Cardlytics: Scaling the Advertiser Side of the Network

Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

Randall, can we take a step back to when you first joined Cardlytics? What was your role and responsibility when you first joined the company?

I was the group president for the advertiser side of the business, working for the two co-founders. Cardlytics has an advertiser and banking side, and I ran the advertiser side with sales, client service, data and analytics; all the various functions supporting the advertisers and their programs on the platform.

What originally attracted you to the business?

Cardlytics had an interesting value proposition for marketers at scale. There are few marketing activities which drive significant top line growth at a high return on ad spend, because of the scale of the platform. There was a handful of interesting things under what they had built. They had the ability to use the anonymized purchase behavior data from credit and debit card transaction data of 161 million monthly average users. They could use that for targeting and could bring in second party data to match to that. They could deliver the cashback offers through the banks and e-banking applications.

That could be either the bank's app or their online banking website, which meant a pristine channel free from bots and fraud, all behind the bank firewall, a safe environment for marketers. The third thing was their closed loop measurement of the actual sales impact of the programs they were running. When you do those three things at scale, with a US audience of 160 million monthly average users, that is an interesting proposition for a marketer.

How would you describe closed loop measurement?

Think about the gold standard of efficacy in the pharmaceutical world. When a new drug is being tested, they use a double-blind test giving the drug to a group of people. Another group of people get a placebo, and nobody knows who is getting what, other than the people running the test. They compare the results between the two and, in essence, closed loop measurement does something similar to that. Cardlytics would typically run a campaign and 10 million people would see the offer, but several hundred thousand people – that were just like the people who saw the offer – were not served the offer. Over the course of the campaign, you can look at how the purchase behavior differs between these two groups.

That is considered the gold standard of measurement in advertising and media. It is very hard to do and few players can do it, but it is the single best way to measure the efficacy of advertising and promotion.

How would you compare closed loop measurement with the multi-touch attribution on Facebook, Google or other walled gardens?

At a high level, multi-touch attribution modeling measures a range of marketing activities. For the most part, they use multiple regression to understand how each of these independent marketing activities are having an effect on some outcome variable, typically sales or revenue. It is a statistical model which allows you to understand each marketing activity's comparative impact on the outcome variable. It is not as precise or definitive as closed loop measurement, but you can see how the activities compare on a level playing field.

One of the issues Cardlytics has, even with their closed loop measurement, is that if they offer a three to one return on ad spend, how does that compare to the other activities? They do not have closed loop measurement for all the other activities, so there is no simple way to do a direct apples-to-apples comparison of all the activities.

So an agency or marketer might spend $100 on Facebook and get multi-touch attribution measurement, but they cannot measure it against Cardlytics' closed loop. Could they see a higher return on Facebook or on a dollar amount that could distort the comparison which makes it hard for marketers to switch dollars to Cardlytics?

As I said, it is hard to make a direct apples-to-apples comparison. Having said that, in market mixed modeling or attribution modeling, for each activity you do, you will see for every dollar I spend, I get X dollars back. If I was doing display advertising on a particular platform, for every dollar I spend, on average, I got $1.23 back. Cardlytics gave me a five to one return on ad spend, but they are different measurement systems so it is simply an approximation. Companies like Nielsen, where I used to work, have done attribution modeling for years and have built up normative databases about the typical returns on different kinds of activities.

The normative data they have for TV, display or online video advertising, are less than $2 for every dollar you spend. If I’m a company like Cardlytics and I can prove I can deliver $4 or $5 return on ad spend, that looks good in comparison to what most people get from TV, online video or display advertising.

How do marketers compare returns with no true apples-to-apples method?

Some simply make an imperfect comparison. There are market mixed modeling providers who enable you to take closed loop measurement results or A/B testing, and include them in the market mixed modeling as what they call a prior. In some cases, you can do a comparison across things, but only certain companies do that today. It depends on the marketer and who they use as a marketing mix or attribution modeling provider.

What could improve the situation for Cardlytics?

Comparison will always be a challenge until Cardlytics allow individual personal level data outside their platform. Banks are uncomfortable with them sharing that data, even when anonymized and no PII is involved. Without doing that, it will stay where it is today. Historically, Cardlytics delivered the offers and measured the sales lift and outcome with their closed loop measurement system. For many marketers, especially those who did not spend large amounts of money on the platform, that was probably okay.

There were also marketers who, on principle, said, you cannot measure yourself. They could not spend lots of money on your platform, have you do the targeting, deliver the offers and measure yourself. It was not that they did not trust you personally, it was simply that, in principle, that was not the way they wanted to run a business. The same thing is true of the big media platforms. If I am a big brand and I spend a bunch of money on YouTube, there is a third party there to measure what I got, which is typically Nielsen, ComScore or others. One thing we did when I was there was to bring Nielsen in for third-party measurement. If a marketer spent above a certain amount of money on the platform, or if they were a new marketer on the platform, we would use Nielsen as a third-party measurement company to show them what they got, based on measurement by Nielsen who is a trusted third party.