Interviewer: Today I am very pleased to welcome Brian Kardon the CMO of Lattice Engines. Brian is responsible for Lattice’s market positioning, demand generation, thought leadership and integrated marketing. So, you might ask, what does Lattice Engines do? Lattice is pioneering the predictive applications for marketing and sales. Lattice helps companies win more customers by applying sophisticated data science in a suite of easy to use Cloud applications. So, welcome Brian.
Brian: Thanks Glenn I appreciate that.
Interviewer: I’m fascinated by this whole topic of predictive analytics. When you and I were speaking earlier you said it feels a little bit like the marketing automation world did a few years ago.
Interviewer: Tell us a little bit about that.
Brian: I was very early into marketing automation. I was the CMO at Forrester Research in 2004 and we were struggling with our lead management process, and stages, and conversions. One of the analysts on my team told me about this little company in Canada called Eloqua, that’s how they pronounced, Eloqua. So, we installed it. We bought the software and it really transformed what we did, and for Forrester the stock price tripled, which was great for me. So, the business grew, win rates improved, quotas improved and then in 2008 I got a call to be the CMO of Eloqua, which was great. So, I’ve seen first-hand the early adopters of marketing automation and they got a real step up in improvements to performance, and conversions, and win rates and so on, and deal size, and velocity. Now we’re seeing that the adoption rates for marketing automation are hovering around 50-70% depending on whether it’s a big company or small company. But, most companies now have marketing automation and marketers are asking what’s next. How do we get the next step improvement in our performance? And we’re seeing that companies like DocuSign and Mindjet and GE and HP, Adobe, a bunch of cool, interesting companies in different sectors are now embracing predictive analytics and predictive marketing. So, that’s why it kind of feels like the early days of marketing automation. The same people, the early adopters of Eloqua or Marketo were a part of that [00:03:23 – Inaudible] are the early adopters now of predictive marketing.
Interviewer: Do me a favor for our audience and define predictive analytics in marketing.
Brian: It’s so funny. First when we talk about predictive analytics it’s all around us and I think people are saying what is this thing. For example, in health care, predictive analytics can help predict which patients are at risk of developing certain conditions. What it does is it looks at patient’s history over millions of records, your cholesterol; your glucose; family history and it can predict diseases. Fraud detection, so credit card companies use it all the time. So, if Glenn suddenly buys 10 flat screen TV’s in Tijuana they would see that that’s probably not something that you do all the time; there is pattern recognition; and they know that the transaction is likely to be fraudulent. It’s used in cable TV. So, a friend of mine works in Comcast and she runs a group that takes customers that are very unhappy and want to disconnect, want to stop the service, and her job is to try to save them by making offers to prevent them from terminating service. I asked her, I said you give $20 away or a free package. She said, no, no, we have an analytics thing that tells me what to offer that will optimize the chances of them renewing. I said, give me an example. She said well if they live in an affluent area and we know what they bought before and what they like to watch, we actually know their viewing habits, we’ll offer them maybe a sports package, or free HBO, Showtime, or two months free. So, they tailor all of those things, so that’s predictive analytics in work. Now, it’s being applied to marketing, so the basic use case is you have lots and lots of prospects and you call them, but can you actually know which ones are most likely to buy, because those are the ones that you ought to prioritize.
Brian: Predictive analytics can tell you where to prioritize and where to put your efforts and where to ignore.
Interviewer: Right because we always want to know who are the most valuable prospects either outside of the pipeline or in the pipeline itself.
Brian: Exactly, now your phrase valuable is really interesting because we all have lead scoring using marketing automation for that, but that’s guessing. So, we’ve all been at a table with marketing sales people where we guess and we say, hey someone goes to the webinar give them 20 points; someone goes to the career page subtract 8 points, they’re looking for a job; someone goes to the pricing page call them immediately. So, we think that is exact science, but in fact, it’s all judgments, and its people basing it on the judgments. Now, there is enough data that we can use data science to actually know which behaviors, which characteristics, which demographics are those that are most likely to buy.
