Tech is on the Brink of a Gross Retention Apocalypse & a Customer Success Renaissance

I’ve said it before and I’ll say it again: ARR traction is a major red herring in the world of AI-native startups.

Jamin Ball’s 2024 Clouded Judgment piece on the ERR (experimental runrate revenue) vs. ARR conundrum was something of a shot heard round the world for me. His argument about experimental budgets versus enduring revenue—that companies were moving quickly to experiment but that once prototypes were validated, they’d consider alternatives in the name of evaluating scalability, cost, compliance, etc.—resonated deeply, especially when compounded by other mounting anecdotes I was hearing in the market. Around the same time, Menlo Ventures reported that 60% of generative AI investment dollars were coming out of innovation budgets, and there were daily headlines about companies mandating employees to level up their AI proficiency. Operators suddenly felt deep pressure to avoid professional irrelevance, resulting in rampant experimentation and ubiquitous pilots.

Meanwhile, I was reading news about startups raising massive Series A/B fundraises. As recently as last month, Carta reported Series A valuations being up 20% YoY and Series B valuations up 15% QoQ. But when I regularly reached out to the customer logos listed on those startups’ websites to better understand how they were using these innovative new products, I’d hear: “We’re probably churning.” I quickly became a squeaky wheel about seeing smoke in the house with respect to looming churn risks—so much so that there was a running joke at Primary that I’d tipped off TechCrunch about the 11x story (for the record, I most certainly did not!).

I’m of course quite bullish on the opportunity set AI will unlock at a macro level, but I have become worried that churn risk will pose an existential threat to the broader ecosystem. Graduation rates and time between rounds are already worsening; will AI experimentation further compound that reality? But despite any worries, I am equally energized. Because I believe that churn risk will also pave the way for a customer success renaissance. What’s old is new again: in the same way customer success initially became consequential in the throes of the on-prem to SaaS evolution and the surrounding implications for lower switching costs, it will once again have its moment as we move from traditional SaaS to AI-native solutions. Companies simply cannot build enduring enterprise value without delivering enduring customer value. In a world of rampant experimentation, a focus on proof of adoption, ROI realization, and value delivery is simply non-negotiable. And the industry is acknowledging this urgency today. Friends such as John Gleeson have championed the concept of forward deployed customer success. Board members are appropriately freaking out about the lag from CARR to ARR and downstream implications.

But truth be told, a year and a half later, I don't think much has actually been done to address the ARR vs. ERR conundrum. Yes, companies have figured out how to navigate contractual levers so they can be more intellectually honest when they write the letters A-R-R. They know how to avoid opt-out clauses, how to accelerate conversion from proof of concept to an initial 12-month contract, and so forth. But the jury is still out on their ability to drive enduring customer value. Many of the AI darlings are PLG or hybrid PLG/sales-led, not enterprise-first, meaning the famous “E” (Economic Buyer) in the MEDDPICC qualification may still be missing in action (an automatic DQ for proclaiming enduring customer value in my opinion). And regardless of the go-to-market motion at play, most AI breakouts don’t have the benefit of rich customer history or learnings about what it takes to drive healthy mutli-year renewals.

So here's my prediction: it may not be 90 days from now, it may not even be six months from now, but the gross retention apocalypse is 100% coming.

Why am I expressly naming this as a gross retention apocalypse? Because most companies can cultivate stickiness with some segment of their installed base. I've recently been waking up in cold sweats about my early days at Sailthru when we were navigating some scaling challenges. Our net retention rate was well over 120%, but our gross retention rate was 60% (not a typo!). There was a segment of fast-growing startup customers driving the NRR curve, but we were simultaneously losing trust with other customers and logo churn proved to be a lagging metric. I frequently talk about the importance of presence of strength versus absence of weakness, but not when it comes to net retention: net retention must be considered in close concert with gross retention.

And yes, for enterprise solutions with robust implementation processes, the switching costs are real: when implementation takes months and reinforcement learning is mission-critical to customer value, lock-in is indeed quite possible. AI-native businesses are also throwing bodies at the problem with forward deployed engineers and so forth. But it takes deliberate company focus around customer success to consistently generate value and more importantly in today’s competitive climate, to do so expeditiously.

For those companies with limited historical data, the looming gross retention apocalypse is particularly daunting, but there are absolutely leading indicators to track to mitigate risk. Time to value (“TTV”) is paramount, but companies need to calculate it correctly. TTV cannot be a measurement of "go live" because let’s be honest, Phase II never happens. Companies must understand the "sticky drivers" of long-term retention—and accelerate customers’ paths to achieving those tipping points and thresholds. Implementation is everything these days, and a rare place where I encourage founders to do unscalable things. (Editor’s note: if you know a startup trying to disrupt traditional B2B implementation / PS workflows, send them my way!) 

