Improving Gong’s customer retention with predictive data analytics

Tarush Aggarwal (00:00):
Hi, everyone. Welcome to another episode of People of Data, where we get to highlight the wonderful people of data. These are our data leaders from various industries, geographies, and disciplines. Discover how companies are using data, their achievements, and the challenges they face. Today on the show, we have Mike Tambe. Mike has spent 10 years in the data and analytics world. He's had stints at Amazon and LinkedIn, getting started when data science was just becoming a field. Most recently, he's a data leader at Gong, where he's working on combining predictive analytics with Gen AI. He's also mastered French cooking. Welcome to the show, Mike.

‍

Michael Tambe (00:44):
Thank you, thank you. I don't know if I'd call myself a master, but I...

‍

Tarush Aggarwal (00:47):
Are you ready to get into it?

‍

Michael Tambe (00:52):
Yeah, let's do it.

‍

Tarush Aggarwal (00:54):
Awesome. So what is Gong? How does Gong make money? How does data play a part in it? I know we've all heard of Gong, but we'd love to hear it from your mouth.

‍

Michael Tambe (01:02):
Yeah. And it's good because I think that we're wanting to get our real message out there. So Gong is an AI platform that transforms revenue growth. And so I think you can kind of think about it as a series of blocks that build on each other. And so at that basic block is conversational intelligence. And this is our bread and butter. So we record calls, email transcripts, use proprietary AI to really come up with actionable recommendations. Like, you're speaking too fast, you're speaking too much, you're not pausing enough. You're not asking enough questions. You're not asking the right questions. And it's really for coaching and ramping up sales reps and making them better. And then that's the basis. Then on top of that, we have forecasting. Because you figure you have all this. We call it the state of the reality data. This is we call ourselves the reality platform. So once you have all that interaction data and everything going on there, you can sort of marry it with the CRM data and start figuring out, OK, for each of these sales opportunities, what's the probability it's going to close? How much is it going to close for? When will it close? And really come up with an accurate forecast. And on top of that, there's always engagement as well. So who should I talk to next? What should I talk to them about? There's a whole bunch of sales engagement tools, but we really wanted to reimagine it really around our core data. And so it's really that core interaction data and how we can reimagine it all.

‍

Tarush Aggarwal (02:05):
Yeah.

‍

Tarush Aggarwal (02:05):
Yeah. And so your customers are more B2B SaaS businesses. Is that the I think?

‍

Michael Tambe (02:27):
Yeah, we ourselves are a B2B SaaS business. We generally, we sell to companies with B2B sales forces. So, you know, tech is a big vertical, but it's not our only vertical.

‍

Tarush Aggarwal (02:39):
Yeah, that makes sense. And obviously, you're a very unique company because data in some ways is your product. You are a data platform fundamentally; the insights and analytics which come out of it get productized. So data is one of the few businesses where data is the product. But how do you then view this in a way where you have a data and analytics group which...

‍

Michael Tambe (02:46):
Exactly.

‍

Tarush Aggarwal (03:08):
...potentially sits outside of the data product, which is the company. So how do you view, what's the sort of, how does this sort of break up on a day-to-day basis?

‍

Michael Tambe (03:18):
Yeah, well, it's both extremely scary and extremely exciting because all the product leaders are masters in your space. And so for me, again, if I just think about myself, I started at LinkedIn. We had this amazing member customer data. So when I went to Amazon, I wanted to leapfrog because the retail signals are just 10x bigger. And so here, I wanted to go to a place that could really challenge, have amazing data. And so for me, it's like two things: one is how do I apply all that industry best practice in terms of metrics definition, root cause analysis, predictive models, like, you know, which accounts should we be prioritizing for sales? How should we be segmenting our customers? What are the churn signals in this with the kind of detailed AI technology that we have? So it's like being a best-in-class analytics and data science team, but also being amazing power users of Gong and trying to create and chart what I hope will be the future of sales analytics.

‍

Tarush Aggarwal (04:16):
Yeah, that makes sense. And what's the culture like at Gong?

‍

Michael Tambe (04:21):
I mean, it's a great culture, I would say. So it's interesting because we're built in Israel and sold in the US. And so we have a very kind of international culture. And so I think it's very direct. One of our operating principles or leadership principles is no sugar. Just tell me what it is. Don't sugarcoat it. Be very plain to the facts and all of that. Very action-oriented, but with a real sense of building for the long term and it's not just about getting customers to use the product, but we want to create raving fans. Like they have to love the product. Like our product leader always talks about, "Look, if you create workflows, you make people use the workflows, they'll use the product." That doesn't mean they love the product. We want them to love the product, not just use the product.

