SDS 481: Performance Marketing Analytics

This is episode number 481 with Kris Tait, managing director at Croud. 

Jon Krohn: 00:14

Welcome to the SuperDataScience Podcast. My name is Jon Krohn, chief data scientist and best-selling author on Deep Learning. Each week, we bring you inspiring people and ideas to help you build a successful career in data science. Thanks for being here today, and now let’s make the complex simple. 

Jon Krohn: 00:42

Welcome back to the SuperDataScience Podcast. I’m delighted to have the charming and knowledgeable Kris Tait joining us on today’s program. Kris is the managing director for the United States at Croud, an innovative marketing agency that is driven by data analytics and machine learning algorithms. Their data-driven approach is reaping dividends for Croud, who have experienced hyper-growth for years and were recently recognized by Campaign Magazine as the 2020 global performance marketing agency of the year. 

Jon Krohn: 01:13

In today’s episode, Kris fills us in on what performance marketing is, the rapidly shifting digital marketing ecosystem, as well as how data and machine learning can mitigate the risks associated with these changes. The sweet spot for augmenting human marketers’ skills with machines, how any firm should define metrics to maximize return on marketing investment, thereby ensuring broader commercial success, the various data related rules a digital marketing agency must have, and the most useful modern data science tools for global digital marketing. Today’s episode will be of interest to anyone interested in driving commercial success through effective data-driven marketing strategies, no particular technical data science background is required to make the most of it. 

Jon Krohn: 02:06

Kris, welcome to the SuperDataScience show. I’m absolutely delighted to have you on. I’ve wanted to have you on the show ever since I saw your YouTube webinar, that’s called ‘How not to get stuck in a performance marketing rut’. I learned so much during it, but anyway, we’ll get into all of that what performance marketing is later. First, let the audience know where you are, what’s going on in your world. 

Kris Tait: 02:31

Thanks, Jon, delighted to be here. I thought you’d never ask. It’s great to be here. I am in Brooklyn. I’m in Brooklyn. Getting through the nearly summer. I’m not too far away from you just over the bridge. 

Jon Krohn: 02:46

We’re just opposite sides of the Brooklyn Bridge. Full disclosure for the audience, I have been friends with Kris for many years. We met in a dog park in the west village neighborhood of Manhattan. We both used to live in the west village, basically nearby there anyway, and our dogs they’re just a month apart. Your dog, Ruby, is one month younger than my dog, Oboa. We started going to the dog park around the same time, and our dogs have always been obsessed with each other. 

Kris Tait: 03:18

Indeed. 

Jon Krohn: 03:19

Begrudgingly, we had to be friends. 

Kris Tait: 03:22

Luckily we got along, wasn’t it, when we were spending several hours in the dog park, pre 6:00 AM. 

Jon Krohn: 03:28

For sure. All right, so you’re here today to tell us all about performance marketing. First off, what is it? 

Kris Tait: 03:36

Great. Performance marketing is a term that has been coined probably over the last decade and a little bit more, but generally relates to the marketing practice of demand generation, lead generation, conversion, sales. So making sure that we can spend media to get people to our websites to convert. The term really got coined when Google and Facebook started accelerating the ways that you can market online with their platforms. Actually, I like to think of it as just marketing, but I think performance marketing describes exactly what we do relating to the media. 

Jon Krohn: 04:16

I guess maybe part of that is the idea that for a long time before there was digital marketing, when it was billboard ads and newspaper ads, it was probably very difficult to measure how effective your marketing was. Now these days with digital marketing channels, probably many of the channels that you use, you can at least have some sense of the performance of the marketing spend. 

Kris Tait: 04:42

100%, that is one of the key factors. Where in the past you would have to do inferred models or big econometric models to try and understand, we just put this TV spot live, we just put this radio spot live. What impact did it have on our sales? Now, we can see it within an hour, or we can see almost in real time if someone clicks, if someone converts. Honestly, that’s excellent in one manner where we can see very quickly what’s going on, and we can optimize. Also, I think creates a bad habit of being so reactive to things straight away. We’ll talk about that in terms of what that means for marketing in general. For me, it’s sort of gone so far down one route that people focus on everything they can see and forget about what we’re actually trying to say to the audience and what we’re trying to sell to the audience. We can talk about that one as well. 

Jon Krohn: 05:42

Nice. We will for sure. Before we get there, I think one of the big distinctions to make in the performance marketing space is this difference between multi-channel versus single channel marketing. What does that mean? What are those two things? 

Kris Tait: 06:00

I think one of the risks, and one of the trends that has been happening over the past few years is I think the stat is something like 90% of all new media spend, goes to Google and Facebook. They are the powerhouses of digital media spend, especially. Certainly just any media that you’re trying to put into the market, it’s very easy to execute it on those platforms. From a performance marketing perspective, they perform really well. They’ve built their ad platforms incredibly well. The machine learning algorithms, again, will come onto this later on very, very, very good. You spend a dollar on Facebook. You’re going to get a good amount of dollars back. As you’ve said, it’s trackable. We can know, put a dollar in, get $3 back. That’s been happening a lot. I think from a single channel point of view, that’s where people are going. They’re going to Google or Facebook. 

