Not subscribed? Sign up to get it in your inbox every week.

PostHog was a product analytics company for about five minutes.

Today it sells session replay, feature flags, A/B testing, surveys, error tracking, a data warehouse, and LLM observability alongside the analytics, and its public handbook lists five more products in flight with six queued behind them. Revenue roughly doubled over the past year, to $57.5 million as of February. The strategy line in the handbook is stated flat out: "Providing all the tools in one is a core part of our strategy."

For most of SaaS history, that sentence was a red flag. Focus was the religion, the point solution was the credential, and the second product was something you earned after a decade of dominating the first. My claim this week: AI collapsed the cost of the second product, and cheap second products change what a software business even is. The durable unit is shifting from the product to the persona. Pick the person you serve, then own every workflow they touch, because if the workflows are adjacent to your data and you leave them unbuilt, someone else's roadmap now reaches them in a quarter.

Parker Conrad spent years defending this idea back when it was heresy. He coined the term compound startup for Rippling's bet: many deeply integrated products built in parallel on one shared platform, sold to one buyer, while the rest of the industry preached focus. The bet paid. Rippling crossed $1 billion in ARR with a $16.8 billion valuation, runs more than ten product lines that each clear $1 million in ARR, and reportedly gets new lines to that mark within five or six months of launch.

What changed for everyone else is the build math. In June I wrote that the build-versus-buy line moved because agents collapsed the cost of internal tools. The same collapse applies to product lines. The second product used to cost two years of a founding team's attention; a credible v1 is now a quarter's work for a team that has encoded its patterns, which means Conrad's contrarian architecture is becoming the default playbook. The scarce input moved from engineering capacity to clarity about the person you serve.

PostHog has that clarity in writing. Their handbook picks products by one filter, the product engineer, and notes something worth stealing: "The persona doesn't change as the company gets bigger." A persona is durable in a way a product category never was. Analytics can be disrupted; the engineer who needs to understand users will exist in every company, at every stage, indefinitely.

The demand side moved just as fast, because your customers know what building costs now. A roadmap that ships one meaningful thing a year read as discipline in 2021. In 2026 it reads as a vendor falling behind, and the buyer doing that math has new alternatives on both sides: a compound platform that covers the adjacent workflow already, and an internal agent that can build the missing piece.

The consolidation pull is real on the buyer's side of the table too. One system means one login, one bill, one place the data lives, and, increasingly, one surface an agent can operate end to end. An agent runs a whole workflow much more easily when the whole workflow lives in one platform. In July I wrote about Omnisend absorbing migration costs as a customer-acquisition expense; the same company shows what full persona coverage looks like at the small end: for the ecommerce marketer, it covers email, SMS, web push, forms, and the automation connecting them, one retention surface for 150,000+ brands. Ramp ran the same play for the finance team: corporate cards, then expense, bill pay, procurement, travel, and treasury, and now agents that run those workflows themselves. The pattern in every case is identical: same buyer, adjacent workflow, shared data.

Here is the honest caveat, because the 2010s already ran the bad version of this movie. The old suites grew by acquisition into bundles of C-minus products that shared a logo and an invoice and nothing else, and best-of-breed point solutions beat them for a decade. Product count was never the moat.

The compound companies that work share a spine. Rippling's products all read and write one employee graph. PostHog's all sit on one event stream. Ramp's sit on one transaction ledger, Omnisend's on one customer purchase history. A second product on the same spine makes the first product better, because every workflow it touches enriches the data all the others run on. A second product off the spine is a distraction wearing your logo. PostHog's handbook is disciplined about this in a way worth copying: "Shipping them in the right order is key to a fast return on investment from every new product."

So the strategic question for a software business comes down to two clauses: what is our spine, and which of our persona's workflows does it already touch.

Sell side first. Write down your persona as a person, not a market ("the retention marketer at a Shopify brand doing $2-20M," not "ecommerce"). List the ten tools that person touches in a normal week around yours, then mark which of those workflows read or write the data your product already holds. That marked list is an expansion map, and at 2026 build costs, each line on it is roughly a quarter of work. For each one, decide on purpose: ship it, or be able to say why it belongs to someone else's spine. The lines you leave undecided are the ones that show up in a competitor's launch email mid-renewal.

Buy side, the mirror image. Count the logins one team persona juggles today. That number is heading down whether you plan it or not, so negotiate like consolidation is coming: shorter terms on point solutions, and a direct question to your platform vendors about what's next on their map for your persona.

The question used to be what to build. The question now is who you build for. Pick your person, then build them everything.

Till next time,
Chris