The Exchanges

Q of HubSpot's growth was to break from just a Cartesian plane where there were two things that you could upsell on, maybe seats and some feature and adding a third and a fourth and a fifth and a sixth, but you can't add too much complexity. How do you make a decision whether to add something as a pricing based utility axis or just let it be included for free?

A Yeah, I think it's a really, it's a really good question. I think that, um, Sometimes putting things in as free can actually add complexity to the customer if they're not going to want to use it and, and also makes them feel that there may be over overpaying because they're getting things that they actually don't want to use. I mean, they, the traditional pricing model calls them, you know, killers. I think, you know, McDonald's did testing and that's why they add the fries, but they don't throw in the apple pie because they add apple pies into their sweets and the, uh, and the sales of it went down regardless if somebody was getting for free. That's the first consideration. There's been a lot of work and a lot of studies on that. Then the question is really is how do you equate the pricing model to value? And an example of that is AnswerBot. So for example, this is an ML product that we bought that will actually solve customer questions for you based upon looking at the question and then finding resources. And we looked at that and The value to a customer of them not having to have that question go to a support agent, which tends to cost, you know, can cost tens, 15, tens, 20 dollars, uh, is very, very high. So we decided to price that separately, and we charge one dollar per resolution on that as well, and it's very easy for the customer.

AI assessment note: “Then the question is really is how do you equate the pricing model to value?”

Q Okay, good. So, pre-IPO, uh, what was shape, you were, you would have been at the company about three years, I think, at that point, or four years. What was shaping product strategy, and how did you think about pricing?

A Yeah, it, it, At that stage, we were, um, we were really just starting our evolution from being a single product company to become a multi-product company, and so that was one dynamic in thinking about pricing. The other dynamic is that we were really starting to move up market for larger businesses at that point in time. Zendesk, you know, was for small businesses to begin with, but really started to acquire a lot larger business. We think of that as people who have more than a hundred seats. Um, what we really started to think about At that time was really, how do we put a pricing model that allows companies to, um, pay for the value they're getting out of the software, allows them to scale super really simply and keeps complexity out of the business model. So a lot of these enterprise software vendors had super, super complex, um, menus of, of, of purchasing options, and we just wanted to keep it very simple. So we, we stayed true. We had a per agent cost. And, um, you know, you could have a plan type of, of just a simple plan or a more enterprise plan as well. And that's pretty much where we are today. We have a few variations on that, but it's where we are.

AI assessment note: “starting our evolution from being a single product company to become a multi-product company”

Q Exactly. Um, in all seriousness though, how do you decide, I mean, that space, I could point to 30 companies, very similar to outbound, very similar teams, very similar tech stack. How do you make decisions around where to focus your time?

A Um, we, we have a reasonable understanding based upon where we, uh, you know, what our customers are asking for, and some of the market dynamics as to where things, where things are moving. Um, and, and then from that, really, it's just a case of, okay, you know, Should we build in order to, should we build in order to, uh, you know, base it on our own tech stack, or are there some people and some tech that's there that can really help to accelerate that and also enhance our knowledge of a particular domain? And, and that's, you know, that's what we found from the outbound guys. But, you know, there's some stuff we built ourselves. Our, uh, Uh, our machine, latest machine learning product, AnswerBot. We, we, we built ourselves, um, you know, and, uh, you know, we have data scientists on staff, and because a lot of the work on that is how does it integrate, integrate very, very smoothly into the other products that you've got.

AI assessment note: “based upon where we, uh, you know, what our customers are asking for”

Q great? I said, no. He's given us a forty-five-day free trial at NathanLatka.com forward slash schedule. That's not gonna stay up forever, so go get it now. NathanLatka.com forward slash schedule. How do you differentiate between what's signal and what's noise, and can you quantify that in terms of a sample size or cohort? You know, if this percent of people do X, we know it's a signal, not noise.

