Online support for families with Action for Children
“Working with DataKind completely opened our mind to the possibilities of data, and gave us access to so many brilliant data scientists and their different perspectives.”
Lynn Roberts
Director of Growth and Service Design, Action for Children
Action for Children is a UK-wide charity that supports parents and caregivers of children. In July 2020, they launched Parent Talk, a free service where parents can find expert written information and advice, and one-to-one parenting support.
The service was launched as a response to the gaps created by funding cuts to universal services for parents and families, a situation that worsened during the pandemic. Between April 2021 and March 2022, nearly 9,000 parents and carers used Parent Talk to speak with a parenting coach.
How could Action for Children’s data help them?
To help them tackle and understand this ever-growing need, Action for Children took part in a pro bono DataDive project in spring 2022. Members of their staff team worked closely with experienced, volunteer data scientists from DataKind UK on a few main challenges:
Understanding parents
Action for Children wanted to improve their understanding of the challenges Parent Talk users are experiencing. Learning what people ask most frequently will help them to provide better information, advice, and guidance that can be accessed 24/7. These insights can be used to train their coaches, as well as to create advice and resources to help people out of hours, when the main coaching service is not live. It will also highlight any gaps in local or national services where people are not finding the support or advice they need – whether that’s from other organisations and agencies, or Parent Talk itself.
Action for Children is also passionate about building their role as an expert and advocate for parents’ and caregivers’ needs. They want to use their experience, reinforced with data, to draw public and political attention to what’s needed to better support families, parents, and caregivers across the UK.
Encouraging user feedback
Like any service, getting lots of feedback is key to knowing how to improve. Action for Children wanted to know what encourages, or discourages, parents from leaving feedback, and how they could increase the overall amount of feedback they receive. As well as steering them in other ways to improve the service, they hope increased feedback will help them understand more about Parent Talk’s impact.
Using webchat data safely
Parent Talk’s coaching happens over webchat, so analysing this webchat data was a key piece of the puzzle in understanding how to better support their users’ needs. This is difficult because webchat data is likely to be filled with a lot of private data (or PII, Personally Identifiable Information) and contain potentially distressing content about parenting challenges.
Action for Children puts the utmost importance on ensuring no personal information is ever seen by anyone outside of the Parent Talk team. To ensure the data was safe for use by volunteer analysts during the DataDive project, DataKind UK helped them to prepare it.
Chat transcripts and any attached ‘metadata’ (such as user IDs and timestamps) were anonymised thoroughly, first through automated methods that removed words related to personal information, and then through manual checks by Action for Children staff to confirm that no Personally Identifiable Information was included.
Action for Children then went even further to safeguard the data science volunteers who would analyse the data. The webchat conversations were turned into ‘bags’ of words (literally called bag-of-words modelling). This means that the chats are no longer conversations with sentences and phrases, but instead become a table showing the frequency of each word in each message. Each message is represented in a row, and could be from an Action for Children coach or user. This removes some of the meaning behind the words, as it strips away their context. Bags of words let data scientists detect the common words and themes, while providing a layer of protection from sensitive content.
| help | the | baby | in | sleep(s) | cot | |
|---|---|---|---|---|---|---|
| help the baby | 1 | 1 | 1 | 0 | 0 | 0 |
| sleep in cot | 0 | 0 | 0 | 1 | 1 | 1 |
| the baby sleeps | 0 | 1 | 1 | 0 | 1 | 0 |
| help baby sleep | 1 | 0 | 1 | 0 | 1 | 0 |
An example of a bag of words showing short phrases turned into a table, where columns are the words used in the phrases, and rows represent each phrase, with numbers appearing beneath each word used in the phrase to show how often it occurred in that phrase. The phrase ‘Help baby sleep’ appears as three columns, help, baby, and sleep, with a one beneath each word to indicate its frequency.
What they found out
Finding themes
Once the bags of words were prepared, they could be analysed for common topics and themes and compared to the categories Action for Children already uses. After each webchat conversation, coaches add tags such as ‘Mental health’, ‘Development’, or ‘Legal aid’. Conversations can have more than 10 tags each, and there are hundreds of tags and sub-tags for coaches to choose from. Action for Children wondered if the tags captured all the types of needs of their users, or if there were other themes that users spoke about.
To analyse the webchat data, DataKind volunteers used an approach called topic modelling. Topic modelling finds groupings of words that have the same or a very similar meaning, or represent the same concept. Using the bags of words, volunteers excluded common words like ‘the’, then asked the model to group the content into a chosen number of topics.
Bar graph showing the top conversation topics across almost 8,000 conversations. ‘Legal’ was the top topic in the highest number of conversations, followed by ‘family’, ‘abuse’, and ‘SEND/school’
The model doesn’t tell you what the topics are – it provides groups of words, which the team at Action for Children were able to label. From Action for Children’s data, the three largest topics were identified as ‘Legal’ ‘Family’, and ‘Abuse’ (shown in the bar chart below). Action for Children also frequently supports parents and carers about Special Educational Needs or Disabilities, known as SEND. ‘Mental health ‘ and ‘potty training’ were also common topics.
Generally, Action for Children’s existing tags aligned well with the topics they found. But more chats were about legal topics such as ‘legal advice’ and ‘allied services/mediation’ than were shown by the manual tagging.
This suggests that the tags around legal topics could be refined, and that users may want more support in these areas. As Parent Talk cannot provide any legal advice, understanding how they might better signpost parents to the right legal services was insightful in itself.
Volunteers also clustered conversation tags into very broad themes that will help Action for Children arrange information more intuitively on their website, so that users can find it easily.
Trends over time
Next, the team looked at how themes in the webchats had changed over time. They noted a broad change in the type of advice sought as the pandemic progressed, from a peak in conversations about behavioural management, sleep, living arrangements and maintenance payments early in lockdown; to a rise in conversations around education and SEND, presumably as children returned to school. Overall, mental health and SEND issues stood out as increasing dramatically since the beginning of the pandemic.
Different needs at different times
The volunteers also found that parents’ and children’s needs evolve in a pattern as the child grows. Initially, parents are concerned with legal or practical issues, but over time, topics such as mental health become more prevalent, and peak around age 13-15. SEND conversations also increase with the child’s age.
Webchat feedback
Finally, the feedback from webchat conversations showed that longer conversations with lots of replies from the coach get rated more highly by users. This helps Action for Children to make the case for increasing their capacity, and apply for further funding to do so.
What’s next
The project helped Action for Children decide to launch a new contact centre in August 2022. They are working on building their reporting and how they record their impact. They are also hiring for a data role, and have been able to use the skills identified during the DataDive project to build a job description.
Lynn Roberts, Director of Growth and Service Design at Action for Children, said that, “Working with DataKind completely opened our mind to the possibilities of data, and gave us access to so many brilliant data scientists and their different perspectives. It’s something we could never have done by ourselves, and it’s given us a talking point to share the benefits of investing in data within our organisation.
It also created the basis of our first annual Parent Talk report, meaning we are sharing children and families needs with the UK public and decision makers, and raising awareness of the gaps in support.”
With huge thanks to the Data Ambassadors who led the volunteer team: Laura Batchelor, Péter Udvardi, Tanya Rofani, and troubleshooter Sukhil Patel, as well as everyone who volunteered over the DataDive weekend.