Interviewer: You used an example of activity on a website; can you share some other examples outside of websites directly?
Brian: Yeah, there are tons of things. For example, in the case of Juniper Networks, Juniper was one of our customers and they sell switches and routers. I don’t really know what switches and routers are, but that’s what they sell. We looked at about 2000 attributes and the one attribute that was very interesting that they didn’t know about was if someone just signed a lease to move to a new office they’re 20 times more likely to need switches and routers.
Brian: Exactly, wow. So, of course, they were never capturing that information, but you can crawl through websites legal, real estate information. You can find people who have just signed leases to move. So, in the case of Juniper they’re very clever, they set-up an automatic alert to the sales person who has got that account to say, hey, give Glenn a call now and congratulate him; he just signed a lease and is going to move in 90-days. Then the marketing team also has trigger that if someone just signed a lease they send them an email congratulating them on the move and showing a little video of what they have to offer. So, Juniper may be looking at things like your title, and your revenues, and your industry, but in fact data science proves that that’s just noise. The attribute that is most predictive is whether someone just signed a lease.
Interviewer: Tell us a little bit about – how do you find that. I think of that as data mining. Is that what we’re talking about here?
Brian: Exactly it’s data mining, but a lot of it is hiding in plain sight. So, we have crawlers that are going through a hundred fifty million websites every day pulling information. It could be sites like Career Builder, or Monster, or Indeed it could be real estate sites, or government sites about patents. It’s also company websites. It’s LinkedIn. So, that’s one piece of just going through websites. The other piece of the data picture is we have relationships with about 35 data providers. So, we’re getting credit reports from Experian. So, for example, you may think you have a really hot prospect and you’re about to call them, but if you knew that their credit score just dropped from A to C you probably wouldn’t call, because they have no money.
Interviewer: Um hm, right, right.
Brian: So, it’s all those kinds of things. So, it’s credit reports, Lexis/Nexis, a company called Spiceworks, [00:08:14 – Inaudible] Technology, so if you’re selling marketing technology it helps to know if you’re using Salesforce, or Marketo, or [00:08:23 – Inaudible] or any number of things. So, the data is generally available, but it’s really not efficient for any one company to collect it. So, Juniper is not going to go out and be collecting all these real estate leases information. We’re doing it on behalf of our hundreds of customers around the world, so we do it at scale, and then we’re able to build models and tell them which accounts are going to close.
Interviewer: Do we need data scientists to make all this work?
Brian: That’s a great question. There’s been a huge surge in hiring activity of data scientists, sometimes it’s called they have PhD’s in machine learning, data science. It’s funny, the largest employers of these are things like the NSA, and Langley, the CIA, and Google is a big hirer of data scientists. Also, I was with friends at LinkedIn a few weeks ago and I said how many data scientists do you have at LinkedIn? They said we have 60 full-time PhD’s in data science.
Brian: So, they’re all over the place, but what’s happened now is we can put all of that data science in a box and create software, so you don’t have to hire a PhD and have a top school to do it. It’s been automated now. All you really need is really you can use Lattice to do this. We have all the mathematics kind of in a box. Sometimes the math is kind of tricky, but you don’t have to be a data scientist to do it anymore. So, it used to be the domain of these data scientists, but software makes even the most exotic techniques within sight. So, its things like some of us learned in school like linear regression, multi [00:09:53 – Inaudible] regression, and things like regression trees, conditional mutual algorithms, random forest neural networks. A lot of very advanced math is used, but it’s a little bit like Excel, you don’t have to be a genius to understand how all these things work, but Excel has it in thee for you, or SPSS, any of these statistical packages you could buy right now. You don’t have to actually know what an [00:10:19 – Inaudible] value is, or what an R square is, or a Pi square is, you can actually just put your data in and then in common language ask it to do something and it will do it for you.
