Turning Support Tickets Into Revenue - Lessons from Hubspot, And Ernst & Young

There is a room in almost every company that knows exactly why you're losing money. Yeah. And it knows why your customers are angry too. Right. Exactly. It knows what features they actually want. It knows who is about to quit. But here's the crazy part. The executives almost never step foot in it. We are talking about the support department. It is easily the most undervalued asset in the corporate world right now. We tend to think of support as, uh, as the place where problems go to die. But today's deep dive argues that it's actually where strategy should be born. Yeah. And to figure out how to make that happen, we're looking at a fascinating panel discussion. It was titled turning support data into executive insight. Great panel. Really diverse perspectives on it. Totally. It was hosted by Ryan Nichols over at D Y D X capital and he brought in two absolute heavy hitters for this. We've got Somia, who is the VP of revenue operations at HubSpot. So she's the practitioner and the trenches. Right. And then VJ, a partner at Ernst in Young for AI and data. He's the consultant seeing this across dozens of enterprises. So you have the investor, the practitioner and the consultant all in one room. And they're all circling the same, uh, the same core tension, which that you have. The C suite, right? And they are completely obsessed with revenue, with growth, with risk mitigation. And then literally down the hall, you have the support team dealing with tickets and bugs and people who forgot their passwords. Exactly. And the mission for this deep dive is to figure out why those two groups are speaking completely different languages. And how bridging that specific gap can actually turn a, what we usually call a cost center into a literal revenue engine. I really want to start with a sheer scale of this because Somia dropped a stat early on that honestly stopped me in my tracks. The HubSpot stat. Yeah. She said that at HubSpot, 60% of all customer interactions happen in support. 60%. It's a massive volume of data when you think about it. But is it, I mean, is it valuable volume? Sure is 60% of the interactions, but isn't a lot of that just people venting or asking where a button is. I'm just not sure I'd call that strategic executive insight. See, that is the common trap. That's exactly where most companies go wrong. We completely dismiss it as noise. But Somia calls this the honesty factor. She argues that while it's 60% of the volume, it actually represents 90% of the honesty. Okay. Break that down for me. 90% of the honesty. We'll just think about the difference between a sales call and a support ticket. Oh, I so in a sales call, I'm guarding my wallet. I'm posturing right. You're negotiating. You were performing a little bit. Yeah. You might not admit that you don't really understand the software because you don't want to look bad to the sales rep or to your boss who's on the call. Right. But when you are writing a support ticket, the sale is over. You're frustrated. You're alone at your desk at 4 in the afternoon. You were telling them exactly what is broken, exactly what is confusing you and exactly what you need to do your job. So the venting is actually the most raw, unfiltered form of customer feedback you can possibly get. Precisely. It's unfiltered reality. Okay. So if this data is so honest and there's just mountains of it, why aren't CEOs reading these tickets every morning with their coffee? Like, why is there this huge disconnect? Because they literally cannot understand each other. So me used a brilliant analogy here during the panel. She compared it to speaking with her son. She speaks to him in Tamil, which is her native language. And he responds to her in English. Okay. So they're communicating right. They understand the gist. But the mode of transmission creates this constant barrier. So in this corporate version, support is speaking Tamil and the C suite is speaking English. You know what? Yeah. Think about the vocabulary. Support speaks in tickets, right? Bugs, SLA's average handle time, time to resolution. Right. Metrics about efficiency. Exactly. But executives, they only speak in revenue growth, risk and churn. I see. So a VP of support walks into a quarterly board meeting and proudly says, Hey, we kept our first response time under two hours this quarter. And the CEO here's always spent a ton of money fixing things that shouldn't have been broken in the first place. Wow. Yeah. That was a complete translation failure. And that is exactly where the opportunity lies. The data isn't missing. It is just untranslated. The job isn't to stop fixing the bugs. It's to take those support interactions and translate a quote unquote feature request into a product road map opportunity or translate login issues into churn risk. Yes. Yeah. You have to take that raw support dialect and turn it into the language of the C suite, which is always at the end of the day about money. That translation piece seems like the hardest part though. I mean, most companies, I know they just rely on NPS net promoter score, you know, to figure out of customers are happy. If the MPS is high, the CEO sleeps well. And the panel absolutely tore that idea apart. Really? But everyone uses MPS. You get the little survey in your email. You click the smiley face to the