‍

Tarush Aggarwal (05:03):
Yeah.

‍

Tarush Aggarwal (05:03):
Yeah. And I know I've had a team in a previous gig. I worked at WeWork and we had a big function in Israel and we love working with our Israeli counterparts. Fully get the directness. Same time, you know, loving place. How is, you know, what's the culture like in the data team? And, know, the reason I ask this question is very often, so data is just like very multi...

‍

Michael Tambe (05:17):
It's refreshing.

‍

Tarush Aggarwal (05:29):
...faceted from the point of view, you deal with the entire business and we see these sort of data teams have their own quirks and how they do different things. How does the sort of culture of the stuff you're doing around predictive analytics and sort of GenAI, how does that kind of fit into the company culture?

‍

Michael Tambe (05:46):
Yeah. I mean, I think it's interesting because Gong is a data platform and there's so much that you can do in the app. In terms of building this internal analytics function, we've had to sort of establish the value proposition of having a team that has kind of industry best practices and can combine it because there's a certain level of self-sufficiency. Like Gong grew for quite a bit without having a really mature internal data analytics function because of the strength of that platform. And so it's a very, very slim team. Like if you think about what we have in the US between, you know, data analysts, data scientists, data engineers, and just the people who manage the systems, only eight people. It's a very, very small team. So we try to be very, you know, mapped to certain businesses and it's like, "Okay, let's really make sure we understand the strategy and for every top two or three things, let's make sure we're enabling that." Use a combination of internally built things and stuff that we can combine via API with our internal application.

‍

Tarush Aggarwal (06:13):
Yeah.

‍

Michael Tambe (06:15):
And so it's very much like each little function feels like a startup at various stages of maturity. And so it's exciting.

‍

Tarush Aggarwal (06:49):
Yeah. What's the overall company size?

‍

Michael Tambe (06:55):
I'm not allowed to give exact numbers, but you know, in the call of the 800 to 1200 range, if you will.

‍

Tarush Aggarwal (07:01):
So just ballpark, assuming it's 1,000. Typically, you see in a typical business, tech is about 20% of the company and about 20% of tech is data. So roughly 4% of the company is data. At 1,000 people, you would expect that to be about 40 people. So...

‍

Michael Tambe (07:19):
Yeah, well, there are some data teams in Israel, but yeah, we're a lean team.

‍

Tarush Aggarwal (07:26):
Yeah. And once you, you know, once data is your product and fundamentally you are a data platform, all of those bets are off because you have a lot of representation of data skills throughout the business. You know, analytics being a little bit slimmer than what's expected is to be expected. What does the data stack look like again, where you have so many data platform capabilities in the product. Do you still use a warehouse? What are some of the, you know, what's the day-to-day stack look like?

‍

Michael Tambe (07:56):
We do, we do. And we've gone back and forth on this, but we decided we do need the mature stack. So we have your typical Snowflake core data lake house. On top of that, we have typical Tableau visualization. And then a lot of what we're doing in terms of analysis and ways that we're plugging in is a lot of custom Python stuff. And so for us to be able to access data, little bit of internal stuff and be able to serve it to people, let's say via Slack or... within the application, reverse ETL in there.

‍

Tarush Aggarwal (08:27):
Yeah, so do you use any tools for reverse ETL or ingestion? Given that you're in Python, is there any dbt or do you have any orchestrator?

‍

Michael Tambe (08:34):
We do use dbt and there's a bit of orchestration as well. Think we do use a little bit of Workato for a lot of the reverse ETL as well, but yeah.

‍

Tarush Aggarwal (08:43):
Got it. Nice. You know, the predictive analytics side is obviously very interesting. Snowflake, Databricks, all of these folks are super sort of, sort of supercharged on, you know, time and effort and sort of talking about it. How does, what is the data science and predictive analytics and sort of Gen AIs, you're looking at those worlds? Are those powered on top of Snowflake or how do you look at sort of powering some of that technology from the analytics perspective, not from the quality of the product?

‍

Michael Tambe (09:11):
Yeah, maybe. I'm just an old school guy. I like scikit-learn. Open source Python packages, working on that. We've played around a little bit with some of Snowflake's ability to kind of productionalize some of the Python libraries. We're figuring out if that's where we want to go. But again, I think it's going to be very API-driven. So for example, we use AWS Bedrock and so...

‍

Tarush Aggarwal (09:16):
Yeah.

‍

Michael Tambe (09:38):
We can play around with some of the foundation models that we want to play with as well as interact with some of the core AI services of Gong via API.

