How Gain Theory Streamlines Ingestion Of Thousands Of Data Feeds With CloverDX


Gain Theory, part of WPP, is a global marketing effectiveness consultancy that empowers marketers and insight professionals to make faster decisions using data, technology and advanced analytics. The consultancy exists to inspire marketing excellence by creating data informed cultures that drive business growth, profit and market share.

The consultancy provides a unique marketing analytics platform to their clients, and they were looking to speed up and streamline the process of getting huge amounts of data into their models. 

We spoke with Brian Suh, Transformation Practice Lead at Gain Theory, to discover how they’ve used CloverDX to make the data ingest process more efficient.  

CloverDX: Hi Brian, thanks for chatting with us. Tell us a little bit about what Gain Theory does. 

Brian Suh: Our job at Gain Theory is to enable meaningful decision making, at speed for clients. Using data, analytics and technology, our consultants help marketers improve marketing performance, delivering growth in profit and market share.

Our clients have terabytes and terabytes of data which feed into the marketing analytics decision making process. The pain point they face is translating a large volume of data and reports into meaningful action.

How setting up a data ingestion framework helps automate and speed up data onboarding - watch now

So, we are trying to make a huge volume of data more meaningful to decision making, and also transforming how our clients work, helping them utilize data to make decisions. 

How much of the data process you work with is standardized and how much is custom? 

Our modelling outputs are standard, but because clients' marketing efforts are varied, the inputs we get are rarely, if ever, “standard”. So, we end up customizing our data processes to standardize the inputs for our marketing analytics models.

Tell us about the challenges of that...

We gather client data in any format available to them. And the same exact dataset across different clients could all look slightly different. And some datasets require more cleansing than others.

Webinar - How engineers can adopt new data at least 2x faster - register

And we then have to transform it all into our marketing analytics modelling templates. So, even if the data’s perfectly clean there’s still more work to be done. We have custom data processing and cleansing rules – all custom, for each dataset, for each client. Clients can have thousands of data feeds coming in, each with their own rules. There’s a huge variety to handle, and at speed to deliver decision making around marketing programs.

And how were you doing that before you were using CloverDX? 

Since all of the client data is different, everyone in different parts of the organization was building their own technology to try and translate it into a format for suitable for each client. Because the data is different, there’s no way to standardize it, so everyone ended up doing their own thing, but all using different tools. 

Was that an issue?

It doesn’t translate that well into efficiency. If everybody’s using a different tool, then only people who know that tool can work on those particular projects.

So, it was all working, but it was working with a caveat. It was harder to transition teams, it always felt like we could be doing things a bit more efficiently. I knew there was a more elegant way to organize all of this – an opportunity to do better.

"I knew there was a more elegant way to organize all of this – an opportunity to do better."

Then you ended up choosing CloverDX to help with that. Why CloverDX? 

We had our requirements – including a visual GUI, readability, the ability to handle advanced code – and for us, CloverDX offered the best flexibility, checked off the boxes of the capabilities we needed, and pricing was competitive. We knew what we needed, we evaluated the tool, and we did the proof of concept to understand the technical capability to do what we needed.

How are you using CloverDX? 

The main way we’re using Clover is to standardize that data ingest process, make it more streamlined and consistent. It’s enabled us to standardize how we handle custom implementation, at speed.

Data Ingest with CloverDX

Is there anything in particular about CloverDX you liked? 

Part of our core requirement was to have the visual GUI. It’s easier, it’s more human readable, it’s easier for people to analyze code. We had thousands of lines of Python code, and if you didn’t write the Python, and the creator didn’t annotate it very well, it’s impossible to read.  

The flexibility in being able to integrate with our own custom packages and tools we’ve built was helpful. Our data science work requires advanced scripts, so being able to call Python and R from CloverDX, that was also good for us.

Has CloverDX changed how you work at all?

It’s really allowed us to get on the same page. Knowledge share and knowledge transfer has happened a lot more since we’ve been using CloverDX.

As an example - our core business is developing our models, so we have lots of people who are familiar with coding concepts and use different things like R and Python, but they’re not necessarily hardcore developers. With CloverDX it helps them be able to think more like a developer, but without having to learn it. For instance, a lot of people will write inline code rather than create functions. CloverDX graphs almost force you to create a function. It forces you to essentially comment it, or at least make it human readable. 

Because the tool automatically has documentation built in, there’s not an extra step and it’s just easier at that point. 

"It’s really allowed us to get on the same page. Knowledge share and knowledge transfer has happened a lot more since we’ve been using CloverDX."

How else is CloverDX helping? 

It’s helping us organize and be more efficient with what we’re doing.  

Our use case is a lot about width, as opposed to depth. We deal with thousands of data sources on behalf of clients for their marketing effectiveness programs. Each data source is not always going to be very complicated, but the fact there’s 1000 of them makes it hard to manage, so this helps us scale very easily.

In the instances where we do need a lot more depth to build 100-component graphs, we know that it’s all there and we know how to build that. So it gives us both ends of what we need. 

What would you say is the biggest change in what you’re doing now versus what you were doing before? 

The transferability across people. It’s just so much easier now.  

If someone has a use case that’s like ‘How can I do ABC?’ I can just take an example graph, send it to them and say ‘This does 80%, or all, of what you need’ and they’re able to quickly retrofit that to their client. 

Whereas before it would be: ‘Well, ok, you’re using SQL – I have another example that’s in Python, but you don’t know how to do Python’. And if you get Python installed and I send you the thing, you then don’t have any of the packages. And it takes 2 weeks just to get you that, and I should have just rebuilt the whole thing for you in SQL.

And I’d always have to make that decision – is it worth a transfer, or is it less work just to rebuild it all? And if I rebuild it all, I have another headache in the future where I have all these branching codes. 

Therefore the biggest change is not having to do that any more, I can just shift things over.

And people are using a common language. Because of the shareability of the graphs, I can send them an example graph where they can reverse engineer it, or I don’t even always need to send graphs around, I can just tell people the component name and it’s easy enough for them to figure out from there.

"It’s helping us be more efficient with what we’re doing."

Can you summarize the impact of CloverDX for Gain Theory?

It’s really the efficiency part for us. We were doing the work already, and there’s always an infinite number of ways to do things, but there’s often a more efficient way to do things. It has transformed how we were doing things to make us a lot more efficient. 

Part of our approach with Gain Theory's clients is coming up with standardized processes and common language, so we really recognize the importance of those things. When we saw a tool like CloverDX, we saw how Clover could help us standardize our language and streamline processes just as we normally do for our clients. 

And because our business is helping people adopt new tools, we knew the hurdles that come along with that, and we saw CloverDX had a lot of capabilities out of the box that would make it as easy as possible for us to transition to using it as part of our process. 

Download the case study PDF

Discover how CloverDX can help your business streamline data ingest - book a personalized demo

How faster customer data onboarding fuels acquisition and growth

Posted on November 25, 2020
Try CloverDX for 45-days  Full access to Tech Support as if you were a customer

Where to go next