A It's, um, again, it, it, It is, it is bearable. I think in many cases on product usage, you have to look at things for a certain amount of time. Uh, it's amazing how, you know, how time can best, whether it's trial adoption or not, because you are going to get weird things like seasonality coming in, or you're going to get other factors coming in. So I would say looking at things long enough, um, I actually, It's funny you say on that, on using a sample set or not as well, because I feel a lot of people don't use cohorts or samples enough. So actually you can get probably more intelligence a lot of time out of taking a hundred, a sample of a hundred customers and looking deep into what those customers are doing, whether it's when did they adopt, what features are they doing? You've got a big enough sample size that you can then extend that across your whole base rather than trying to do a top down and look at your whole user base. And make assessments, make assessments out of that. So, um, I always encourage our product managers to try and, even if they're looking at broad-based metrics, to look at a customer cohort and go deep and understand what's, what's underlying the numbers as well.

AI assessment note: “taking a hundred, a sample of a hundred customers and looking deep”

Q Over the next 12 months, what will be a bigger driver of Zendesk product and pricing decisions? Um, increasing wallet share across current customer base, or expanding ARPU across current customer base, or going and finding new wallets to acquire completely?

A Yeah, I, the, We're, we're definitely seeing, uh, very strong growth, um, upmarket. You know, our, I think over a third of our MRR now comes from customers with more than a hundred agents, and we are investing, uh, all the time in order to, uh, add capabilities within that area, but at the same time, keep the product super simple to deploy and configure. Um, so that, that, that's one of our drivers. That, That obviously helps us with win rate and new customer, new customer acquisition. There are some interesting dynamics happening though, where we're seeing, I gave the example before about how a large number of people are involved in the support and service process. So we're very interested in how we can support those people and increase the use of Zendesk beyond just the core base, but, but, but, uh, other, um, Uh, other people within the organization as well. So that's potentially, uh, existing seats. And then obviously as we add new value and add new products, then that increases the ARPU as well. So, I mean, you, you nailed the three vectors that we look at from any new product thing. It's, it's win rate seats and ARPU.

AI assessment note: “you nailed the three vectors that we look at... It's win rate seats and ARPU”

Q Do you, when you look at new experiments to run, how do you typically find them? I found it very interesting when I've had some other CEOs that are your size software companies, they'll look at customers that have had the highest lifetime value over the past 10 years, reverse engineer their onboarding flows, and then replicate that for new customers coming on. How do you do it?

A Yeah, I think it, I mean, it's a really good question. And actually, you know, if you, if you were to ask me, What I wish I'd done, you know, six years ago, or if I was founding a company now, actually building and ingraining testing methodology into a business right from day one, Would be super important. And whether it's building A-B testing into structures, into everything that you do, collecting the right product analytics, or, or testing in a geography before you roll out, you roll out to the whole business. Um, we definitely look for signals. Uh, you know, we, we look for, um, and we do segment, uh, uh, for usage across, uh, across different, uh, customer, uh, Different customer segments and different plans. Um, I wouldn't say we necessarily focus on, you know, trying to replicate what is the largest customers. Largest customers tend to be outliers and have very unique requirements in which case you need to overlay certain levels or levels of Service on top of that. A lot of what we're trying to do on product is, um, really solve the problem for our broad base of customers and, you know, and move things in that direction. Typically it's around product usage.

AI assessment note: “we do segment for usage across across different customer segments and different plans”

Q from light to a team plan, you do have some utility based metrics or number based metrics that are not per agents. For example, two triggers, Two departments. Those then turn unlimited at the 30 dollar price point. So I'm trying to get a sense of, are these always under testing mode or have you identified that you should only have one number based metric tied to a pricing plan?

A Yeah. I mean, you know, this is the beauty of SAS. You know, when we started talking, I just fell in love with SAS and its business and somebody who's comes from the commercial side and you can enter into these decisions with your eyes very wide open as, as, as to what are people using? And, and, and set some thresholds there as, as to, uh, you know, the types of businesses and, uh, and what the long tail is on some of this stuff. And so actually typically where a long tail exists, then there's an opportunity to actually add the premium price, uh, to those into those features because they have very specific utility to a small number of customers.

AI assessment note: “set some thresholds there as, as to, uh, you know, the types of businesses”