Another metric I'm tracking in my own portfolio is time to expansion: how quickly do enterprise buyers expand into adjacent use cases? What does that mean for cohorted net revenue retention rates even in year 1? Companies such as 1mind and Lyric are putting impressive numbers on the board on this front.

And of course there’s customer ROI. Most conversations about AI these days include some reference to MIT’s recent State of AI in Business report and the POC chasm, but a few things in that report drove me crazy, including this line: "...these tools (referring to ChatGPT, etc.) primarily enhance individual productivity, not P&L performance." Bullsh*t! Every business initiative must be tied to a P&L outcome. If the customer can't get there on their own, it's on the vendor to help the customer build the business case and map the ROI. Murky ROI is guaranteed gross retention risk. (Editor’s note: An easy hack for CS teams is to beat customers over the head with ROI and value explanation. I recommend at least a weekly report sharing what's been accomplished in the past 7 days. Customers will undeniably argue the attribution of the ROI calculation many times over, but perception is reality, and putting something in front of them is far better than nothing.)

A few weeks ago I spent an entire cross-country flight digging into ICONIQ's State of Software 2025 report—quite the field day!

One interesting visual for me was the juxtaposition of the AI-native sales funnel vis-a-vis non-AI companies. What was not explicitly named is that within the <$100M ARR segment, if you actually multiply out New Lead > Closed Won (the product of the first three metrics), AI-native companies actually perform worse (2.18% to closed won versus 2.25% for the non-AI group). Demo conversion is also worse but perhaps unsurprisingly, the AI-native companies outperform when a trial is offered—because a trial is certainly more powerful than a demo.

But what happens after these companies convert from POC into the initial paid contract? Do they stick around? Do contracts grow? If the company does not treat customer success deliberately and seriously, I assure you the answer is no. Needless to say, I can't wait to see this slide updated for logo and revenue retention rates at Years 1, 2, and well beyond!

Similarly, while new logo velocity for <$100M ARR companies appears eye-popping, I'm itching to see a regression analysis that compares early YoY new logo and ARR growth with lagging gross and net retention rates. I phoned a few analyst friends to see if anyone had seen any sort of analysis for new logo growth vis-a-vis lagging/cohorted NRR but came up dry. Sure, POCs create value, but will they offer enduring value? I've got my popcorn and will be ready to watch!

So, if funnel conversion hasn't fundamentally changed but AI-native new logo growth is still through the roof—are buyers just buying more stuff? Yes, they most certainly are! Buyers are in an era of rampant experimentation, and vendors are making it easier to try and buy than ever before. But here's the news flash: if you're easier to buy, you're also easier to replace.

Companies are certainly growing at unprecedented clips, as Bessemer pointed out in their recent State of AI 2025 report. Growth absolutely matters. There's a reason the growth-adjusted rule of 40 metric is even a thing: growth always commands a premium over profit (today that premium is approximately 2x). But growth alone does not guarantee a formidable company.

Customer value is the key ingredient in unlocking enduring enterprise value. As I always say, if you do everything in your power to ensure your customers are successful with your product—within your financial constraints, of course (hello to my CFO readers)—you'd have to catastrophically screw something up to not succeed as a business. Plus, it’s well-understood that buyers rely on their networks for market intelligence, and referrals are the fastest and easiest deals to close. 

Obsess over your customers. Don't rest on net retention laurels—if you are losing customers, understand why. If you're evolving the ICP, that's fine and dandy, but make sure the sales team is actually operationalizing that and changing their sales process to only sell to the latest and greatest ICP.

At Primary we’ve been talking extensively about the importance of jaw-dropping customer experiences (JDCEs – if you want a primer, I recommend this recent Invest Like the Best episode with Neil Mehta from Greenoaks). When I consider the concept of JDCE, I think about it from the lens of the “whole product” perspective introduced in Crossing the Chasm, not simply a dead-simple UI or a killer demo…which is an important reminder that customer success is everyone's job. What can every team do to deliver a JDCE? Our portfolio company Lyric runs a successful “POC Factory” team that can stand up POCs for F1000 enterprises in a single business day. Other portfolio companies have moved implementation under the Product/Engineering teams to treat onboarding with the same level of importance as any other product feature or launch. As my friend Harini Gokul, Chief Customer Officer at cybersecurity unicorn Entrust, offered at Pavilion's recent Women's Summit: "AI is just another means to an end of driving end value to the customer."

Growth matters but is one of several important fundamental metrics that must be evaluated in tandem, including time to value and gross retention. Those who weather the gross retention apocalypse will be the ones who treat customer success not as a function, but as a core operating philosophy.

Become a Topline insider by joining our Slack channel.

We want to hear your feedback! Let us know your thoughts.

This Made Us Think: Curated by Kyler Verney.

Imagery by Neil Topinka.