people don’t have a child maintenance arrangement in place.
-
Why fathers find the service stressful
Research asked why fathers feel anxious and stressed by DWP’s child maintenance service.
-
How money could reach children faster
The team looked for what DWP could do to speed up how quickly money is paid to children.
Apply for Child Maintenance passed 93% take-up in its first month
DWP Digital used Wordnerds to scale its research into how users felt about child maintenance, beyond regular surveys and researchers travelling the country. Stress and anxiety in 300,000 tweets made policymakers pay attention, and DWP Digital’s new Apply for Child Maintenance service passed 93% take-up in its first month and reached 97% satisfaction.
Working with Wordnerds kept DWP Digital agile and let it deliver at pace. And once the policy team could see how users felt, the question changed from whether there was a problem to how to make things better.
About the Department for Work and Pensions
The Department for Work and Pensions is responsible for welfare, pensions and child maintenance policy. As the UK’s biggest public service department, it administers the State Pension and a range of working age, disability and ill health benefits to around 20 million claimants and customers.
About Wordnerds
Wordnerds turns what customers say into what organisations do. We ingest feedback from surveys, complaints, reviews, calls and social; apply transparent, explainable AI to surface themes and drivers; and serve the insight two ways from one semantic model: full-detail Power BI for analysts, plain-language AI-chat answers for everyone else. Every result traces back to what customers actually wrote, so it's evidence you can act on, not a black box.
Questions CX leaders ask
How do you understand user needs when a service has to work for everyone?
Add what users write unprompted to what surveys and interviews tell you. DWP Digital found surveys and researchers travelling the country slow and costly for services that must work for every age and demographic, so it had Wordnerds analyse 300,000 tweets about its child maintenance service and combined the findings with user research.
How do you get policymakers to pay attention to how users feel?
Show them how users feel at scale, in their own words. When Wordnerds analysed 300,000 tweets about DWP’s child maintenance service, the overwhelming sentiment was stress and anxiety. That shocked DWP Digital’s stakeholders, made the policy team pay attention, and turned the discussion to how to make things better.
Can AI analysis replace user research?
At DWP Digital, AI analysis showed where to point the research. After Wordnerds found stress and anxiety in 300,000 tweets, the team used both AI and real people to research three questions, including why fathers feel anxious and stressed by the child maintenance service and how money could reach children faster. What customer feedback analysis can't tell you looks at when you need research as well.