The economics: why caught expansion compounds hardest

Expansion ARR is the cheapest revenue you will ever book, and that is the case for building this loop before you build another acquisition channel. There is no ad spend, no SDR cold-call, no month of nurture; the customer is already inside the product, already paying, already feeling the pain you are about to relieve. The only cost is the latency between the signal and the offer, and the agent's entire job is to drive that cost to near zero.

Run the second-order maths and it gets starker. Expansion does not just add revenue, it lengthens customer lifetime and lifts LTV, because an account that has upgraded into a higher tier and spread across three teams is structurally harder to rip out than the single-team account it used to be. Every caught expansion compounds in two directions at once: more revenue per account now, and a longer, stickier relationship that keeps paying. That is why a point of NRR is worth more to enterprise value than a point of new-logo growth, and why software at 120%-plus NRR commands the multiples it does.

The acquisition comparison is the part founders underrate. A team obsessing over LTV:CAC on the acquisition side while leaking expansion is optimising the expensive half of the ledger and ignoring the free half. The expansion offer the agent catches has an effective acquisition cost near zero, which means even a modest conversion rate on it returns more per hour of effort than almost anything your top-of-funnel can do. For a lean team, that asymmetry is decisive: you cannot out-hire a sales floor, but you can out-sense one, and sensing is where the agent has the structural advantage.

The uncaught version is the silent tax. Every 80%-seats account that times out into a discounted renewal is expansion you funded the product to earn and then handed back. It never appears as a loss line, which is exactly why it persists for years in companies that would never tolerate the same leak on the acquisition side. Wiring the agent is how you make the invisible leak visible, then close it.