‍

Tarush Aggarwal (09:46):
Yeah. Yeah, that makes a ton of sense. What's one achievement of your team which you're really proud of? Whether it maps back to a high-level business objective or culture. We see a variety of answers to this question because data teams are so unique and different. But I'm very curious for a company like Gong, what does an achievement to you look like?

‍

Michael Tambe (10:12):
Yeah, so I'd say, and I've been in the role for about eight, nine months, the first one that I think is really coming to mind with me, which is a project that started before me, but I've been able to kind of see it all the way through, is really around retention strategy. So in 2012, I think the entire SaaS industry saw some headwinds. So there's a lot of focus on really understanding drivers of retention. And so it was really about combining this sort of best-in-class predictive analytics in terms of, "Okay, let's look at not just usage, but usage by persona, usage by different types of things to understand what patterns of usage and who they're using and who does them are really indicative of value creation for the product as well as retention." So to be able to fine-tune on that and then deploy it within the sales team's operating rhythm. So...

‍

Tarush Aggarwal (11:12):
I was just curious because this is retention, right? So why is retention with the sales team and not with the sort of CS team?

‍

Michael Tambe (11:18):
It is well, it's a mixture of both, right? The sales team owns success, the CS team owns that. They both own the renewal.

‍

Tarush Aggarwal (11:26):
Got it. Interesting. And what does the sort of practical application, so an ROI perspective, look like? How do you measure this? And what does success look like in a, obviously, one metric of success around anything to do with retention is obviously lower churn. But how do you really measure this? And what impact have you seen this as driven?

‍

Michael Tambe (11:48):
Yeah, so I think the impact measurement is something we're just starting to get into again. We're relatively nascent in this. So just being able to see the health in terms of usage, being able to see this go down. We have had some conversations about A/B testing and all of that. It's always a little tricky with B2B sales about how much you're going to try to do some sort of causal method versus like pre-post, but that's an open conversation that we're having.

‍

Tarush Aggarwal (12:12):
Yeah, if how in order to execute a project like this, you know, given that you have an eight-person group, which is broadly speaking across all of data and analytics, what's like the, you know, what's the sort of resources needed on your side to basically go deliver something like this? You know, is it group ownership? There are one or two people focused on this? How does what's the sort of day-to-day look like working on an advanced analytics project at Gong?

‍

Michael Tambe (12:44):
Yeah. Yeah. And again, this is just us being a very slim team. So I'd say probably, you know, I have, you know, if I think about on the data science and analytics side, you've got about two in sales, two in post-sales. So there's going to be two in this kind of post-sales area that's thinking about predictive retention and 50 to 60% of the time is going to be on like core foundational data metrics enablement. And the other, you know, 40% of the time is on things like this. And so it's, you know, it's the fact that like, these are well-known like scikit-learn has a mature algorithm. So you can take that first set of features, run it through and get a quick model where it takes a lot of time is trying to figure out the next set of features to really optimize. And this is where Gen AI can really help because you can build a model, you get false positives, false negatives. If you can just ask the model, "Hey, why did this account churn? It had great usage," or "Hey, why was this account retained? It had horrible usage," and you can actually really quickly get to that level of like generating new hypotheses, getting into the data, making it work. And so you can already start to see a little bit more of the efficiency of these tools coming together.

‍

Tarush Aggarwal (13:25):
Yeah.

‍

Michael Tambe (13:41):
Yeah. And does the company have a big appetite to really go play around with AI? Obviously, we're living in a world today where a lot of people have AI anxiety.

‍

Michael Tambe (13:43):
Yes. That is like a completely... there's another team that does that. So we, in AI, we use it for photos, for editing, for models wearing lingerie so that we do not make more photos and we do not have more photo shoots. We have a set from which we get inspired and then extrapolate. And that saves money. That's something, yeah. The face is real, that's not a problem, but the rest can be a little bit manipulated in different outfits. There are also some other AI opportunities for optimization, but I don't know if I can talk about this particular one, because it's not mine, but you can do... In the area of marketing, I would say there's a lot of opportunity as well.

‍

Tarush Aggarwal (15:26):
Yeah.

‍

Michael Tambe (15:43):
You can build a customer lifetime value and they can do bidding on the margin or on the RFN segmentation group or on the entire revenue or on so on and so forth.

‍

Tarush Aggarwal (15:49):
Yeah.

‍

Michael Tambe (15:49):
Yeah. That makes a ton of sense. What's one area which has been sort of challenging? It's a small data team working across a large group. What's one area which is a challenge which you're extremely focused on?

