Sales

TrustRadius @ Dreamforce: Interview with Shashi Upadhyay, Lattice Engines CEO

Lattice Engines’ CEO describes how predictive analytics can transform marketing and sales processes and unclog the B2B funnel. Over the past few years, Lattice Engines worked to achieve a surprisingly fast deployment rate, which continues to be a win-win for customers and company growth.

 

Can you give us some background on Lattice Engines? What’s your vision?

We are pioneers in the predictive marketing and sales space. We were the first company that ever started talking about predictive marketing and sales for B2B. We start with the premise that if you can predict who is going to buy, what they’re going to buy, when they’re going to buy, and how much they’re going to buy better than the average person, then you can use that to improve everything about the sales and marketing function.  So for example, if you have perfect information about what a particular person is going to buy on a particular day, you don’t have to spend your marketing dollars and sales dollars on anybody else, you would just go to them and talk them.  Obviously, you can’t have perfect information but what you can have is very high-quality probabilistic information—this person is 80% likely to buy, versus this other who person is 70% likely to buy. We embed that into the workflow of everything marketers and salespeople do.

In B2B, the marketers and salespeople think in terms of a funnel. Our vision from day one has been: we want to use predictions as a way to unclog the funnel at every stage.  The approach depends on what part of the funnel is clogged. For example, a standard way in which funnels get clogged is the very poor interface between marketing and sales. This involves complaints about the quality of leads, etc. So our lead prioritization solution, or what used to be called the lead scoring solution, unclogs that part of the funnel. It’s a neutral thing that uses data science to identify the best leads, which then get to sales, and identify the leads that are not that great, which then get pushed back to marketing. That way the infighting between the two teams stops. Everyone is happy. Once that’s unclogged, teams can move on to some other problem, for example, identifying completely new contacts or selling more to the existing customer base, which is upsell or cross-sell.

How does Lattice’s lead prioritization work, and how does it differ from other approaches?

Our vision is across the funnel and our product is across the funnel.  We believe that just focusing on lead scoring or finding new accounts is not enough—you have to do the whole funnel, beginning to end.  In terms of how the whole thing is done, the idea is, at some level, very simple. You start by collecting all the data on every known business in the world. There’s a ton of data available on the web about who these companies are and what they are doing. Are they growing, are they shrinking, are they hiring, have there been executive changes, how well do they treat their employees, and what technologies do they use? We collect a huge amount of that data using our search engine.

We also partner with companies like Dun & Bradstreet to provide us very specific kinds of information in addition to that. We have an open ecosystem, so our approach is very different from the other vendors who have taken a more closed approach. They only focus on their data, whereas we have found that taking an open approach has allowed us to be global on day one.  For any given country in the world, you can find data with us.

That external data then gets combined with the data that our customers have in their own systems. This gives us a sense for who their customers are, as well as who they would like to have as customers. Once we have that, the algorithms then do the work of sifting through and identifying who is a great lead, who is not a good lead, who is a great contact, who’s not a good contact, what’s a great cross-sell option, and what’s not a great cross-sell option, things like that. Those judgments get pushed back into the CRM and marketing automation system. When salespeople take action, their actions get recorded and then reported back on, and from there you get a sales learning system that just gets better and better. That’s the basic idea.

So where is your focus, and what kinds of initial success have you had?

We now have close to 200 deployments in place.  We tend to focus on the midmarket to large enterprises.  Others in the space focus more on SMB, whereas we have a truly enterprise-class system that can scale all the way to a company like SunTrust Bank or Dell or HP. But it can also support high growth companies, whether it’s Qlik or FireEye or DocuSign. These are all customers of ours, and they’re all very sophisticated buyers. They’ve done thorough A/B tests to evaluate our product. (For example, when we predict that something will convert at a high rate, does it actually convert it a high rate?) We have done very well on that front.  As the results have come in, it’s become very clear that Lattice Engines works. Our solution tends to improve the performance of the team.

What kinds of specific improvements are your customers seeing?

These organizations care about three metrics: conversion rate, time to close, and average size of the deal. One great thing about predictive technology is that you can improve all three metrics simultaneously.  Most other products and strategies tend to improve one thing, but then hurt the other things. Here’s a good example—let’s say you’re buying a list, and you decide you want to spend less money and get very clever about it to improve your conversion rates. But very often leads that convert fast also tend to be smaller deal sizes, so it’s actually hurting your average deal size.

