Your Martech Renewal Now Has One Question: Can You Rip It Out?

Look at the software line on your marketing budget and ask which of those tools you could cancel on Monday and rebuild yourself by Friday. That list is about to get expensive for the vendors on it, and a seed investor who has been on the other side of the table for eighteen years just said so out loud.

David Frankel is Managing Partner and co-founder of Founder Collective, which seeded Uber, Coupang, PillPack and Airtable. On August 8, four days after Airtable agreed to sell to Bending Spoons for $1.28 billion in cash against a 2021 peak valuation over $11 billion, he sat down with Harry Stebbings on 20VC. Fund one still holds Airtable, he said, at [00:41:23].

He was asked whether the AI wave produces casualties. At [00:20:00]:

“will there be road kill from this wave? Oh my god, there’s going to be a lot.”

Then, at [00:43:29], he gave the rule for telling the casualties from the survivors. It isn’t features and it isn’t AI strategy:

“the more embedded you are, I think the more overdone that SaaS apocalypse may be.”

And the other half, at [00:43:54]:

“The less embedded, clearly, right? The easier you are to kind of turf out and play around with Claude, you name it.”

That is a purchasing test, and marketing owns more software than any other department. Embeddedness isn’t how much you like a tool or how many seats it has. It’s how much of your operation would break, and for how long, if it went away tomorrow. Your CRM with nine years of contact history and every integration pointed at it scores high. The tool that finds internal linking opportunities, or reformats a report, or watches a competitor’s pricing page, scores near zero. It has your login and a monthly invoice, and everything it knows arrived this week.

The evidence that the second group is already being cancelled isn’t new and it isn’t anecdotal. Chiefmartec and MartechTribe’s Martech for 2026 report found a 35% YoY decline in mid-market renewals for single-function martech tools. Eric D. Noren’s analysis of the chiefmartec census found 1,367 tools removed from the industry in a year, 51.7% of them companies founded between 2010 and 2019, and 80% of those with fewer than 50 employees. Noren’s summary of what killed them:

“The 2010–2019 SaaS cohort lost an architecture fight, not a feature fight.”

Christopher Penn, co-founder and chief data officer at TrustInsights.ai, gave MarTech the version from inside a services business in April:

“I was talking to someone this morning, they work in a marketing agency, and their agency has replaced 80% of their software subscriptions.”

Penn also named why the usual vendor response doesn’t work:

“There’s no such thing as making it defensible if you’re trying to defend software.”

Now the part where the investor’s own example doesn’t survive a check, which is worth doing because he invited it. Making the case that the SaaS selloff is overdone, Frankel reached for Veeva, the pharma industry’s system of record, and said at [00:43:08] it’s “at a $30 billion market cap now” and has come down “I don’t know how at least 50% or more.” He flagged his own uncertainty, and he was right to. Veeva’s market cap was $37.98 billion on August 9. Its highest year-end value on that record is $41.23 billion, in 2020. It bottomed at $25.12 billion in 2022 and has climbed every year since.

So the most embedded company he could name isn’t 50% down. It’s roughly where it was six years ago, having recovered about half its trough. Which strengthens his argument rather than weakening it, and tells you the “SaaS apocalypse” isn’t one event happening to one category. Deeply wired-in software is flat. Loosely wired-in software is being consolidated at a ninth of its peak valuation or shutting down. It’s the same split we found when platforms started charging for outcomes instead of seats: the vendors who could prove they held something charged more, and the rest discovered they were features.

The mechanism is simple enough to explain at a budget meeting. A tool that only transforms data has a job an AI assistant can now do on demand, so its price is capped by the cost of asking. A tool that holds data, and that other systems write into, is expensive to leave for reasons that have nothing to do with how good it is: migration, retraining, broken automations, a quarter of degraded reporting. Vendors have always known which category they’re in. AI made the customers know too.

Frankel’s caveat, at [00:44:01], is the honest one: “But I think we’re underestimating that last 5%.” The last 5% of a workflow, the exceptions and the edge cases and the thing that has to reconcile with finance, is where rebuild projects go to die. Anyone who has replaced a small tool with a script knows the script is fine until the month it isn’t.

Which gives a marketing team the actual test, and it’s cheaper than a procurement review. Before the next renewal, sort the stack into two lists: tools that hold data other systems depend on, and tools that process data on request. Defend the first list and negotiate hard on the second, because the vendors on it are watching the same renewal numbers you are, and the only card they have left is the last 5% you haven’t hit yet. If a vendor can’t tell you what breaks when they leave, they’ve answered the question.