Interviewer: And that’s assuming that what the marketer is using is designed for a marketer to use. Let’s talk a little bit about that. We talked earlier about predictive campaigns and mass customization. Tell us more about that.
Brian: Yeah, so, very often marketers have campaigns that go to everyone, but if you look through historical data you can actually determine through data science which campaigns are most effective across which customers. It could be by industry, or title, or even your past engagement rates. So, let’s say a simple one could be Glenn has been getting a lot of campaigns for the company, but the only ones you engage with are the ones that are videos. So then, the systems are very smart and know to only push you videos, because those are the highest level of engagement, or the system may know where you are in a certain stage because it’s looking at your CRM, the company CRM, and it knows that you’re actually in negotiating stage. So, maybe it sends you information relevant to that. So, it’s a way to sort of automate the communication based on data science and knowing where you are in stages of your buying cycle. That’s sort of predictive campaign management. Also things like predictive renewal, so if you have a subscription business it would be great to know which of your customers are going to renew and which ones are not going to renew. So, we’re seeing companies use predictive analytics to actually predict who is going to terminate and they can take action long before the contract comes us, so it could be factors like they’ve used software for a while; they’ve complained on the help desk; they’re in a certain industry that is a high rate of churn. So, a lot of subscription companies, SaaS companies, software companies, magazine companies, anyone that’s got like an annual subscription is very conscious of their renewal rate. Using data science to predict the renewers and then focusing your efforts on those who are likely to [00:12:16 – Inaudible]. Just like Comcast, they know who is likely to [00:12:19 – Inaudible] based on data usage, where you live. They know that the churn is higher in certain neighborhoods than others. They know what moving rates are. They know what credit scores are for different families, and so they’re able to understand who is going to [00:12:31 – Inaudible] and who is going to renew and they can understand that and offer packages to renew.
Interviewer: Can you share an example with us of predictive analytics before someone enters the sales funnel; before you know who they are at a name level?
Brian: It usually happens at an account level, not a name level. There is not that much information at the Glenn Gow name level, but there is at Crimson Marketing. So, at the account level I’ll give you a good example. We have a few customers. They give us a list of their current customers; let’s say a couple hundred names. We’re able to find companies that look like those companies that have the same characteristics. So, this is a company that isn’t even in our pipeline, that isn’t in your database, we can find more like that. They could be competitors of your current customers. They could be people in the same industry or have certain characteristics. So, we can start to identify accounts that are likely to buy based on the characteristics of people who have previously bought from you.
Interviewer: Right, right, so looking at history and trying to find matches of common companies that look like that.
Brian: Right, but it’s not a human being looking at it. It’s using mathematical techniques and things that are very valid statistically. I think in the past there has been a lot of – I know you guys talk about Moneyball sometimes and marketing moving more towards science and towards – you know, the kid’s got a good grip on the bat or something, but most marketing and sales organizations there is sort of this tribal “knowledge”. I use knowledge in quotes where they think they know what is a good lead. The loudest guy in the room says I only care about C-Suite. I don’t care about any managers or directors, it’s got to be C-Suite, but in fact you look at the data and it proves otherwise; or they say healthcare, it’s all about the healthcare industry. But, in fact, you see your conversion rates are financial services are four times that. So, successful companies are trying to trust the data and don’t rely so much on tribal instincts. Something we do that’s very interesting is after a customer signs up we’ll have a workshop with the customer and we’ll say what do you think are the characteristics or the attributes of the ideal customer? They’ll list a whole bunch of things, the title, the industry and they went to the website and looked at this. Then we say, that’s very interesting; here is what the data shows us, and in many cases it validates some of those things, but in many cases it says you’re exactly wrong. The classic one is the C-Suite. They get very excited about targeting C-level people, but in fact, C-level people very often approve something later but not engage in the process.
Brian: We had a customer who would never look at someone with a very low title. We found in big companies in particular that a senior person would ask a junior person to do all the research.