frowny face. Why is that bad? It isn't bad. And apparently it's just too late. The panel argues that metrics like NPS and CS cut customer satisfaction. They're lagging indicators, lagging, meaning the damage is already done. Exactly. Think about when you actually take the time to fill out one of those surveys. Usually it's way after the interaction is over or maybe it's just a random quarterly blast. If a customer cares enough to give you a zero on an NPS survey, they have probably already mentally checked out. So you're reading an obituary instead of giving a medical diagnosis. That's a great way to put it. You need a leading indicator. Something that tells you the customer is getting sick before they actually die in churn. Okay. So how do we do that? What's the leading indicator? Vijay from EY laid out a really specific pattern for this. He calls it the trifecta for churn. It's a pattern hidden in the support data that predicts a customer is leaving long before they ever get an NPS survey. Walk me through it. What are the three signals in the trifecta? It's a specific counterintuitive sequence. First product usage dips, second sport ticket volume spikes and third sentiment drops. We hold on. Let's back up to that second one. If usage dips, if they are using the product less, why would ticket volume go up? Right. It sounds backwards. Yeah. Usually if I stop using something, I just stop complaining about it. I just ghost the company. I don't file more tickets. And that is what Vijay calls the death rattle. This is the key insight he found across all these enterprise clients. Yes, they are using the core product less. They aren't logging in every day to do their normal work, but they are trying to do one last, very specific thing. Like what? Usually something like exporting all their data or trying to figure out how to migrate their contacts to a new system. And because they're already checking out, they're super impatient. They hit a snag. They get frustrated and boom. They file an urgent ticket. I got it. So that spike in tickets isn't help me use this cool new feature. It's helped me get my stuff out of your system. Exactly. Or it's something like, why am I still being billed for these seats? That high friction exit attempt is the loudest signal you will ever get. That makes so much sense because if you just look at usage, you might miss it. If you just look at the volume of tickets, you might think, Oh, wow, they're super engaged with the platform. Right. But if you cross reference them and see usage down while tickets are up, that customer is gone unless you intervene immediately. That context is everything. And speaking of context, Vijay also mentioned something about who is doing the complaining right? Yes. This is the other major flaw with relying on NPS and NPS score going up 10 points across the board. Looks like a massive win in a board deck. But Vijay asks a critical question. Who actually gave that score? Was it the intern or the CEO precisely? Was it the end user who logs in once a month to check a dashboard? Or was it the executive decision maker who actually signs the renewal contract? Support data gives you that exact context. So if the decision maker is the one opening tickets about billing errors or reporting downtime, that is a five alarm fire. Even if the 99 end users are perfectly happy and giving high NPS scores. Exactly. You have to translate who is experiencing the pain, not just how much pain there is overall. It's easy to talk about this in theory though, right? Just translate the data, look for the trifecta. But surely it's a lot harder when the house is actually on fire. Samia shared a story from HubSpot about exactly that, didn't she? She did. And it's perfect case study. This is about three years ago. The macro market turned the economy tightened up and companies everywhere went from growth at all costs to cut costs immediately. Right. Everyone was slashing budgets. So for HubSpot that meant churn was probably starting despite big time. So they weren't just sitting around looking for cool insights for fun. Yeah, they needed to stop the bleeding immediately. So what do they do? They launched an initiative internally called beat the plan. Samia went digging deep into the historical support data to figure out exactly why customers were leaving. And she found a staggering correlation. What was it? She found that 50% of the customers who churned had a specific type of support interaction shortly before they actually canceled their contract. 50% that has a massive chunk. What were they asking about? It goes right back to that death rattle concept we just talked about. The tickets were requests for migration assistance, exporting contacts, data portability questions. But wait, before this beat the plan analysis, how were those tickets being handled? They were just treated as standard technical support tickets. Seriously, a support rep sees a ticket asking, how do I export my entire CRM to a CSV file? And they just think, Oh, easy question. Here's a link to the help article ticket closed. And the rep thinks they did a great job. They close the ticket in 10 minutes. Their handle time looks amazing. But in reality, they just politely held the door open for the customer to walk out and leave the company. Wow. The translation failure in real time, realizing that those aren't technical questions. They