Kris Tait: 07:03

There was a study done by a gentleman called Mark Ritson, who… Look him up if you’re interested, really entertaining speaker, but he looked at hundreds of award entries for this award ceremony called the Effie’s. He found out that campaigns that used one channel were 37% less effective than campaigns that used multi-channels. When we say multi-channel, we mean you’re using video. You’re using Google search or Bing search. You’re using display ads. You’re using social ads, and you’re creating that multi-channel picture, holistic picture for someone to see your brand. If you’re only on Google, people aren’t noticing you, unless they’re searching for you. The stats are really compelling. I think that’s one of the things that we try to work with brands on is to give them that sort of holistic marketing plan, rather than say, let’s just spend our money on Google. Let’s just spend our money on Facebook. 

Jon Krohn: 08:05

Nice. I think that probably ties into the next point that I wanted to talk about, and I think you wanted to talk about, which is that diversification probably helps us avoid some of the big risks in performance marketing. It seems like in life, whether it’s a financial portfolio or career strategy or business strategy, diversification is key to success and resilience. 

Kris Tait: 08:31

100%. The risks of creating your whole business and by the way, there are a lot of companies in this current environment that have sprung up, like a lot of D to C companies, for example. They’ve sprung up. They’ve got really good product. They’re trying to compete against a razor or a shampoo or whatever it is. They’re still coming out daily. You see a new one. They built their businesses a lot of the time on performance marketing because they’re super smart people. They know how to be accountable, and they know how to use these platforms really, really effectively. The risk is that if something in the industry changes, if there’s a regulation that changes, if there’s a technology that changes, if for example, the antitrust lawsuit that’s going on in the government right now around Googles, the Amazons, the Apples of this world, if something changes with these platforms, your marketing strategy is open to a lot of risk if that’s all you’re using. 

Kris Tait: 09:39

That’s one of the big things that we’re trying to show people is that risk. One of the large things that’s happening in industry right now is the app tracking transparency. If you’ve updated to iOS 14.5 on your phone, and you’ve started updating your apps, you’ll see that they’re asking you if you would like to be tracked or not. It’s really interesting to see how that data is impacting people and brands. What that means for the industry is that we can’t see the websites or the brands on the apps can’t see the websites that you’re going on to visit. They sort of getting rid of cookies, which is what advertising and performance marketing has been built on for years. We cookie you Jon. You go to this different website. We can see that you’re interested in this. We can see that you spent money here. Then we can retarget you. Retargeting is the most effective channel because we know a lot about you. 

Kris Tait: 10:36

Again, so if your marketing strategy has been built a lot on that and iOS 14, Apple say, no, we’re looking after user privacy now. We’re not having that. That has a huge impact, a huge impact on your marketing strategy. Every brand that we’re working with right now is working through that. They’re thinking we’re going to have less data on our customers. What do we do about that? That might bring us on to what machine learning and what data science can do about that. 

Jon Krohn: 11:08

I’ll ask about that in one second. There’s one term that you mentioned. You said that one of the most effective marketing strategies is retargeting. What does that mean? 

Kris Tait: 11:17

Retargeting is effectively, just to provide some context. The terms that get talked about and banded around in our industry are prospecting. Prospecting is when you are looking for new customers that might’ve never heard your brand before mid funnel. Well, top of funnel is prospecting. Mid funnel is when you’re reengaging with people and warming them up to your brand. You know that they’ve seen one of your videos, but maybe they haven’t been to your website. You know that they’ve maybe seen one of your ads, but there’s no real connection there. The middle of funnel is you’re engaging these people again. You’re showing them what you do. Then the bottom of the funnel and this is real classic marketing talk. The bottom of the funnel is effectively retargeting. The people that are super close to converting with you but haven’t quite yet. When we’re using cookies, we can say something like, Jon has been to our site. 

Kris Tait: 12:18

We’re going to retarget him aggressively in the first three days, less aggressively for the next seven, and then even less aggressively for the next 30. What we do there is we lower how much we’re willing to pay for you. You can do loads of different optimizations within that. It’s the one that obviously returns the best marketing return on ad spend, or ROI. That’s obvious because it’s doing the easiest job. It’s just flipping people when they’re ready. 

Jon Krohn: 12:49

Nice. If retargeting is the last one, it’s the bottom of funnel where you have data that indicate that this person is very likely to convert and then you can optimize accordingly. Got it. 

Kris Tait: 13:02

That’s right. 

Jon Krohn: 13:04

Transitioning to the next point that you’ve already started to talk about, so with things like Apple allowing a lot more privacy for users, if they want it, cookies maybe are going to be gradually phased out in years to come, probably something that wasn’t anticipated at all just a couple of years ago, but now maybe something that’s going to happen. With fewer third party data on what other websites people have visited, maybe what other ads they’ve seen. What can your clients do? What can people who are advertising do? 

Kris Tait: 13:41

The big buzzword, the big trend in marketing, certainly in our industry, probably in yours too is first party data. What do we know about our current customers? What do we collect on them? What can we legally collect on them? I think I’ve been talking about… I do a predictions piece every year which I can point to later. For the last three years has been more investment in data quality. The reason I say that is because with this performance marketing world, we’ve been really spoiled. We’ve been really spoiled in the way that we can spend money on these platforms like Google and Facebook. Generally, they work incredibly well, and they are built to be successful. That’s how they keep increasing their media dollars. That’s how Google keeps increasing their stock price. They’re incredibly good. 