The great thing about Lattice Engines’ technology is that you can actually do all three things better, simultaneously. Our vision is that over time this will be a must-have.  So if you have a funnel, you’re going to have something to unclog it. Our technologies help sales and marketing do that.

What does a deployment look like? How long does it take to implement, and how do you ensure that all of these different data sources are talking to each other correctly?

How does it do what it does, right?  There are two things going on. One, because we focus on the sales and marketing domain, we’re not like a machine learning company that’s trying to solve an arbitrary problem. There are very well understood notions already—there is such a thing as winning a customer, there is such a thing as losing a customer, there is such a thing as converting to opportunity, and so forth. People use the same language, and because the language has been established, all we have to do in the deployment is map those notions to specific columns that a customer has in their own database and that’s it. That’s the only thing we need to do, and we can do it really fast. For the last couple of months we’ve been doing a 24-hour proof of concept, where a customer can give us access to Salesforce in the morning, and by evening they’ll actually have a live application. So, it’s not a convoluted exercise. The reason is because Salesforce has a very well-defined notion of what’s a “customer,” what’s an “account,” and what’s the “lead.” We use that to make the deployment really simple.

Two, for the data that we provide, we curate it ourselves. For example, we can set up a trigger of business change or a trigger of a growing company. We have taken all this data and created the triggers, so the rules that say they must carry this course. But we have an index, for example, that tracks the growth rates of companies and those then become inputs in the modeling.

Who’s your target customer? What type of company is a good fit?

We like sophisticated companies as customers. We’re a good fit for companies that have at least 100 employees. I think if they are much smaller than that, then there are a bunch of other vendors who could provide them a solution.  We work with enterprises all the way up to the Fortune 500. The higher, the bigger, the more complex the environment, the better suited we actually tend to be. But we can also help a single product 100-person company.

The main thing we are looking for is that they are thinking about predictive technology as a transformation; they’re going to do things differently, versus simply looking at it as a check-the-box-exercise.  If they just want to do one little thing—just score the leads for me and go away kind of thing—we have found that the value they get can be a bit patchy because they aren’t involved in creating real change in their processes.

We found that the solution applies quite nicely across verticals. So we are in technology, but we’re also in business services and financial services and manufacturing and distribution. In terms of size, we seem to be very well suited to a pre-IPO company and above, but not necessarily a 10-person type company. The mix of customers reflects that: it’s very good brand name large companies, and very high growth pre-IPO or post-IPO companies.

You’ve tripled your customer base in the last year—can you give us some insight into how you’ve managed to scale adoption so quickly?

We had a very dramatic increase in the number of customers last year and that’s continuing. Part of the reason for that is because we’ve achieved a significantly lesser deployment time.  A couple of years ago, it would take us several weeks to deploy things. The longer it takes to deploy, the longer it takes to sell, right? But if you can start doing proof of concepts in a day, then all of sudden, that uncertainty—‘will this work for me?’— goes away.  That’s been key to our adding customers at such a fast pace.

The other part, which less visible, is that “unclogging the funnel” business. Once our customers are successful with that, they’re very happy to then point predictive at their next problem. As a company, we’ve done a very good job of expanding inside the customer base. We can only do that because we’ve made the first deployment successful. So we have quite a few customers that are now using us in five or six different business units because they found success.

You described a pretty fast deployment, but does it take some time after that for the predictive to really start working? Does the system have to watch trial interactions?

We use history to learn, so the answer is no. For 95% of our customers, a quarter or two of your business history is all we really need for the model to work. 

You mentioned that there were a few points in the funnel that could get clogged where Lattice Engines could help. What are some other use cases in addition to lead scoring?

The lead scoring one is actually fairly obvious. But the problem gets worse both further down the funnel, as well as further up the funnel.

Let’s start with further down the funnel. If you have one product and you are selling installations, it’s very easy, you don’t have a problem. But, most of our customers will typically have at least five or six products, and then they get into the same problem: one product that has 100% penetration by definition and then everything else has 5% penetration. They build all these things but nobody’s buying them. So now they go to the installed customer base—we think of them almost as leads, because you know them and you can call them, but you have five-pitch choices every single time. For a larger company like VMware, there are 35 choices. For a distribution company, there might be literally a thousand choices. So what product are they going to call about? What are they trying to get in there next?

