Teach First

Our in-house “statistical Batman”, the brains behind the Ambassador Survey

Our community of ambassadors is one of our most valuable resources. Not only do they devote skills and energy towards addressing educational inequality in the classroom, they also provide insights and opinions that help improve our work. Every year we run our Ambassador Survey to gather this feedback – it’s a mammoth undertaking, taking weeks to design and weeks to analyse thousands of responses. 

Since data gathering algorithms are so pervasive now, it’s easy to imagine the results of a survey not resulting in any human action, but this isn’t the case with the Ambassador Survey. In fact, every single result is logged and studied by our Ambassador Impact Data Officer, Ewan Wakeman, who describes his job as “trying to understand what ambassadors do, want and need from Teach First.” He loves the thrill of discovery that accompanies each survey. “All sorts of interesting things start appearing when you spend enough time with some data, and in my role you’re always the first person who gets to see that,” he says.

Because of the valuable insights we’ve gained from previous year’s surveys, we are becoming more and more of a data-led organisation, a change that Ewan and his team have been pivotal in driving.

“The problem has always been the delay between investment in data and gaining the rewards from it,” he explains. “But I think Teach First is now really starting to realise, not only how data can help us to understand and articulate the problem of educational inequality, but also how we can use it to ensure our strategy is focused where it needs to be.”

Gathering data is risky business, especially through a survey. It’s easy to introduce minor mistakes in the design of questions that result in very misleading outcomes. Ewan is full of examples of the things he has to avoid.

“There is a common error, particularly in surveys, where people ask questions that are different to what they want to know,” he says. “Commonly this is things like asking whether someone ‘enjoyed the training’ when really what they wanted to know was whether the ‘training was high quality’. These two things may be highly correlated, and you may be able to ascertain that if more people enjoy the training in Session B than Session A, then Session B was also of a higher quality – but you’re better off sticking with the quality question to be sure.”

Ewan has a reputation within Teach First as being a kind of statistical Batman, capable of feats that are beyond the reach of most humans, and he applies the full weight of these skills in working through the responses to the survey. It goes beyond just designing pivot tables in Microsoft Excel – he’s even taught himself to code in order to be able to do more with the data.

“I’ve been learning more about how to conduct various types of regression and cluster analysis, and on the technical side I’ve just taught myself M and DAX, which are Microsoft data coding and analysis languages,” he says, as we nod and simulate understanding. “I’m now in the process of learning Python and some more advanced R.”

Despite all the arcane details, what drives him to turn data into knowledge is an impulse we can all recognise – the desire to progress through understanding.

“My friend Alice just learned how to teach R to recognise tumours from images of neural scans of the brain, which was pretty incredible to see,” he says. “Working with data is a constant series of challenges, to try and extract meaningful information from seemingly meaningless streams. When you manage to get that to work, you can give people answers to their questions and enable everyone to do their job better – there’s a joy in that.”

What would his one wish be for the Ambassador Survey? To have even more data to work with.

“I think in the modern world of data science, quality comes from quantity,” he explains. “Datasets are becoming so large now that your ability to infer meaning from one thing to another becomes incredibly easy to test and predict. We can look for simple things like concurrence across millions and millions of rows of data – for example, if a person buys X product how likely are they to also buy Y product or if they use X word how likely are they to follow it with Y word. This is basic, but if you have a large enough amount of data it becomes really accurate, really quickly.”

Without that accurate data, we won’t able to do the best for our ambassadors – so if you’re one of our community, don’t delay in filling in your Ambassador Survey. Ewan is eagerly awaiting your responses – don’t let him down.