Julian Tait, the chief executive of the Open Data Manchester CIC, opened this discussion by saying that “a lot of data comes from a very top-down, managerial perspective.” It “tries to put people in boxes” that “don’t fit their lived experience” and that leads to “poor decision making.”

So, he asked, “how can we as data practitioners make sure data better represents the people we want to serve” and “we don’t get so much sh*t policy.”

Specifically, he said, his organisation is working on a project to tackle violence against women and girls. Often, initiatives in the space focus on better lighting, or more police, when the feedback is this doesn’t work – and those affected might have completely different ideas.

One problem is that data can be funded by organisations with agendas. Two participants talked about doing research with offenders, with a view to addressing some of the challenges they face when they leave prison.

They said the Home Office wanted to roll out a prescriptive programme, while homeless charities had their own view of what services should look like. Yet neither matched closely to the needs of the ex-offenders themselves.

At the next level, one participant suggested that what is needed is a “mindset change.” Instead of labelling people as “hard to reach”, researchers should be asking “how do we reach them?”

Another picked up on this, and said it’s important to think about where data is collected or interviews are conducted. People without their own transport might not be able to get to a public building. Some spaces will be more gendered than others.

Similarly, it’s important to think about when data is collected. People who work are not going to make a meeting in the middle of the afternoon.

Overall, though, the session surfaced the big, underlying issue as trust. People may not trust researchers or what will happen to their data, and may have good reasons for that. “You have got to create trust,” Julian said.

And there was some agreement that the best way to do that is to get down and personal with people. Don’t try and run huge data projects, run small projects, but run them frequently to pick up trends.

Ask the community to collect their own data, and share it (if there is a community to work with, and leaders really represent the people they claim to speak for). And, if nothing else, make sure it’s clear why it matters.