Kris Tait: 14:34

When you layer on, well, people are stopping the tracking. We’re not going to be able to know as much about these customers. Then we go into so how can we solve this for brands? Number one, they have to invest in their data quality. Basically, they have to have clean data internally. The ideal scenario is they have a technology which would have a single customer view of a customer within that portfolio. Then they can use that. We can go on to use cases of predictive modeling and the way that we might use data science on that dataset. Before we get to that, the other side is so we have less data on our customers. In a previous world, we could say we could control all of this manually. We could say we know age demographic. We know time of day. We know day of week. We know the weather today. We know what site you just visited. 

Kris Tait: 15:33

There’s quite a lot that we could optimize on. Maybe five or six clear data points that we could optimize on. The way that we would do that in the past is say this week, we’re going to look at all of the data. We’re going to say, we’re going to optimize the bids by X percent because we’ve seen people convert on a Tuesday more than a Monday. That’s our optimization for this week. Let’s do it. Let’s give it another week to bid in, the week after we do another one. We’d say age of people or gender of people, and we’d optimize that. It was quite a formulaic process to optimize media. Now, we’re leaning into machine learning and algorithms provided by Google and by Facebook and by other kinds of programmatic media buying platforms. Programmatic media buying is effectively just being able to buy in real time and being able to buy on platforms that you can execute very quickly, rather than, oh, can I buy this TV spot? Let’s plan for it for a few months. You’re going to buy that spot at 7:00 PM, et cetera, et cetera. We can do that very quickly on a programmatic media. 

Jon Krohn: 16:44

You talked about investment in data quality and having lots of first party data. Can you use tools provided by Google, Facebook, other programmatic options with your first party data?

Kris Tait: 16:57

100%. The key thing is if you feed bad data, you’re not in a good scenario there. That’s when you get suboptimal performance, if you put bad data in. If you do concentrate on that clustering of customers and having a really clean database around when people last purchased with you, how much they spent. This has been a focus for years within our environment. I think people have gotten quite lazy around there and haven’t kept their databases clean and easy to read. What ends up happening is that you go, “Can we have an email list of people that have purchased in the last 30 days?” That was quite easy to get for brands, but it’s not that advanced. It’s like so these people have purchased in the last 30 days. Let’s look for more of these people. Good. What happens there is we say, here are 3,000 people that purchased. Here you go, Google. Can you find more of them for us? 

Kris Tait: 18:06

They have an incredible machine learning algorithm that does that. It does it very effectively. What we’re saying now is okay, but what if we knew… Those people that purchased in the last 30 days, what if we could segment it further and say, well, these people purchased over $3,000 worth of product, these people between one and two, and these people under a thousand. Then well actually we don’t make that much money on these under 1,000 people. Why don’t we find more of the 3,000 people? That’s just a really simple way of thinking about it, but we can get much more advanced than what we’re feeding, Google and Facebook. 

Jon Krohn: 18:45

Cool. This episode is brought to you by SuperDataScience. Yes, our online membership platform for transitioning into data science and the namesake of the podcast itself. In the SuperDataScience platform, we recently launched our new 99-day data scientist study plan, a cheat sheet with a week-by-week instructions to get you started as a data scientist in as few as 15 weeks. Each week, you complete tasks in four categories. The first is super data science courses to become familiar with the technical foundations of data science. The second is hands-on projects to fill up your portfolio and showcase your knowledge in your job applications. The third is a career toolkit with actions to help you stand out in your job hunting. The fourth is additional curated resources, such as articles, books, and podcasts to expand your learning and stay up to date. To devise this curriculum, we sat down with some of the best data scientists, as well as many of our most successful students and came up with the ideal 99-day data scientists study plan to teach you everything you need to succeed. 

Jon Krohn: 19:53

You can skip the planning and simply focus on learning. We believe the program can be completed in 99 days, and we challenge you to do it. Are you ready? Go to www.superdatascience.com/challenge, download the 99-day study plan and use it with your super data science subscription to get started as a data scientist in under 100 days. Now, let’s get back to this amazing episode. 

Jon Krohn: 20:19

Somehow I didn’t know that you could do that with your own first party data. That makes a lot of sense to reiterate the importance of the data quality piece. That is something that has been getting a lot of press in the data science world lately. A lot of preeminent data scientists like Andrew Ng have been talking about how absolutely critical it is to have high quality data. For me personally, I just had this week, someone start, a data scientist who is 100% on improving data quality. Instead of having this data scientist focus on models that are in our platform, this data scientist is using models to improve data quality so that our downstream in platform models have richer, more accurate data to work with. Absolutely it is, you can’t possibly, no matter how good a machine learning algorithm is, you cannot have it perform well unless the quality of the data going into it as high. 

Kris Tait: 21:26

Definitely so. I think another thing to mention here is the really advanced best case scenario of what we could do with this is that we could actually use it in a real time basis to optimize based on lifetime value metric. Say we could, here’s one example that we’ve been working on recently. Working with a university to try and understand what the value of a application is to that university. The way that we’re sort of discovering and preparing that data is effectively looking at student records. They have a lot of data on student records, like application details, fees, margins dropout rates, et cetera, and then pairing that up with user behavior online. Like what have they done when they come to our website? How long did it take them to apply? Then obviously you have all of the demographic and data behind that, where they are in the country, et cetera. 

