Kill the System-of-Record Slide

Jason Lemkin published the cleanest teardown of the system-of-record moat I have read this year, and I want to argue with the last paragraph.

Not the diagnosis. The diagnosis is right and the numbers are his. The destination.

The short version

"We are the system of record" is a retention fact, and it is being presented in board decks as a growth claim. Jason Lemkin's ServiceTitan example shows the gap: ServiceTitan shut off Podium's integration for roughly 1,000 shared customers, kept essentially all of them, and added no new revenue. His conclusion is that growth moved down a layer to Snowflake and Databricks. Mine is that Zero Copy proves the data layer is rentable too, and that Stripe's platforms data points at the real test, Eric Noeth's "what breaks the day the customer turns it off." That is a loop question, not a storage question. Kill the slide and replace it with two lines.

Lemkin's facts

ServiceTitan delisted Podium after a nine-year partnership. Roughly 1,000 shared customers, a partner Lemkin puts at $100M in AI agent ARR built in under 24 months, much of it inside ServiceTitan's own customer base. ServiceTitan kept essentially all of those accounts. The contractors stayed because their jobs, invoices, customer history, and technician schedules all live in ServiceTitan.

That is what the record buys you. It is a lot. Renewal rates, pricing power, net revenue retention above 110%.

Then he looks at the growth rate, and this is the part worth sitting with. By his read of Salesforce's quarter, Agentforce Apps, the sales, service, marketing, commerce, and Slack business, the actual record, grew 7% in constant currency. Data 360, Headless Platform, and Other went from $2.95B to $3.68B. His line: the record layer grew 7%, the data layer grew 25%. Same company, same customers, same quarter.

And the largest system of record has already conceded the architectural point. Of the 52 trillion records Data 360 ingested in a quarter, 35 trillion came in via Zero Copy, meaning they never physically moved into Salesforce. Zero Copy volume grew 277% year over year. Salesforce Engineering's own Data 360 post from the same week reports 104.3 trillion connected records, 81.5 trillion of them Zero Copy rows.

The fastest-growing data motion at the biggest system of record is the one where it does not hold the data.

His conclusion from all this: systems of record convert into retention, not acceleration. Retention is the floor. The incremental dollar went to Snowflake (product revenue up 34%, NRR 126%, by his figures) and Databricks (over 80% growth at $7B ARR).

Cutting off Podium did not add a dollar of new revenue to ServiceTitan. It prevented the loss of some future dollars. He calls those different businesses, and they are.

Where I get off

If Zero Copy proves anything, it proves that the data layer is rentable too.

Follow the logic. The record used to be the moat because moving the data out was painful. Zero Copy is Salesforce saying the data can stay where it is and participate anyway. Fine. But the same property cuts the other way: if the record can be federated, the warehouse can be swapped. Snowflake and Databricks are growing fast right now because agent workloads generate an absurd amount of exhaust and somebody has to hold it cheaply. That is a volume story. It is not obviously a lock-in story, and Lemkin's own post is the evidence, since he is describing a world in which the data is increasingly decoupled from the system that produced it.

So "growth moved to the data layer" is a description of this year's revenue. It is not an answer to the question the board is actually asking, which is where the incremental dollar sticks.

I wrote about this a week ago from a different angle, in Context Is King. Written Context Is Rented. Switching costs didn't disappear. They moved, from the database to the decision record. The archive ports, and LLMs made porting it dramatically easier. What does not port is the earned record of what the product decided, on what evidence, and what happened next.

Stripe published the other half of the argument in the same week, and I think it is the more useful half.

Stripe's platforms

Stripe added more new platforms in the last three months than in the final six months of 2025. New platform businesses are up over 180% year over year in that period. Platforms that went live in January 2026 are reaching $1 million in payment volume faster than any previous cohort. This is the same industry that shed roughly $1 trillion in market cap in 30 days early in the year while the headlines said SaaSpocalypse.

Stripe's read on what the durable ones share is three traits. They run key workflows, scheduling or inventory or intake. They retain valuable business information, customer histories and pricing logic. And they connect day-to-day operations to money movement.

Notice that "hold the record" is one of the three, and it is the weakest one on its own. The other two are about action.

Then Eric Noeth of Advent gives the test, and it is the line I would put on the slide instead:

"A high-signal indicator of platform defensibility is what breaks the day the customer turns it off."

If operations keep running, the product is exposed. If claims don't pay, cars don't sell, trades don't settle, the product is hard to dislodge.

That is not a storage question. Nobody's trades stop settling because a database went away, they stop settling because a loop went away. Did the product act, did the act produce evidence, did the evidence change the next act, and can the customer see that chain. ServiceTitan passes the test not because it holds the contractor's history but because the technician's Tuesday is dispatched from it.

The version I shipped wrong

I ran product at Salesforce, across Marketing Cloud, Quip, and Slack, so I read the 7% with some history. The record is a real asset. The slide that presents it as a growth asset is the one I am asking you to kill.

I have also built the failure mode myself. Heidi's Signals layer held everything. Every source connected, every signal resolved and displayed cleanly, side by side. It demoed beautifully. It compounded nothing, because signals sitting next to each other on a screen are not correlated with each other, and a signal that never triggers an act never generates the evidence that would make the next signal smarter.

We held the data. We did not run the loop. By Noeth's test, the day a customer turned it off, nothing broke. That is the honest version of what "we are the system of record" often means in a product that has not yet learned to act.

Kill the slide

"We are the system of record" is a retention fact. Keep it in the retention section, where it belongs and where it is true.

It is not a growth claim, and your board can read the growth rate.

Replace it with two lines.

First: what breaks the day they turn us off. Not what they lose access to. What stops. If the answer is "they would have to export a lot of data," you have a retention moat and you already knew that. If the answer is "their dispatch stops" or "their invoices stop going out," write that down, because it is the actual moat and it has nothing to do with where the bytes live.

Second: what did the product decide last quarter that it could not have decided the quarter before. This is the compounding test. A record that grows is an archive. A record that changes what the product does next is a loop, and the loop is the only thing in this whole argument that does not port.

If you cannot fill in the second line, that is the finding. Not a reason to keep the old slide.

One thing to try this week: open the last board deck, find the moat slide, and rewrite it as those two lines. If the second one comes out blank, you have your next quarter's roadmap.

This one belongs to the running argument on SaaS to AI Business Models: software priced per seat is priced against a labor cost that AI removes, and the moat structure moves with the pricing model.

Related answer: Is being the system of record still a moat?

Sources: Being a System of Record Helps With Retention. But Alone, It Won't Equal Growth, Jason Lemkin, SaaStr, September 2026. SaaS platforms are surging despite the SaaSpocalypse, Stripe, September 2026. How Data 360 Builds Trusted Context: The Enduring Layer for Enterprise AI, Raveendrnathan Loganathan and Tobias Muehlbauer, Salesforce Engineering, September 2026.