‍

Michael Tambe (16:06):
Volume. We all have a lot of projects, so the to-do list is unlimited. So one has to prioritize the things that are impactful now or in the next month and work strategically, prioritize strategically. Because I also have curiosities, like I don't know, two weeks ago I wanted to know the return reason for everything and I want to see what, why, what, why. Like nobody did a deep dive if there is a...

‍

Tarush Aggarwal (16:15):
Mm.

‍

Michael Tambe (16:36):
Finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:22):
Finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
Finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
...finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
Finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
Finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
...finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
...finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
...finer trends or if you, let's say, that you spread the vouchers in the city center for Berlin and this is a one-day event and if you put it in an MMM model that will count as a one-month event if it's a monthly granularity. If it's weekly, then it's weekly, so you're reducing a little bit this the effect of it because while positive, it's also overinflated if you go to the monthly granularity.

‍

Tarush Aggarwal (16:36):
Yeah.

‍

Michael Tambe (16:36):
...so it’s about that.

‍

Tarush Aggarwal (16:40):
Yeah. Yeah. So it sounds like you have the flexibility to go sort of try out all of these things, which is awesome. How do you sort of prioritize? And if you're structured in a way which everyone is structured as a sort of senior consultant, there are sort of many different trains of thought and opportunities. How do you sort of prioritize to make sure that what everyone's doing is very aligned?

‍

Irina Ioana Brudaru (17:06):
Would say this is something we're still working on because we are so, I mean, the data team is centralized. A lot of other services are centralized. Marketing is obviously not because it's also so cultural and so different in every country, what works and what doesn't. I would say project managing people from different cultures, the companies themselves, different goals, different ways of communication. My god, know, the Dutch are direct. I love it. To me, it's great. It's clinical. It tells me what to do, while maybe the French are not. And you know, you have to adjust the alignment and the project management based on people. People are the hardest part. Data is not.

‍

Tarush Aggarwal (17:38):
Yeah.

‍

Irina Ioana Brudaru (17:38):
Yeah. Yeah.

‍

Tarush Aggarwal (17:38):
Yeah, it sounds like the culture is very open to it and the diversity in this case really helps and creates this multi-faceted culture which it really sounds like. The challenge on one side over there is it's a little bit free-for-all right now and a little bit less top-down. These are the focuses across the group but it sounds like the awareness around how to go do this is coming.

‍

Irina Ioana Brudaru (18:17):
Yeah, when we're going, we're going there. I mean, the companies go through different phases. First, it's like, let's set up the business idea and then, it works. Let's make it the skill. Let's put more volume behind it. And that works. And then, you know, companies get to a certain stage when you can no longer wing it and you need to bring up like experts—experts in CRM, experts in personalization, pricing, and so on.

‍

Tarush Aggarwal (18:30):
Yeah.

‍

Irina Ioana Brudaru (18:45):
So right now it's at that tipping point where things become a little bit more specialized and we need to be more focused on where we go and how we do things.

‍

Tarush Aggarwal (18:53):
Makes perfect sense. As you look at the future, what are some areas of investment, particularly on the data side, which can have outside impact in the business?

‍

Irina Ioana Brudaru (19:04):
So many. More people would be the first one. Second one would be like an alignment of what type of tracking we do in the company so that we don't have one person having to learn six tools. So alignment, coalescing the tools. Another part, there is still a lot of volume in order to support the stakeholders in the day-to-day business decisions.

‍

Tarush Aggarwal (19:20):
Mm.

‍

Irina Ioana Brudaru (19:32):
My dream would be to have reports customized per role sent every day about your performance of this, this, this, this. Being, what's it, being served with the things that you have to do every single day in related to your business. I've experienced this in a team at Google where we had such tools that would give us alerts and tell us what we need to do that day that's very urgent. Or if something broke.

‍

Tarush Aggarwal (19:54):
Yeah.

‍

Irina Ioana Brudaru (20:01):
And I feel like in this business, we're not there yet, but that is where I want to go. I will make business so happy. Just transparency. And so we need more people for that. We need more people to build alerts and dashboards and pipelines.

‍

Tarush Aggarwal (20:04):
Yeah.

‍

Tarush Aggarwal (20:04):
Yeah? No, that makes a ton of sense. If you're watching the show and you're looking for a new gig, the EQOM group working with Irina sounds awesome. You should definitely give it a shot. Irina, thank you so much for being on the show and adding value to our listeners.

‍

Irina Ioana Brudaru (20:26):
A pleasure.