The same formula that worked before also works in this situation, except the question now is: given a customer, what product am I selling to them next? It’s the same thing with renewals, but with more analytics and less process. How likely is someone to churn?  There are many signals, many reasons people churn. If they’re not using an application, that’s an obvious one.  But a lot of times they’re likely to churn because they’re actually going out of business, and there are plenty of indicators on the web that they’re going out of business. We can provide that, so people can use it as a signal.

If you go further up the funnel, it gets even more interesting, because the further up the funnel you go the less you know about a customer.  The person in marketing who is managing this database with a million records has no information about a single record. So this person is running a series of campaigns with, basically, no idea if they should be even running that campaign against that particular person.  They’re just having fun with Marketo or Eloqua, right?  The most basic example is someone in the financial services vertical with between 10-100 employees is more likely to convert than another vertical. If you knew that for a fact then you would actually create campaigns for precisely those people, with content that is about the financial services vertical. If you run campaigns that way, then you improve things all the way downstream.

Then we can even go further out of the funnel, with all of these people who you don’t even know about—the Dark Web equivalent of the marketing database. If I’m a storage company, you should be interested in people online who are reading articles about storage. There is a pretty rich ecosystem of ad targeters who are cookie-ing these readers who might be buyers. In the B2B world, if you can trace that cookie back to what IP address that search came from, you have some information about the content someone from a business is looking at, and that becomes a part of our data source. So now we can not only identify things in your database that are interesting, but we can also say, here are all these people out there looking for your stuff. You should be adding them to your database.

That’s what I mean by transformation. Every activity these marketers or salespeople do can be made better with this kind of predictive approach.

Which employees at a company are buying Lattice Engines, and who are your primary users? Are you sales technology, marketing technology, or both?

We are a revenue technology. I think the marketing/sales distinction is going to go away over time. It’s kind of an artificial distinction, especially as marketers have become much more revenue-focused and less awareness focused. Our applications span the marketing to sales spectrum. Because of that, if they are buying some of the funnel solutions, the buyer will be someone in demand generation—Director of demand gen, for example, or VP of demand gen or sometimes even the CMO. The further downstream we go, the more buyers are looking for campaigns around sales, so that would typically be a purchase by the VP of inside sales. But in most organizations, those two people (the inside sales leader and the demand gen person) tend to work very closely together anyway.  So we do have two kinds of buyers, but our position is that their needs are more similar than different.

On the users side, we have one set of users who are demand gen marketing analysts—basically the same people who would use a Pardot or Eloqua or Marketo.  The second set of users is data scientists. The third is the sales reps. We have an app so that a sales rep can see predictive insights in their Salesforce instance.

What is your place in the Salesforce ecosystem?

We integrate with them, and about 70% of our customers are Salesforce customers.  But we also integrate pretty closely with Marketo and Eloqua, and we see all three as equal partners in that sense.  We have seen some of the other players focused too much on one partnership versus the other. Because of the diversity of our users, we have found it’s best to work with all three of them. With Salesforce specifically, we have a completely native app built on the Salesforce one platform. It’s mobile, and it does all the right things.  But we’re not completely aligned with them, is what I would say. For example, we haven’t taken investment from Salesforce. We’d be a little uncomfortable with that.

What category do you see yourselves in?

The category that seems to be emerging is, unfortunately, called “predictive marketing,” even though it is really “predictive marketing and sales.”  I heard somebody call it “predictive revenue generation,” but I think that’s too many words.  If you force me I would say that the language that Gartner has been using is pretty consistently “predictive marketing” now. I would say we belong there, at least in the short term.  We’re not, for example, in the sales acceleration category where InsideSales has been. I see that as a separate category.

Who do you see as Lattice Engines’ competitors?

In the mid-market, we have a number of competitors. Infer and Mintigo are probably the two we see most often, and then sometimes we see EverString and LeadSpace. In the large enterprise, we have sometimes seen 6sense, which is an emerging enterprise-focused company. We haven’t really seen anybody else in the large enterprise segment. We’ll see how that goes. Also, in enterprise, there is a buy-versus-build decision.  So we always compete with, “Hey, I can do this at home.”  That’s the elephant in the room, and we focus on addressing that issue more than we worry about competition.

We hear other spaces talking about predictive analytics, especially incorporating predictive into marketing automation products. How does this affect you? Does everyone have the same definition of predictive analytics?