Interviewer: That’s right.
Brian: In smaller companies you do care about the C-level because they’re more hands-on than in larger companies, so we set-up the algorithms that way. So, in larger companies we would engage with a lower level title, but in smaller companies we would [00:15:37 – Inaudible] the higher title, so those kinds of nuances.
Interviewer: That’s an interesting conclusion. We’ve validated that through research, in addition to predictive analytics like you have. When a company is aiming for the C-Suite, people they rely on the most to inform them are their staff.
Interviewer: So, that is not the C-Suite.
Interviewer: Often, you’re right, companies don’t want to target there because they don’t feel that they are the decision makers, and yet, they are the influencers.
Brian: Glenn, think about all the wasted time of all these dialers, these business development reps calling because the head of sales said get the C person on the phone. Well, you can’t get them on the phone. I feel sorry. All this opportunity costs, all these hundreds of hours are wasted trying to get some senior person. I get calls all the time, probably like you do, sales reps trying to sell me something. I don’t buy that much. As it happens I love to listen to the calls because I’m a bit of an anthropologist. I take great pleasure in listening to their pitches. I hear horrible ones, I hear good ones, I hear occasionally a great one, so I take notes on how they do it. So, it’s very interesting, but I just find it very funny that – I said why are you calling me? They said you’re the C-level, you’re the decision maker. I said, not really (Laughter).
Interviewer: That’s right.
Brian: But, they’ve been instructed to do that.
Interviewer: Brian you had suggested you had a couple of other very quick stories for us. I want to give you a chance to pull those out.
Brian: Yeah, so another customer is interesting and it’s Autodesk. Autodesk sells CAD/CAM software, and they were using traditional lead scoring and it just wasn’t improving their conversion rates. They came on board, we did the analysis and, as it happened, any company that is hiring design engineers and if the phrase CAD/CAM is in the job description are 10 times more likely to need more Autodesk licenses in the next 90-days. So, what we’re crawling through is Monster, Career Builder, Indeed, other company websites and we’re crawling and harvesting all that data. Who has posted job descriptions? Is the word design engineer in the job description, and is the word CAD/CAM in the job description? Then we’re immediately letting the sales people know that these are good targets because they’re hiring new design engineers.
Interviewer: Right, right.
Brian: That’s pretty dramatically improved Autodesk’s conversion rates, because before they were calling a bunch of companies and they would say, I don’t need any more licenses, I’m not hiring.
Interviewer: Right, right.
Brian: So, it seems very basic, but it’s actually the companies are not looking at these things, nor do they know which are attributes to look at. One sort of final example from another industry is Staples. Staples is the office supply company and we know them for the big box stores, but they also have a B2B team. They have 1000 phone reps and they’re calling small companies and two years ago they would call a small company and say, hey this is Staples, need anything this week. The customer would say no I’m okay. Okay thank you I’ll call you next week. Now they call and they say something like, hey Glenn we know you’re hiring new people, you’re going to hire four people so you ought to buy desks and chairs now because it’ll take a couple of weeks to deliver. You say, oh great thank you, and by the way Glenn we know you opened up an office in Austin, Texas. We can help you with signage, getting a printer in there and stationary and filling up the supply closet. Terrific. Or, it could be something like hey Glenn you probably need more paper. You say, well how do you know that? Well, you’ve been buying paper every 90-days and 200 days have gone by and you haven’t bought paper, so you’re probably running low, why don’t I put that on the order too. So, suddenly the sales rep at Staples is not flying blind asking the customer what they need, but actually suggesting and being proactive using big data. They know if they’re hiring people, they knew that the customer just opened up and office in Austin, Texas, they know that they’re running low of paper. It’s all these kinds of things. Suddenly the order size goes from $100 for Staples to $2000, and so we’re seeing much higher conversion rates and much higher checkout rates from Staples because of big data and predictive analytics. That’s a great story, so not only do I know a lot more, I’m a lot more educated when I reach out to that potential customer and I can be more helpful in my communication with them as a sales rep, but in addition, there are opportunities where I may know more than the customer to use your example of you need more paper. That’s really powerful.