Kris Tait: 22:28

Once we’ve unified that data into a single customer view, then we calculate a lifetime value for every single student that comes through there. Once we’ve done that, we can then do a couple of things. We can do predictive modeling based on when a lead comes in, do we think it’s going to be valuable, or do we not think it’s going to be valuable based on 10 years of history? We can also do more clustering. We can say here’s a cluster of people that have applied to the six year program that all live on the west coast. They’re female. They’re in the top 10% of lifetime value. Then once we have that data, we can start using really advanced stuff, probably a little bit above my head in terms of the practicality of it, but effectively, simply it is pushing that data into online platforms and using the lifetime value to bid in real time. To say, we see someone searching on Google. Within 20 seconds, we can say, we think this type of lifetime value is down our bids. This is all done within the sort of machine learning algorithms that are afforded to us. Then we feed the data into that. 

Jon Krohn: 23:44

Nice. Just to clarify what lifetime value is, that’s the value… It’s the predicted value of a given person in terms of how much they’re likely to pay your company at any given time, or I guess over their entire lifetime? 

Kris Tait: 24:02

It’s a really good question because people define lifetime value differently, depending on what business you work in or how it’s been done in your business in the past. For the students at the university, it will be the value of that student for the lifetime of when they go to when they graduate. For e-commerce, people do things like, how much a customer would spend within a year, how much they would spend in two years. Effectively, it’s trying to work out a number that says if we know that number, we have much more power in terms of how much we can spend to acquire that person. We have much more… Basically that much more power about our budget. 

Kris Tait: 24:48

Whereas in the past, what people have done is say here’s an example. Someone converts in our website. We spent a hundred bucks to get that person to convert, and they spent 200 bucks with us. There’s a good return there, but what if they come back several times in the next year? It’s not getting factored into the calculation. We’re limiting how much we can spend per acquisition. This is a big topic of discussion in the marketing industry. I think a lot of what you do as a profession and the data scientists at iCloud can play a huge role for brands like powering their marketing teams.

Jon Krohn: 25:30

Nice. Digging into that a little bit more, you’ve talked about some of the big advantages of allowing machine learning algorithms to allocate budgets. Some of these things are obvious to me without having been working in marketing recently, myself, things like being able to in real time, adjust bids for a particular person based on demographic information, past website views, where that cookie information is available, maybe layering in this first party data that may be available as well. Basically, machines can be 24/7, 365 days a year be adjusting how much to spend, to get an ad or a Google search result presented to a given person based on their characteristics. Things like their lifetime values that you’ve described. That’s the obvious advantage to machines. Also just being able to use trillions and trillions of historical data points to be able to optimize allocate that spend efficiently, especially if you’re working with one of the big players like Google or Facebook. I can see tons of advantages to machines, but can humans still provide value anyway? 

Kris Tait: 26:51

Yes, of course. With automation in every industry, when people hear that word, it’s like how can we make these jobs more efficient? People think it’ll get rid of a truck driver or whatever it is, or a farmer if we can increase automation. That’s the same in marketing. We’re coming from a world where there’s a lot of manual hands-on keyboard work. Now, if you tell me well, I can do the same thing by setting something live and forgetting about it. That saves us a lot of time. Now, in the marketing industry people are time poor. It’s 24/7. It’s always on and there’s always, always work we can do. We have to stop somewhere. We have to have a life outside of work. If we didn’t have lives, there’s always something we can do. What my vision and what I think everyone is sort of thinking within this is that, we have a lot of smart and creative people that know marketing, that know the practice of marketing. We don’t want them spending all of their time in these platforms pushing buttons to optimize media. 

Kris Tait: 28:05

That is a much better job for a machine to do. What we’re going to do is we’re going to release those sort of people into more creative tasks, into more strategic tasks, into more what is this data telling us tasks? We might come on to sort of measurement later, but interpreting the results, that’s a huge… Thus machines haven’t got there yet. They haven’t got to the stage where they can do it and tell us what to do next. I think it’s like, we need to understand that data, understand what the business objectives are, and then make a decision on where to go next. That is some of the stuff that falls down and some of the stuff that people don’t have time to do right now because of the labor intensity of digital marketing. I think it will balance out. I think the overall industry, the productivity of it and the output of it will get much better over time as automation continues to roll out, which it already has in a major way. 

Jon Krohn: 29:04

Makes perfect sense. You folks at Croud have had an enormous amount of success lately. I know that you’ve had huge double digit growth year over year for many years now. The New York office that you lead, which is the whole US operations, has in particular seen a lot of growth in the last few years. What does Croud specifically do? How does Croud managed to get this balance right of using machines to optimize, like you’re saying to be getting all the nuances 24/7 taken care of to free up time for human creativity and strategy? 

