Using Data to Work Smarter, Not Harder: Lessons from Collective[i] CEO Tad Martin

May 31, 2018

Tad Martin

When Tad Martin first joined in 2000, he entered a company with a lot of valuable data waiting to be used, but lacking the technology to use it. As COO, Tad led a lengthy and pricey big data overhaul to get the company up to speed.

After he left Overstock, Tad wanted to empower other organizations to harness the power of big data in an easier way. He started Collective[i] for a profession that he thought was underserved in big data solutions: sales professionals. Using predictive analytics based on their network of sales professionals, Collective[i] makes recommendations to its users on things like which buyers to target and when.

In this episode, we explore:

  • How sales professionals can make more money while doing less work using predictive analytics.
  • What it means to create network effects with data.
  • The strides big data technology has taken over the past two decades.
  • And the things sales professionals get wrong when they don’t rely on the data.

Want to make your own company more data-driven? Learn how Indicative can help.

Full Transcript

Lauren Feiner: You’re listening to Deciding by Data, the podcast that brings you into the C-Suite to learn how data drives successful businesses.

Today on the show, we’re exploring how a few data points can be amplified across a network to generate powerful insights.

We’re talking with Tad Martin, the co-founder and CEO of Collective[i]. It’s a platform for sales professionals and managers that collects data about the sales process from all the companies in its network. It uses that data to make predictions about the most effective ways to sell to target companies. Tad told us about how far big data solutions have come within the past twenty years, and what to look for in the future.

This is your host, Jeremy Levy.

Jeremy Levy: Back in 2000, the world learned that it was not ready for what the Internet had to offer. At the time of the dot-com bubble burst, Tad Martin had recently started running’s merchandising group after it had acquired his startup, He would later become Overstock’s COO.

Tad Martin: Half, if not more than half of the Internet world just went away because you had this massive dot-com recession where people realized that there are some basic business fundamentals that you need to run a business. And I think we were very fortunate at Overstock that we ran our business by those principles. We were very data-driven in the decisions we made. Everything that we did was supported by data in one way or another, whether it was how are we going to acquire customers, what types of products are we going to purchase, and how do we price them to optimize the revenue we’re going to get from them. Things that we ultimately got to the place where we started to say, what’s the best location for a product in the warehouse so we can optimize pick paths? And so if you look at one of the reasons that Overstock was able to survive that recession of 2000, I think a big part of it was how we were able to make decisions by data, which led us to be very capital efficient.

Jeremy Levy: The Internet forced companies to confront massive amounts of data like they had never seen before.

“…we actually got to the place in 2003 where the size and scale of our data outgrew the systems that we were using to analyze it.”

Tad Martin: If I look back at the size and the scale of the data we’re talking about, I mean it seems so small now compared to the way we have now. But at that time, thinking about being able to analyze terabytes worth of data, well it was a huge, it was a huge undertaking. And it’s interesting because we actually got to the place in 2003 where the size and scale of our data outgrew the systems that we were using to analyze it.

Now, it’s probably not going to surprise you that as an earlier stage or a growth stage-type business thinking very much like a startup, our data systems or analytical systems were SQL and Excel. And this goes back, I mean in the early 2000 when Excel even had the row limitation. You can only have 65,000 lines. And we had 700,000 skews, and we were repricing our products every week, and so you start thinking about, how do you do that? And so if you think back to 2003, we had been public for a year. We went public in 2002. We were growing at a 100 percent a year. We were doing $250 to $300 million of revenue. And all of the sudden we started to get to this place where we couldn’t produce the analysis we need to do to be capital efficient.

Jeremy Levy: By 2003, when Tad was the COO of Overstock, there were only a couple choices when it came to solving the company’s big data challenges.

Tad Martin: One, you could go out in one of the Big Five consulting firms and hire them to come in and do an analysis on your business, which would result in a presentation of a strategic roadmap. But that didn’t necessarily help you on the tactical side of things. The other option was to go down the path of an enterprise business intelligence implementation, where you would go out and buy a data warehouse, you’d go out and buy an ETL engine, you’d go out and buy an analytical engine, you’d go out and buy a visualization engine, you’d go out and buy iron, you’d go out and have to hire consultants to help you integrate this or model it and build it out. And that’s what we ended up doing.