are glaring churn signals. So once they realize this, How did HubSpot change the playbook? >>Instead of just answering the question and closing the ticket, those specific types of interactions started triggering a save initiative. >>Like an alarm goes off. >>Exactly. The system would flag it. Maybe instead of just a generic support reply, a customer success manager reaches out proactively. They say, "Hey, I see you're trying to export your entire database. Is there a problem? Is there a feature you're missing that we can help with?" >>So, you completely changed the posture. You move from being a librarian, just handing out information to being a firefighter, actively saving the account. >>And that right there is the fundamental shift from tracking support metrics to tracking revenue metrics. >>Speaking of metrics, we really have to talk about how we measure this pain. Ryan, the host of the panel, asks them for the one metric that matters most. Vijay voted for Effort Score, which is pretty standard, right? Like, how hard is it to do X? >>Yeah, customer effort score is very common. But Somia has a different one that sounds a lot more visceral. >>The pain value index. >>Yeah. >>PVI. >>Yeah. The pain value index. >>I mean, it sounds like something out of a medieval torture chamber, not a modern sauce company. >>It really does. But it is brutally honest, and that's why it works. It measures the complexity and the friction of using the product or interacting with the organization. But the real difference between PVI and a standard effort score is where that data actually goes. >>Okay, where does it go? >>It bubbles straight up to the product team. >>Not the support managers. >>Right. If the PVI is consistently high for a specific feature, it is no longer the support team's job to just handle the complaints better, or write better health articles. It formally becomes the product team's job to fix the root cause. >>That completely closes the feedback loop. Instead of support just acting as a sponge absorbing all the customers' pain, they weaponize it and send it back to the engineers who built the flawed feature in the first place. >>Which is how it should be. It forces the organization to actually fix the product, rather than just hiring 100 more support reps to constantly apologize for it. >>Okay, so we've covered defense pretty thoroughly, preventing churn fixing pains, stopping the bleeding. But this next part of the panel discussion is what I thought was the most futuristic. Can the "complete" department actually play offense? >>You mean generating net new revenue? >>Exactly. Can support actually sell? This is definitely the new frontier, moving from save to sell. And it relies heavily on what the panel called the Ambient Agent. >>The Ambient Agent? What exactly is that? >>It's essentially AI working silently in the background to support conversations. It's not a chatbot talking to the customer. It's listening to the human-human chat. And it's not just looking for keywords, it's listening for intent. >>You need a concrete example of that. How does an AI listening to a chat turn into actual money? >>Okay, let's say a customer is chatting with support and they ask, "Hey, is there a way to set up single sign on for my team?" "We're growing and we need SSO." But that feature is only available on the enterprise plan. And this customer is currently on the basic pro plan. >>Okay, so normally the human support rep would just say, "No, sorry, you need to upgrade to get that." And maybe they drop a link to the pricing page. It's an awkward interaction because support reps generally hate selling. They just want to fix things. >>Exactly. And most customers don't click that link. But the ambient agent is listening. It detects that intent automatically. It recognizes user wants SSO and SSO equals enterprise feature. It can then automatically generate a sales qualified leave and SQL and route it directly to the sales team with all the context of the chat attached. >>So the support rep doesn't have to be a salesperson at all. They don't have to do an awkward pitch. The AI does all the prospecting behind the scenes. >>Right. The rep just answers the question honestly and the system handles the business opportunity. And Vijay gave an even better example of this in action involving a client of his in the communications space, a CCAS provider. >>TCAS, so that's contact center as a service, right? >>Yep. So they provide phone systems for businesses. >>Oh, this is a miss calls example. >>Yes. They use their support data to show customers exactly how many phone calls they were missing after business hours. The ambient agent could look at the logs and flag it. So a message goes out saying, "Hey, you missed 50 calls last night between 6 p.m. and 9 p.m." >>Honestly, that is terrifying for a small business owner. Every missed call is lost revenue. >>It's terrifying, but it is also an immediate checkbook opener. The pitch basically writes itself at that point. If you buy our AI receptionist add on, you would have captured those 50 leads instead of sending them to voicemail. >>Wow. >> Vijay noted the ROI on that was so incredibly obvious. They didn't even need a human salesperson to make the call. They just said an automated email with the hard data and a link to upgrade. >>You literally lost money last night. Click here to stop losing money