Kris Tait: 29:48

Just in terms of what Croud does. We are a digital marketing agency. We’re a performance marketing agency. We’re a marketing agency. I think those terms are interchangeable, but we do build ourselves as a performance marketing agency. That means the channels that we run in are all of the Googles, the Facebooks, the programmatic media and display online media. We do a lot of app advertising for various finance brands, and Fintech brands, et cetera. We also do the gamut of creative and content services. One of our really, I think we’ve talked about it all day today, but the data solutions department, which is effectively data analytics, reporting and data visualization, data science and measurement and effectiveness is a really important growing department for us because as the automation continues to come in within our industry, some of those sort of media buying practices will be taken over more and more by algorithms and by machines buying that media. 

Kris Tait: 30:54

The skills that we can really help brands on are number one, what is our overall media plan? Where should we be spending that money? How much of it should we be spending on those different channels? Then how do we interpret all of this? I think the data solutions department where it’s like we’re visualizing this in a really good way. You can visualize stuff in a really bad way as well. It needs to be readable, and it needs to be actionable. Then also helping brands go through this environment of like, we’re getting less third party data on our customers. Our first party data isn’t great. How do we sort of transition into this new world? Croud has been a position where we can provide a lot of services across that brand. We don’t just do this media buying performance element. We can grow with brand. We can start with the media buying performance and then we can grow, and we can launch brands internationally. 

Kris Tait: 31:51

We can do every language and market in the world as well. It’s worth touching on. In this environment, it’s quite interesting because Croud has a model of sort of decentralizing our workforce. We have five hubs globally two in America, LA and New York, two in England, London and Shrewsbury, and one in Sydney in Australia. There are 230 people in those offices, but we also have 2,400 people everywhere. Anyway, it doesn’t matter where you are. We use those people through our technology to execute tasks for us. We might say, we need to analyze this data. We might say we need to drop this creative. We might say we need to… We might need this person for five hours a month on this specific skill set. We’ve created this… It was quite easy for us to transition into a work from home environment because we were already doing. It was 2,000 people. That has been incredibly useful for us. I think what Croud does is it helps brands, I guess, manage the media, but try and navigate this new world of how do we read it? What do we do from a reporting and optimization point of view, and how do we scale the company, both in the US and globally? 

Jon Krohn: 33:09

Nice. Then that explains the name Croud, so C-R-O-U-D. It’s a mix of cloud and having, I guess… If you have a big crowd, C-R-O-W-D in the cloud. 

Kris Tait: 33:25

Exactly. Pretty good name, actually. It’s fair play to the founders. I think that’s spot on. 

Jon Krohn: 33:31

That is, actually. I didn’t, until now, appreciate how great that name really is. Nice work founders, indeed. It seems clear that humans can provide a huge amount of value as you’ve outlined, even things like being able to define what lifetime value is, what that means for a particular customer. Obviously, no machine can do that. That requires a lot of thoughtfulness, a lot of working with the clients to figure out exactly what their needs are. That brings me to my next question, which is this is a data focus program. What are the best key performance indicators? What are the best marketing metrics or general strategies that people can be following when they’re trying to market their product? 

Kris Tait: 34:19

This is a really good question. Just to provide some context, what has happened in this sort of performance marketing environment is everything is trackable. We can track how long you’ve been on a website, how you moved through that website, where your mouse went. Did you click on this? Did you click on that? The risk of that is that you get way too distracted by every single metric. What we recommend is that we focus on a core KPI. Ideally there needs to be a business challenge. What is the business trying to solve? There is something that we’re trying to solve. How can marketing help you solve that? There needs to be a business objective, a marketing objective. The marketing objective should be very, it should be measurable. 

Kris Tait: 35:14

It should be time-bound, and there should be a sort of before and after. Here’s a good core KPI for our business. We would like to increase our revenue 30% from 1 million to 1.3 million in FY 21. You’ve got the, what you want to increase revenue, by how much, from what to what and when. That is a very, very good core KPI for business. What tends to happen is that people concentrate too much on the, what’s the click-through rate of our ad? How much did we pay for that click? Where do we rank on Google when someone searches this? These are all really good in-flight metrics. They’re good to sort of diagnose what’s going on. People take them incredibly seriously and change their marketing strategy based on them sometimes, which is not a good thing to do. Our sort of signal is don’t get too sidetracked by the things that you can track every minute. Make sure that you’re setting up your business for success with like a core KPI that can feed down. 

Kris Tait: 36:23

When we get this core KPI from a business, and then they say, “Hey guys, can you launch this campaign on Friday evening? We need to do it really quickly.” We ask, how does this relate to your core KPI? If it doesn’t, why are we all wasting time doing it? That’s one of the key things that I think. Stop looking at everything. Look at the things that are important. 

Jon Krohn: 36:49

Nice. That makes sense. I guess to summarize, it’s have probably fewer metrics so you’re not trying to follow tons of different kinds of metrics that end up leading you all over the place. As much as possible, have a small set of key performance indicators that are directly tied to a business objective. As you mentioned, they are time bound. They’re relevant obviously to the business objectives. That’s the, what [inaudible 00:37:20] talked about the when the time down is, and then how much, so we have this quantifiable target. That makes perfect sense to me. With all of the data available, as you’re saying, you can end up probably chasing yourself around in circles if every day you’re reacting to different changes. They might not be meaningful changes, and so you end up wasting a lot of time and effort and ultimately maybe even counteracting some progress that you’ve been making. 