It’s a hugely costly endeavor and it takes a lot of time and a lot of effort. Those two things combined, lead to things like statistics that say 72 percent of all these BIS installations fail. No one has the ability to stick to that because they take so long. Overstock was very fortunate in that we did stick with it. Two and a half, three years later, we launched with this program and if you look at what’s happened with Overstock, setting aside their endeavors into blockchain now, they’ve been profitable, I think 11 over the last 14 quarters, doing nearly two billion dollars of business, and it allowed for that growth to accelerate for that period.

Jeremy Levy: This laborious process that allowed Overstock to gain value from its own data informed Tad’s next business, Collective[i]. His now-wife Heidi and her brother, Stephen, had sold LinkShare, an early affiliate marketing company, which had to go through a similar process. The three of them wanted to create a solution for other businesses with similar problems making use of their own data. They chose to focus their efforts in an area integral to business, but whose needs had yet to be met: sales professionals.

Tad Martin: That’s one of the things I really liked about going after sales and the sales domain, is selling has been happening for centuries and at its fundamental level, although every organization will tell you that their sales process is different, it’s not different. Everybody goes through the same steps in the selling process. You have a[n] introduction phase, you have a discovery phase, you have a[n] exploration phase and a proposal phase, and then there’s negotiation and there’s a closing. Everybody goes through that. There may be different lengths, there may be different activities that happen, but when you start thinking about the process in and of itself, you can start to see what are the types of activities that are more specific to the company or the organization you’re selling to that might be different from other things that you do. And that’s what our machines are looking for is, what are the things that are specific to the process that you’re in that are going to give you the insights to help you succeed?

And that’s going to be different for almost every organization because the types of products you’re selling are different, the types of salespeople you’re using are different, the types of people you’re selling to on the other side are different. And so, what the challenge is for sales organizations are right now is they’re starting at ground zero and trying to figure that thing out through the sales process, which is really really hard and I have a tremendous amount of respect for salespeople because how hard that is. But if we can shortcut that process and provide them with those insights earlier, we can improve their likelihood of success, we could shorten the sales cycle, and we can make their time more efficient.

Jeremy Levy: Tad and his co-founders wanted to create a true system of record with Collective[i].

Tad says that CRMs like Salesforce, where sales professionals are often required to log their own activity manually, have been held up for years as the gold standard of record-keeping.

In reality, he says, they represent mistake-ridden records that take an unnecessary amount of time to fill out.

Collective[i] connects to the tools sales professionals already use regularly to do basic tasks like setting up meetings, so the logging is done for them.

Beyond record-keeping, Collective[i] gives sales staff and managers valuable insight into their own selling process.

Tad Martin: Sales managers can have complete transparency into everything that’s going on around a sales rep’s opportunity including seeing what emails are being sent back and forth. Because if you think about how sales managers manage right now, they’re managing relatively blindly because they’re relying on what a salesperson’s entering into Salesforce as the starting point of a conversation, but then they have to go into the room with a sales professional, ask them, “Okay, well tell me what’s going on. Is this deal really going to close next week?” and the sales professionals say, “Yeah, I had a conversation with this person, and he says we’re on track.”

But, then the manager has to go through this inspection process to make sure that A), the rep doesn’t have happy ears, B) that they’re talking to the right person, and C) that there’s nothing that was left out, like maybe we didn’t send them a contract. And I think it’s the nature of sales professionals to be very optimistic because you have to be, you’re told no 90 percent of the time. But, without that context sales professionals, sales managers have to spend so much time really finding out what’s going on, whereas we can expose it so a sales manager can now go into that meeting having full context of what’s going on with these opportunities, so, the conversation can turn to, “Hey, let’s talk about how we can improve our likelihood that this is going to close this week.”

Jeremy Levy: Collective[i] connects to the sales tools professionals use and offers them predictive analytics on how best to sell to their target companies. Collective[i] is able to bring transparency to the selling process because it operates on a network of data.

By using Collective[i]’s tools, companies also opt into sharing data around their sales process with others in the network. The data is anonymized in Collective[i]’s platform, but it can be used to make suggestions to others in the network about the buyers they should target, when to reach out, and more.

Tad explains the concept by comparing it to Waze, the app crowdsources driving data to help users avoid traffic.

“…every one of us had the experience where you look at Waze and go, “I’ve done this drive 100 times, why would I want to go that way?”  And yet, you either don’t do it, and you find out that you were wrong, or you did do it because you’ve trusted this network.”

