Data science at Attest: an Interview with Erik Mathiesen-Dreyfus

The way Attest's data science team delivers deep insights is split into two parts: 1) by increasing the quality of the data we collect and 2) by automatically discovering insights in the results. We interview Attest's Head of Data Science to learn more.

There’s something smart going on behind-the-scenes. You don’t see it or hear it, but it’s there, and it’s constantly working under the surface of every survey our clients run. It’s making our results more accurate, it’s helping our surveys fill quicker and it’s learning all the time. It’s already so smart it can read survey results in milliseconds and make reasoned judgements about them. 

Erik Mathiesen-Dreyfus is Attest’s Head of Data Science, and his team is responsible for the wizardry that goes on in the background of your consumer insights. The data science team explores and implements cutting edge ways to bring Attest clients the best possible data and platform experience, so they can get on with making business decisions every day. 

In this article we’ll hear directly from Erik. Exploring data science at Attest, learning how his team has already improved data quality for Attest clients, and what they’ve got in the works, too. 

But first, some intros…

I guess I was a data scientist before the term was even coined. After finishing my PhD I started out in finance in 2007. I spent 8 years in quantitative finance (what you’d call data science today!) before leaving to set up my first startup. I co-founded two businesses, both with a data science focus, which are still alive and well, and then moved into market research. Having worked for so long on financial data, where things are often far removed from everyday life, it’s fascinating to work with data so close to our shared reality.

From there, via leading the data science team at another consumer insights company, Erik found Attest.

I was immediately attracted to how the company – and especially the founders – always put the long-term vision first, and have an ability to avoid getting caught up in short-term thinking.

Product-led vision

Our goal is always to build scalable, easy to use functionality. We don’t get involved in projects that seem ad-hoc or too niche; if it doesn’t help the majority of our clients gain better insights about their consumers, we shouldn’t be working on it. That approach differentiates us from a lot of other companies in the space, who take a much less product-led approach to their data science function. You may gain some quick wins but, you’re not building for the long-term, and you can end up with an incoherent product in the end.

For Erik’s team, that gave them a clear, definable goal; deliver deep insights to our clients using the techniques of data science. And to bring the richest data and deepest insights to Attest clients, the data science team has two strategies.

The way we deliver deep insights is split into two parts: 1) by increasing the quality of the data we collect and 2) by automatically discovering insights in the results.

Part 1: Increase the quality of data collected

We’re actively working on lots of strategies in this space: fraud detection, quality scoring and assisting with survey creation. We’re a relatively new team, but we’ve already made an impact when it comes to improving data quality by enhancing fraud and low quality detection methods. We have rolled these out internally with our in-house research team, ACE, and are currently working with ACE to automate the manual checks we’re already doing in the platform itself. These updates will make us best-in-class in data quality, and will provide immediate value to those who use the platform.

Part 2: Automate the discovery of insights

We’re working on a number of novel techniques – mainly in the language modelling space – as well as automated, modern versions of techniques traditionally used in the market research space.

The ability to derive insights from a range of disparate data sources in an automated fashion is where the market research sector is heading – not just as look-ups in the data, but as logical deductions, and with clear explanations associated with them. This is what most data scientists dream about in the field. We all have access now to enormous amounts of data, so the challenge is how to use all of this data and how to deduce insights that take all of this data into account – and in a way that you can trust and backup with an explanation! The technology for doing this is still underway, but it’s something we are deeply committed to at Attest and hope to make some industry-leading contributions in the not-so-distant future.

I’m super excited about our work on language modelling, and how we can use that to analyse and reason about unstructured data sets.

For Attest clients, the benefits are tangible.

We have several things in development in the data analysis track that will allow clients to more easily discover meaningful insights across any answer type; be it open text, multiple choice, rankings and so on. This will be the second wave of data science functionality in the platform and we are not at all far away from starting integration work on this.

Building a world-class data science team

Our data science infrastructure – which is truly state-of-the-art – has been a huge game changer for us. It has allowed us to, in a very short time frame, evolve a project from an idea to a production ready piece of functionality.

And this means more rapid updates to the Attest platform, and faster evolution of functionality for all our clients, meaning you consistently get the best quality data possible. So, how’s it done?

At Attest the product and technology teams work as cross-functional squads, in the style of Spotify. However, us data scientists organised ourselves as a separate team, more loosely attached to squads because of the different ways of working and timelines we work to.

A typical project starts with an idea, most often coming from direct client feedback. We then take these ideas through a number of ‘tests’. Our first test is whether it fits with our overall strategy. If it does, we then validate the need with our internal teams. If that’s all good we proceed to build a prototype. We’ve invested a lot of time and effort into building a smooth prototyping process that allows us to rapidly integrate new functionality into an easy to use frontend for internal testing. We then take the prototype on a ‘roadshow’, pitching it to the different product squads, and showing how it can help them achieve their target metrics. If it is good enough to convince a squad to take it on, we then work with the squad to get it live in the product.

In many ways we function as a startup within a startup!

It’s thanks to the unique structure and focus of the data science team that they’re able to take client feedback and so rapidly test, iterate and implement features that improve the access to great quality consumer insight for all our clients. 

Keep your eyes peeled for cutting edge data science functionality coming to the Attest platform soon. 

And sign up to our platform  if you’d like to get access to the highest quality, most insightful data out there.


Content Team 

Our in-house marketing team is always scouring the market for the next big thing. This piece has been lovingly crafted by one of our team members. Attest's platform makes gathering consumer data as simple and actionable as possible.

See all articles by Attest