Kris Tait: 37:55

Yeah. Think about a business when the CEO has some objectives, the CMO, your director, and your direct report. They will have different KPIs. That’s 20 KPIs. What do you do? It’s a really challenging thing because sometimes the director hasn’t got the purview of the business objective that working down here within that kind of little bubble within the business. We try and challenge people to say, okay, but has your boss told you what those business objectives are and how is your marketing campaign going to help towards that? A lot of the time it’s like, well, we just need to deliver business performance. It doesn’t mean anything. We need it to be quantifiable, and we need it to be clear so that when we’re executing marketing campaigns, they are going towards that business growth, that metric. 

Jon Krohn: 38:53

Nice. Makes perfect sense. Thank you for illuminating this all, Kris. Clearly you and Croud are doing really interesting things. You’ve got a unique and a place in the market. You’re obviously having a huge amount of success with that. You’re getting the right balance of machines and humans. Are you hiring right now? What roles are you hiring for? If you are, what do you look for in the people that you hire? 

Kris Tait: 39:23

We are hiring. We are hiring across every department in New York, which is great news. It’s a great time. It’s a great time to be in digital marketing. It’s super exciting. A lot of media investment is moving from more traditional places like TV and radio into digital, and even connected TV. We all have Hulu. We all have Roku. We can buy ads on that. The places that we’re hiring are all across the biddable media team, and that is operational people that are hands on keyboard or telling the algorithms what to do. Also, within the data analyst’s sort of role as well, digital analytic’s roles. One of the ones that might be interesting to your audience as well is we are looking for a data strategy director that will lead that data solutions department within New York, work very closely with the UK, and look after data visualization, data assigned teams, predictive modeling teams. They don’t need to be specifically a practitioner, but they will be overseeing the sort of forward looking strategy for Croud’s data solution in New York. 

Jon Krohn: 40:32

Nice. Perfect. You’ve told us a lot of hiring, particularly that last one sounds very interesting the director of data solutions, if I remember that correctly. What do you look for? In somebody that is say this director of data solutions, when someone comes in or you see their resume, what makes you feel like, yeah, this is definitely the right fit for us? 

Kris Tait: 40:57

Definitely a good question. We look for people that are interested in building something. We are an agency that are still quite young in New York in the context of this market. We offer equity, and we want people to feel like they’re building something that they’re proud of and that they can… They’re having a great story about their career. We’re not the agency that is uniform. We’re not the agency that is 20 years old and has every single process worked out. We’re trying to create something that’s new. I think someone that’s open-minded to like this model of, yes. We have 200 people in the offices that work with clients and speak to clients. We also have a huge workforce behind you. The benefit of that for an employee is that they get to spend more time on the interesting stuff and less time on all of the manual work within the platform. People that get that love Croud. People that don’t get that they don’t, and that’s fine. It’s totally fine if you don’t. That’s what attracted me to here, and I’ve been here nine years. It works. 

Jon Krohn: 42:06

Basically, relative to other marketing agencies, there’s an opportunity to really move up the value chain. 

Kris Tait: 42:13

Definitely so. 

Jon Krohn: 42:15

Their strategy and kind of thing. Crazy you might be in a lightning storm too. I’ve never been in a lightning storm while I’m filming a podcast. That was a- 

Kris Tait: 42:22

Oh my God. 

Jon Krohn: 42:22

I’m noticing. I could even see… Sometimes, if you’re watching the YouTube version of this, my whole face goes blue because of the lightening that’s crashing right outside my window. 

Kris Tait: 42:30

That is insane. 

Jon Krohn: 42:37

Cool. The last key question that I have related to the marketing work that you do is what kinds of tools do you or your team use at Croud day to day that allow you to take advantage of data and maybe even some modeling? 

Kris Tait: 42:53

Good question. I once got asked this question, slightly different question. What’s your favorite channel within Croud? My answer was analytics because it powers everything, what we do. We’ve obviously talked about that today. Google Analytics is one of the most used tools in our repertoire, analyzing site data, analyzing user and audience behavior on a website, analyzing conversion data and revenue, super powerful. That’s a free tool to use this. When you start to get over a million visits a month, you need to pay for the premium version to get all of the data. There’s a long way to go to that for most businesses. Alongside that, we use the Google Marketing Platform, which is, it’s a stack of tools that you could build your whole business on from a marketing perspective, from a tracking perspective, and from a measurement perspective. You have things like Google ads, Google Analytics, DV360, which is a programmatic DSP where we can go and buy media from New York Times or Wall Street Journal. We can buy it programmatically in real time. 

Jon Krohn: 44:03

Just to jump in there quick to let people know what a DSP is. It’s a demand side marketplace, right? 

Kris Tait: 44:07

Yeah, demand side platform. 

Jon Krohn: 44:08

Tell us a bit about why that’s relevant in marketing. 

Kris Tait: 44:13

It’s relevant. We can use a demand side platform to connect to loads of different exchanges of where inventory is. When I say inventory that is, relevant inventory on a website. When we go onto the New York Times, for example, and you see a display ad which might be a retargeting ad, or it might be a prospecting ad, we can see that inventory within the DSP and we can buy it. We can buy it saying, we want to buy it here, or we can let an algorithm buy it as well. That basically that DSP would plug into thousands of those exchanges. Maybe it’s a gaming website, maybe it’s YouTube, maybe it’s whatever it is. The benefit of it is that we can buy all of that media in one place, rather than going to the New York Times and going to the Wall Street Journal and the tracking is everywhere. We can buy it in one place and we can track it in one place. 