Tad Martin: Waze is a data network. Every user that is using Waze is contributing their data to this network, that data is then being used to help guide you to where you want to go. And every one of us had the experience where you look at Waze and go, “I’ve done this drive 100 times, why would I want to go that way?”  And yet, you either don’t do it, and you find out that you were wrong, or you did do it because you’ve trusted this network.

What Collective[i] is doing is that same thing. If you think about how sales works today, sales professionals are trying to figure out who on the buying side I should be talking to? Who’s going to play what roles? How do I navigate the organization? Which lawyers are going to be good? And how does the buying process of that organization work?

Because we’re a network, we see those things because there’s inputs coming from all different parts of our network. So we can provide that roadmap for the sales professionals to know, who are the people that should be involved in the deal? What roles are you going to play? When are things going well, when are they not going well? What are the types of objections you can be seeing? The things that will help you shortcut a lot of the problems and things that you can’t see right now.

Jeremy Levy: Help me understand, what are the actual insights that a sales professional would get from using Collective[i]?

Tad Martin: From a sales professional point of view, it’ll help surface things that you wouldn’t be aware of, other than being a part of this network, so recommended buyers is probably the first example. You get a warm introduction to someone in your organization and you’re starting a sales process with this organization. We start to map in all the different activities that you’re performing, and then we look at the network and we may say, “Hey, it’s great that you’re talking to Mary, but we’ve seen her in these buying processes over the last six months and she’s merely an influencer.” The decision maker is most likely to be Mike, and then we can then let you drill into Mike and we might be able to tell you, “Okay, well Mike is connected to this other person that you used to work with at this and this company. You may want to reach out to Mike to be able to get that information.”

Jeremy Levy: I think about so much of the- at least the enterprise sales process is being based on relationships and having good communications with your prospects. Does this replace the notion of grabbing drinks after work with who I’m selling to get that sort of back channel information? Is that what this provides?

“…why not let machines do the things that they’re good at and free up a sales professionals time to do more of the things that they’re better at.”

Tad Martin: No, it’s just the opposite of that actually. If think about what I consider the most valuable things that salespeople do, it’s understanding people, developing empathy, and creating relationships. And if they’re spending 20 to 30 percent of their time not doing that by doing data entry, or sitting in meetings, or whatever it is that’s not helping contribute that, then why not let machines do the things that they’re good at and free up a sales professional’s time to do more of the things that they’re better at. There’s always these discussions around, are machines going to be able to replace people. In a lot of cases that’s true, in certain instances that’s not going to be the case. I think the interpersonal skills of salespeople is going to be very, very hard for machines to replace. And so, if we can free up their time so that they can spend more time going out and grabbing drinks, and figuring out, am I talking to the right person? Or, how do I make sure that I keep this deal on track? That’s one of the things. The second thing is, does is it helps uncover people in your network that can help you do that more effectively.

Lauren Feiner: We’re going to take a short break, but when we return, Tad will debunk common sales myths with data. Stay tuned.


Lauren Feiner: Welcome back to Deciding by Data. We’re talking with Tad Martin, co-founder and CEO of Collective[i], a data network that helps sales professionals make more intelligent, data-driven decisions. Tad explained why a network is a better predictor and system of record for sales professionals than standard CRMs. Now Tad, will tell us what sales professionals get wrong about selling when they don’t rely on the data.

Jeremy Levy: Have there been insights that you’ve been able to gather from the data that you’re collecting that have been counterintuitive to what my instinct would be as a sales professional?

Tad Martin: [Laughs] Yes. I think the, well, the best example I have, outside of what we usually see for the first couple of weeks, is the first thing we always hear is well, “Your system’s wrong. That can’t be my data.” And it’s because it’s the first time they actually get to see their data in it’s true form, and that’s part of our onboarding process. Once they start to realize it, they start to drill into some of the things. But one of the biggest insights that most organizations find in the first 30-60 days is they realize that their pipeline is not what they think it is. And there’s a reason that the rule of thumb in most organizations is I need 3X to my pipe to make my numbers. It’s because two-thirds of my pipe is crap and I just don’t know what it is.

“And there’s a reason that the rule of thumb in most organizations is I need 3X to my pipe to make my numbers. It’s because two-thirds of my pipe is crap and I just don’t know what it is.”

Jeremy Levy: Yeah.