A little bit of this is just the market catching a buzzword. But we have found that some buyers are very sophisticated and they can cut through the marketing lingo really fast—they know exactly what it means to be a predictive solution, and what it doesn’t mean.  We don’t worry, because we have confidence that our customers understand what it means. To some degree, because the space is getting more coverage, people have stepped up to do a lot more explaining around what predictive means.  So, I think that confusion will take care of itself.

Now, embedding predictive analytics in an application is a very good idea.  In fact, I think there’s a big opportunity out there today to go build the next Marketo or the next Eloqua, but do it predictive-first.  First build the predictive lead, then build the marketing automation, so it’s a smart system on day one—it immediately knows who to contact, etc. But with where we are right now, we have to try to fit into the other existing solutions people are using.

What kind of technology is necessary for predictive to work?

That’s a really good question.  There are two aspects: building a predictive model and running a predictive company as a cloud service. As a cloud service, we ran into issues when, for example, it was the night before Dreamforce and everyone was trying to run their models and figure out who they should send out to their booth and who they should be calling on. The ability to deal with those spikes—that’s hard.

The modeling side is hard because it requires a number of steps. One, you have to have a search engine, because we deal with 150 million business websites.  It’s not Google scale, but it’s a hard problem. You need something that can tell if it’s a business site or not, if it’s a consumer site or not. Then it has to go and collect a bunch of data, convert into something useful, etc. Two, you need the thing that takes all that data, and takes the data from your customers (often structured data), and converts it all into something that a machine can understand. So you need a fairly complex stack of for matching and cleansing; they call it a data pipeline. The third thing you need is the algorithms themselves, and the algorithms have to be good enough that no matter what data quality, they still give you something reasonable. The fourth thing you need is integrations to all of these other systems and so on.

You’ll notice that the space will end up raising a lot of money because in a way, it’s really four different product that have been brought together to build a company.  You can build an entire company out of just doing a search engine.  You can build an entire company out of just moving data around.  You can build an entire company out of just doing cleansing and matching, or a company that has tools for doing prediction.  But putting all that into one roof makes it hard. This is why we don’t worry so much about marketing automation guys getting into it, because—well, do they really want to go build all that stuff from scratch? Why would they do that? 

Can you tell us more about your new Buyer Insights app?

One of the things we found was that just putting the predictions into Salesforce was not sufficient to engage salespeople. It was very important for engagement that not only do reps get their predictions, but they also get some notes with reasoning and context to explain the predictions. We figured out how to automatically generate that, so reps don’t have to type in their notes by hand every single time. So, Buyer Insights is really a recommendation system for our salespeople, embedded entirely in Salesforce. They log in to Salesforce and they can get it. It’s completely mobile, and no matter where they are, they can look up an account and the app will immediately tell them is this likely to buy, why is it likely to buy (or why is it not likely to buy), what are similar accounts, and what things they should be talking about if they do make a call. It’s instant help!

The whole idea is to make everyone seem smarter, and allow them to have that contextual conversation with people.  The cool thing is that we build it entirely on Salesforce, so it’s a very native experience. You can’t tell where Salesforce ends and Lattice begins—it’s very, very seamless.

Why is this kind of context important for B2B sales?

It’s pretty clear that buyers only want high-quality interactions. It’s not that they don’t want to interact with salespeople at all, but they don’t want vacuous conversations talking about sports equipment and golf, or whatever. That era is over.

What they do want to talk about is how you will help them be successful. Buyers can go do a lot of searching and learning before they talk to salespeople, and they might talk to multiple vendors or run multiple POCs. The discussion should really be about the remaining part of their journey, and salespeople can’t have that if they’re going in blind, just kind of smiling and dialing. Buyers should feel like the vendor knows their needs and can talk in their language, and that the vendor has their best interest in mind. Our predictive technology enables that.

To share your own user insights about Lattice Engines, write a review on TrustRadius.

Megan Headley

Megan is the Research Director at TrustRadius. Her mission is to ensure we gather the highest quality data from authenticated reviewers, and provide useful curated reports for prospective software buyers. Prior to joining TrustRadius, Megan was Director of Sales and Marketing at a media company. She holds MA degrees in Journalism and Latin American Studies from the University of Texas.

Leave a Reply

Your email address will not be published. Required fields are marked *