Jon Krohn: 45:11

In addition to buying and tracking, that probably also affords you much more breadth and potential with what you’re doing with third party data, first party data, machine learning models. If you, in one place have access to thousands of major websites with lots and lots and lots of ad inventory, then you can much more efficiently allocate your purchase of that ad inventory than if you were constraining yourself to just the New York Times. 

Kris Tait: 45:38

100% and where we’ve got a really good example of that, where we’ve used a custom bidding algorithm, which we created like a custom script which effectively gave value to each impression. An impression is when I see an advert on a website. That’s an impression. We give a value to an impression based on who the audience was. We uploaded that script and the script… I think I’m just looking at the data now. The script outperformed the standard algorithm, Google standard algorithm by 350%. 

Jon Krohn: 46:15

You guys built that in house. Your data science team created a model, and then you can upload the model into the Google Marketplace platform into things like the DV360 demand side platform. 

Kris Tait: 46:28

Exactly. It goes into DV360 and then it uses that script to buy and change bids, based on some of the lifetime value and audience data that the brand had. This was for a financial website trying to get subscriptions, so super powerful. The other tools that it’s worth just mentioning, because these things might not be things that your audience are sort of aware of, so we do a lot of planning work and that sort of media planning and audience planning work. What you can do is you can use platforms and tools like GlobalWebIndex and another one, which is a surveying tool. They’re doing surveys constantly, and they’re asking people questions. You can go in and look at all of that data around what platforms are people using? Which medication are people using? There’s a lot of stuff in there. It’s all anonymous, of course, but we can see people who go to physiotherapy elicit these kind of behaviors online, really, really clever stuff. 

Kris Tait: 47:31

The last one is Similarweb, which is a platform, a tool that we can look at competitor data. We can see a website, and we can see what traffic goes to that website. By the way, this is when we don’t have access to their platforms like their Google Analytics. It’s incredibly good for looking at a competitor and understand how they’re advertising in the market. What creative they’re doing. How their traffic makeup is. They’re really important tools. The data science team uses Python, uses R and all of the sort of databasing tools that you guys are used to. I’m definitely not an expert in that, but. 

Jon Krohn: 48:13

Nice. Makes perfect sense, Kris. That’s a really nice rundown. I learned a few things there. I didn’t know about GlobalWebIndex or Similarweb, or DV360. It’s really nice to know that. There was a time many years ago when I worked in digital marketing, but the ecosystem changes so quickly that I am easily out of date. 

Kris Tait: 48:35

It does. 

Jon Krohn: 48:36

Nice. Loved everything that you’ve told us about in this episode, so far it’s clear that there’s a huge amount of potential in performance marketing, in being thoughtful about your KPIs. Not tracking too many KPIs. Not being reactive, using the kinds of tools that you just mentioned. To ask you one last business related question, what kinds of clients can benefit most from working with you guys directly or working with a similar kind of approach? 

Kris Tait: 49:11

Good question. We can work with any client, and we do. We work with a very varied range of clients. We’ve worked with a lot of retail e-commerce DTC brands. I think the reason for that is it lends itself very well to what we do. Of course- 

Jon Krohn: 49:30

DTC, in case the audience member isn’t obvious is direct to consumer. It seems like venture capital firms really love direct to consumer companies. You mentioned just before we started recording that it seems every day in just New York, I think there’s a new direct to consumer company every day. It allows in this highly digital marketing world. Where I think there’s a lot of tools that allow for seamless distribution and all these kinds of things, it allows a company to pick a very specific niche. I think this is another thing you mentioned just before recording. I’m stealing your thunder here. Things like you can be like, we’re going to make a razor that is better or cheaper, and we’re going to market it directly to customers. We’re not going to do it via a store, a third-party store, which was for most of the last century, the primary way of getting a product out there. 

Jon Krohn: 50:28

Now you can cut out the middleman. You can go direct to a consumer. Anyway, I’ll stop stealing your thunder. You can fill this us in a bit more about it and why that’s a particularly good opportunity for the performance marketing that you do. 

Kris Tait: 50:41

Well, direct to consumer in nature is really good because they’re speaking directly to their consumers and then not having to go through a… It’s not like a P&G where they’re selling in a CVS or in a Walmart. They don’t have any direct relationship. That is a very powerful when we’re talking about performance marketing and first party data, of course. These companies build, or they build their businesses on knowing that they can input this and get out this. Harry’s is a very good example that you sort of alluded to their shaving company. They’ve just created a razor that competes with Gillette, a very, very old traditional brand and sold the company for 1.4 billion. It’s a razor. They’ve obviously created a lot of different… I think I bought some Harry’s clay yesterday for my hair. They’ve obviously gone into a lot of different products. We work with a lot of those brands because the performance marketing works very well from a Facebook perspective, from a Google perspective. 