Tad Martin: Well, with an application like this where you can start to look at real odds of success. And if you look at most CRMs, they’re set up to map odds to a stage, which is only as good as people are accurately reporting those things, but because we actually use AI in systems to have predictions on what’s going to sell, you can start mapping up odds of success with how long a deal has been in the pipeline. And it’s not uncommon for organizations for the first time to see that they’ve had this deal in their pipeline for a year and a half, two years, and they have a three month sales cycle. And the odds of success are 10 percent. So, as a manager, you can have the conversation, is why is this deal even in your pipeline? Why don’t we remove these deals and free up your pipelines so I can actually put deals that you might have a better likelihood of succeeding with. And it’s just the nature of how things work. Everyone is very short-term focused. How do I make my number, this month, this quarter? I’ll worry about the future after that.

Jeremy Levy: The 3X rule still stands, but it’s only how accurate the actual opportunities in your pipeline are.

Tad Martin: Yeah. Well, I would argue that the 3X rule should become much more efficient. If our systems are doing a job predicting what is going to succeed and what isn’t going to succeed, managers will start to be able to say, “I’m going to kill these three deals so I can put three more deals in there to help you succeed.” So, you’re going to either drive more efficiency and get more out of your representatives or you are going to drive more revenue because you’re going to be able to have more of a bigger impact in what they’re doing.

Jeremy Levy: You mentioned the AI aspect and I think you said ML also. Talk me through what does AI actually mean in the context of Collective[i]? Because I asked this a lot of people, and the term AI is like the cloud from 10 years ago. No one knows what it means. What does it mean to you?

Tad Martin: I’m not going to get into a discourse of AI, and I would recommend that for those of people listening who haven’t listened to the podcast you do with Matt Turck because that was probably one of the most effective descriptions of what AI is. Because in my perspective, AI is kind of the new big data.

Jeremy Levy: Yeah.

Tad Martin: It means a lot of different things to a lot of different people. It’s the subsets of that. It’s the machine learning. It’s the neural networks. It’s the deep learning. It’s the things, it’s how you apply different methods depending on the objective you’re trying to solve for. But because we’re selling to a line of business, a lot of the people that are running sales teams don’t have that next level of understanding. And so, you have to find that middle ground of, “Okay. Well, how does that work?”

So, in most organizations, when we’re talking, we actually don’t even talk much about AI because it’s the predictions and the network and giving people information they didn’t have access to that’s valuable to them. When we talk about it internally and how we’re going to design what we’re doing, it depends on the problem we’re solving. When we start thinking about mapping users in our network so that we can connect them, we talk about neural networks. When we talk about making sure that where our predictions, whether it’s around forecasting or probabilities, or predictions are accurate, we talk about machine learning. So, it really is, what’s the best approach to solve the problem that we need to solve for? Again, it’s lots of different things and we’re using lots of different things depending on what part of the application needs to consume it.

“[AI] means a lot of different things to a lot of different people.”

Jeremy Levy: Tad said that one technology Collective[i] is testing is Natural Language Processing. This is a type of AI that allows computers to understand human language.

Tad Martin: What you’re starting to hear a lot more is how effective NLP can be in being able to understand dialogue in conversations, we’re doing some really interesting things related to that. It’s still early in some of the applications but-

Jeremy Levy: Do you mean like conversational interfaces or just interpreting sentiment from natural language from freeform?

Tad Martin: It’s not just sentiment, it’s also being able to map intent. It’s being able to understand how different communication styles work with different people, and ultimately, even being able to recommend a specific way to reach somebody. So, if I’m a sales professional, I probably have my canned response that I do as a thank you, as a follow up, that I customize the two from and maybe a line of something personal I met.

But if within our network, our systems can start to say that this person will respond to this type of emotional construct, then we can start recommending communications that will give you a higher likelihood of getting the response that you need. So, there’s some really interesting things going on there. It’s a really hard problem to solve for not only technical reasons but psychological reasons. There are some interesting things going on there.

If you look at what Alexa and Cortana and Siri are doing in the processing of language and being able to interpret and understand, I think it’s interesting. There’s companies that are doing nice jobs of taking that as a form to be able to be a data input. And again, as you’re able to effectively transcribe that, then it becomes available for these other technologies. But I haven’t seen a lot out there that’s really integrated something into an end-to-end solution other than these tools that help you do some of these things. That’s where we’re really focused on, how do these tools help us provide the insights to get… because if you’re just doing more effective data capture, there’s a value in that, but the ROI is limited.