tonight. >>And people clicked. >>Yeah. >>It turns raw support data into a highly targeted sales weapon. >>Now HubSpot takes this a step further though when it comes to the human element, right? So Polyvia mentioned they actually pay their support reps for this. >>They do. At HubSpot support reps have actual targets for generating these leads. And they get spiffed. They get incentivized with real cash bonuses just like the sales or success teams do when one of their leads closes. >>That is a massive cultural shift. I mean usually support gets a pizza party if they survive the month without a mental breakdown. Now they are getting actual commission checks. >>And so may have pointed out why it's such a good investment. Leads coming from the support team have the absolute highest conversion rate of any channel in the company. >>Because they aren't cold leads. The customer is already a user. And they are actively asking for the feature right in that moment. It's the hottest lead you could possibly get. >>Exactly. >>I can hear the listeners right now though. They're drying or walking the dog and they're thinking this sounds amazing guys. But my data is a complete mess. >>Oh yeah. >>Right. Like it's in silos. We have spreadsheets from 2015 that nobody understands. And I can't just snap my fingers and deploy an ambient agent tomorrow. >>And that is the harsh reality for 99% of companies out there. The panel was actually very grounded about this. They know it's hard. VJ's advice was extremely pragmatic. He said, don't boil the ocean. >>So don't go to the CEO and ask for 10 million dollars to rebuild the entire data stack from scratch. >>No, you'll get laughed out of the room. Pick a very specific segment to start with. VJ suggested looking at renewals coming up in the next six to eight months. Just focus on that specific group of customers. >>Okay. >>And then filter that group down even further. Only look at the ones of the highest lifetime value of the LTV. >>Got it. So you're looking at the absolute VIPs who are about to make a buying decision very soon. >>Exactly. Run a pilot program just on them. Look at their support tickets manually if you have to. Read through them and see if you can spot the trifecta patterns we talked about earlier. See if you can spot the death rattles. >>And Ryan and Somia added a piece of advice there that they actually borrowed from Paul Graham over at Y Combinator Right. Do things that don't scale. >>Yes. We get so completely obsessed with automation and AI right out of the gate that we forget to prove the actual business value first. So Maya's point was that it's okay to build the spreadsheet manually at first. Read the 50 tickets yourself. >>Because if you can walk into that executive board meeting and say, "Look, I manually track these 50 high value customers. I use support data to predict that 10 of them are going to churn. We intervened and we successfully saved eight of them representing a million dollars in ARR." >>Nobody is going to argue with you about the budget for the software after that. >>You have to earn the right to automate. You do that by proving the translation works on a very small scale first. >>And it really all comes back to that core mission of the panel. You aren't asking for money to, quote-unquote, "improve" support response times. You are asking for budget to secure and expand revenue. >>Because that is the only language the C-suite actually speak. >>So we've gone from the Tamil and English language barrier to the death rattle of churn to the ambient agent actually selling upgrades while we sleep. If you had to distill this entire deep dive down, what is the one main thing you want people to take away from this? >>It's that support data is the literal rosetta stone of customer intent. It translates complaints into corporate strategy. The entire shift is moving from asking how quickly did we close this ticket to asking what did this ticket tell us about the future of our business. But I do want to leave you with one final slightly provocative thought that came out of the discussion on AI at the very end. >>That's here it. >>So if the ambient agent really works, if AI can automatically translate frustration into sales leads. And if it could automatically bubble up product flaws directly to the engineers to fix, do we eventually reach a point where the support ticket, as we know it just completely disappears? >>You know more tickets at all. >>Think about it. A ticket is by definition a reactive cry for help. >>Yeah. It means the system failed and the user had to manually raise their hand and fill out a form. If the system becomes smart enough to anticipate the need to offer the SSO upgrade before the user even asks or to fix the bug before the next user reports, it does support stopping a department entirely. >>It just becomes the central nervous system of the product itself. >>Exactly. We might be moving toward a world of complete anticipation, not reaction. The best support is the support you never actually have to ask for. That is a wild future to think about. But until the robots do take over completely, I'd highly suggest you go give your support team a raise or at the very least start listening to what they have to say. >>It's absolutely the smartest money you can spend. >>That's it for today's Deep Dive. Thanks so much for listening and we'll catch you on the next one.