Brian: Yes, very, very powerful. And you know these things are not in the future. It’s happening now. In the consumer world we see this all the time, so Netflix, they tell us what movies you should watch next. I find it amazingly accurate, so they upsell there. Amazon is the king of collaborative filtering it’s the heart – Amazon last year did about $90 billion in revenue, a third of it came from the recommendation end. They’re saying, hey Glenn, you probably need a new pair of sneakers, or a tennis racket, or have you thought about tenor saxophone reeds, or whatever you like to do. So, they’re getting more and more business. What’s interesting about these systems is that the more you buy, the more you look, the smarter they are, they know more about you. So, that’s what’s happening in marketing. The more customers that buy, the more accurate the predictions. Over time, the Amazon recommendation engine, and of course Google is the largest interest engine in the world and it’s all personalized based. If I type in clubs in Boston it knows that I like jazz, so it’s going to recommend jazz clubs, not dance clubs for example, because it knows my history. The more I search for it gets smarter and smarter.
Interviewer: Or golf clubs.
Brian: Or golf clubs exactly right. Fascinating, or a golf club that’s a great example. It knows that you play golf or I don’t and it will recommend things based on your history that you searched for golf clubs over and over again, good point, that’s great.
Interviewer: So, last thing here. Tell us a little bit about the future. What do you think we’re going to see in the next one or two years Brian?
Brian: I think we’re going to see the world of marketing really move to more automation and data science and less about anecdotes. I think the best analogy is what happened on Wall Street about 20 years ago. I can assure you I’m no fan of Wall Street, but 20 years ago human beings were looking at screens and they were making trades based on what they saw on different screens on their desk. It could be five screens, ten screens, but the data was coming at them so fast between the Dow and the Hang Seng and Nikkei and foreign exchange and options and interest rates, you couldn’t keep up with it. Now who’s one on Wall Street? Well, it’s the guys who wrote all the algorithms. It’s the programmatic trading. They wrote all the math, and not only does the math tell you what trade to make, very often the math will make the trade for you. I think this is happening more in marketing. So, rather than every month, or every week, the marketing get together and looks at the data and says, hm this campaign was good let’s do more of that; or this list didn’t perform very well, so then it wasn’t very good. It seems to be very episodic how marketers look at it. I think it’s going to be automated and the machines will start to read the information and optimize it better and better. Just like your example with Google and golf clubs you know. Like, there wasn’t a human being that actually looked at your search history that said wow Glenn’s a golfer so clubs means golf clubs. It’s like a computer that’s smart enough to recognize that. So, I see computerization just merging as a very big force in marketing and algorithms and automation much more.
Interviewer: Well I think that’s good news if we wrap everything you just said into maybe as a marketer I don’t need to be a data scientist if the software that I have available to me can do that for me, and in many cases it can do it automatically. That’s makes for a very exciting future.
Brian: It is an exciting future. I will say that it just makes creativity even more important. So, great marketers still have to have great messaging, great campaigns, great targeting, but you’re really assisted by all the computerization. I love Watson, which is the IBM product that actually beat Ken Jennings in Jeopardy. And computers now, big blue was able to – a computer program was able to beat the chess masters. So, I see the world of automation, computerization, algorithms just starting to embrace marketing and sales and it’s going to be just a great force to improve conversions, and we’re starting to see the early adopters, like the companies I mentioned, Staples, Autodesk, Bank of America, Juniper a lot of those companies are doing this today and it’s not futuristic it’s happening today.
Interviewer: Well Brian, thank you very much for teaching us and helping us understand what is possible with predictive analytics. We appreciate your time.
Brian: My pleasure Glenn. Thank you very much.
Interviewer: Alright, talk to you soon.