Kris Tait: 51:44

They also, if you think about an e-commerce brand, they can be on social. They can be on TikTok. They can be on Facebook. They can be on Instagram. They can be on Snapchat. They can be on Google. They can be everywhere, and they can be on affiliate websites. They could be on Alibaba in China. They could be on any platform. With retail brands, they’re quite complex, because they have a lot of different products, a lot of different pricing they have stopped to worry about. That just means it’s more advanced in terms of what they need from an agency like us. Therefore, we can provide more value or a lot of value versus a business to business website that has one conversion action, get an inquiry. That is infinitely more easy and scalable to just do, rather than all of the complex nature of DTC brands. DTC brands have been great for us. We work with some really good ones in New York. One called Burrow, which is a sofa company, which you know well, Jon. I think you have one. 

Jon Krohn: 52:50

I do. 

Kris Tait: 52:52

That was not because I recommended them, but they are great. They are scaling into loads of different kind of niches within the home. They’re becoming more of a lifestyle brand now. Those kinds of clients are great for us, but we can work with anyone based on what we do. 

Jon Krohn: 53:09

Beautiful. All right, so starting to wrap up, Kris, do you have a book recommendation for us? 

Kris Tait: 53:15

A book recommendation, so I actually recently wrote seven lessons for seven years in New York post, which went down quite well. I mentioned a couple of books in it. The one that really, really impacted me, and I thought it was fantastic is one The Sales Acceleration Formula. I started out in sales, and I’ve got a real passion for sales. I’m also probably not someone that you think about when you think about a sales person, because I really like the sort of data-driven approach to it. It’s all about input, output numbers and convincing people. The book is called The Sales Acceleration Formula. It’s by a guy called Mark Roberts sorry, Mark Roberge. He was the head of acquisition, the head of demand generation for HubSpot, which is a marketing platform that quite a lot of people will probably know. It’s how he took it from zero to 100 million in everything that he did. 

Kris Tait: 54:21

When I first landed in New York, 2015, I was like where do I start? That book really helped me. It was fantastic. Everyone that I’ve recommended it to has really enjoyed it. Quick read and it will help you put in your mind how you could start a company and scale it. 

Jon Krohn: 54:39

Beautiful, that’s a great practical recommendation. Kris, so much insight that you provided for us in this episode, so much potential for working with you, either by being hired by you or being a client of yours. How can people get in touch? 

Kris Tait: 54:58

I’m pretty active on LinkedIn. I’m relatively active on Twitter. You can connect with me on LinkedIn for sure. You can follow me on Twitter and see what I’m complaining about or absolutely loving, probably soccer, football and technology. You can drop me an email as well. I’m sure we can leave our email and details in your notes. 

Jon Krohn: 55:21

Nice. Absolutely. I will be sure that we have all three of those, your LinkedIn, your Twitter, and your email address in the show notes. Kris, thank you so much for being on the program. This was such a great practical episode on marketing analytics. Really appreciate you taking the time. 

Kris Tait: 55:36

Thanks Jon. Thanks everyone. 

Jon Krohn: 55:43

I love what a cool composed and articulate speaker Kris is. Not too many years ago, I worked as a data scientist in the digital marketing world, but it’s amazing how quickly the marketing ecosystem has evolved. There was a ton for me to learn from Kris during filming of the episode. Kris filled us in on what performance marketing is, including the advantages of a diversified multi-channel strategy. How privacy measures are beginning to shift marketing power from third-party data sources like Google and Facebook toward first party data on your customers, as well as how machine learning can make the most of this new paradigm. We talked about how algorithms surpass human capabilities at optimizing advertising campaigns, but also free up human time for a higher value activities around creativity, drawing inferences and strategizing. 

Jon Krohn: 56:32

We talked about how refined, quantifiable and time bound marketing metrics are the surest path to commercial success. We talked about the huge breadth of data roles in modern performance marketing agencies like Croud across specializations in analytics, visualization, data science and data solutions. As always, you can get all the show notes, including the transcript for this episode, the video recording, any materials mentioned on the show and the URLs for Kris’s LinkedIn profile and his Twitter account, as well as his email address and my own social media profiles at www.superdatascience.com/481. That’s www.superdatascience.com/481. If you enjoyed this episode, I’d of course greatly appreciate it if you left a review on your favorite podcasting app or on the SuperDataScience YouTube channel, where we have a video version of this episode. 

Jon Krohn: 57:24

To let me know your thoughts on the episode, please do feel welcome to add me on LinkedIn or Twitter, and then tag me in a post to let me know your thoughts on this episode. Your feedback is invaluable for figuring out what topics we should cover next. Since this is a free podcast, if you’re looking for a freeway to help me out, I’d be very grateful if you left a rating of my book, Deep Learning Illustrated on Amazon or Goodreads. If you gave videos on my YouTube channel a thumbs up or subscribe to my free content, rich newsletter on jonkrohn.com. To support the SuperDataScience company that kindly funds the management, the editing, and the production of this podcast without any annoying third-party ads, you could create a free login to their learning platform at www.superdatascience.com, or consider buying a usually pretty darn cheap Udemy course published by Ligency, an affiliate of a SuperDataScience such as my Mathematical Foundations of Machine Learning course. 

Jon Krohn: 58:25

Thanks to Ivana, Jaime, Mario and JP on the SuperDataScience team for managing and producing another amazing episode today. Keep on rocking it out there, folks, and I’m looking forward to enjoying another round of the SuperDataScience Podcast with you very soon.