“[I]f within our network, our systems can start to say that this person will respond to this type of emotional construct, then we can start recommending communications that will give you a higher likelihood of getting the response that you need.”

Jeremy Levy: A network of data means there’s a lot of data to protect. We wanted to know how Collective[i] handles that responsibility in a time where data security is top of mind for every business.

Jeremy Levy: It’s a really interesting time right now with data privacy. Both GDPR being rolled out, there’s a lot of controversy around Facebook in Cambridge Analytica. You’re not collecting data on individuals, more companies, but is there a responsibility from a privacy perspective around some of the data that you’re collecting and how do you think about that?

Tad Martin: So, first and foremost, everything that comes into our system gets anonymized when it goes into the network. So, when we’re analyzing it, even the systems don’t even know any individual information. I mean, we’ve been building this for quite a long time and based on our backgrounds with what we did it at Overstock and LinkShare, we’re very familiar with systems that you need to have security and deal with PCI, and these other compliances. So, from the very beginning, we’ve been very cognizant and aware of making sure that our systems have that security involved. We also need to do so in the application and what you share [with] users. And so, if you’re not part of the network, you can’t come in and find certain things about opportunities and deals. It’s just the nature of how we protect users from the system. You never get to see any other user’s information unless you’re invited into an opportunity. We have this concept of team selling where if I’m selling a company and you have a relationship there, I might invite you into my opportunity so you can help me sell them. And so, I would share the information related to that opportunity, but that’s the only thing you would see. We are very careful about making sure we restrict what people have access to, to protect that information from companies.

Jeremy Levy: Are there risks with this model? I mean not specifically with Collective[i], but the idea of just aggregating data on scale on individual people, there’s really powerful applications of it, but it strikes me there’s also enormous risks. Are you thinking at all about whether or not it’s appropriate from a legislations perspective to create laws around privacy?

Tad Martin: We talk about it and we consider it in all of our application development and the products that we build. But again, because we’re dealing with business data, it’s a different question for us because it’s less around privacy and it’s much more about security. And I’m not saying that’s an easier or less complex problem to solve, but I think it is a little bit different. We do have the ability for individuals to come up and then sign up themselves, so there is that element of privacy that we’re always staying on top of. But I’ve given up on trying to predict what legislators will do and whether you agree with GDPR or not, you think it’s too overreaching, it is what it is, and figuring out the best ways to make sure that you’re complying with it …

Jeremy Levy: Are there GDPR implications for Collective[i]?

Tad Martin: No. Again, from the very beginning, we’ve always thought about our users, our security, and keeping information within the right context of what you need it for. And so, I mean we’ve had to update some policies. We’ve had to make some tweaks in how we do certain things with technology.

Jeremy Levy: Yeah.

Tad Martin: But we have not had to re-architect anything bigger than that. It hasn’t been a huge tremendous effort that’s cost us a lot to be able to do it. And I think- I always look at things from the perspective of, if you do what’s right for your users and you do what’s right for your companies, and you’re cognizant about the best ways to keep people protected and provide the most value, something really crazy has to happen for it to be a big miss for you.

“I always look at things from the perspective of, if you do what’s right for your users and you do what’s right for your companies, and you’re cognizant about the best ways to keep people protected and provide the most value, something really crazy has to happen for it to be a big miss for you.”

Jeremy Levy: With all the technology we use from the moment we wake up in the morning, Tad feels that sales technology has lagged behind. He hopes Collective[i] can make sales professionals more data-driven by modernizing the tools they use to make decisions.

Tad Martin: A lot of people wake up in the morning. They pick up their phone. They check the weather. They walk into the kitchen and they ask Alexa what’s the latest news. They talk to Alexa and say, “Hey, here’s my Spotify. Can you play this music?” They get in the car, they turn on Waze, and before they start driving they log into their Starbucks app so the Starbucks coffee is waiting for them.

“It feels like you’ve kind of left this world of modern applications to go back into these enterprise technologies which have not kept up with how users are now used to being able to do that.”

And then they get to the office, and then think about the experience that you have when you’re in the office. You stand in front of your desktop. You open Excel, you open Salesforce. It feels like you’ve kind of left this world of modern applications to go back into these enterprise technologies which have not kept up with how users are now used to being able to do that. And so, the only way we saw that we were going to be able to use that and do that was to take ourselves out of the paradigm of how people think about CRM and develop a standalone application that people would want to interact with because it’s more akin to how they interact with software and applications outside of their enterprise day-to-day work.