Upsell

Growing revenue was treated as a matter of landing new customers for a long time. Then the cost of landing one kept climbing year over year and the arithmetic changed. Getting noticed, becoming trusted, and clearing someone's approval process from scratch turned out to cost several times more than raising the amount at an account already inside the door.

Asking an account to pay more, though, accomplishes nothing on its own. That became visible at Pagedeck when Grace asked why revenue growth was slow while new contracts were up. Every deal had been signed at five seats, and those accounts sat on the same tier a year later. What the higher tier contained was extra storage and a higher seat cap, so customers who ran out of room deleted old documents instead of upgrading.

This is the point where upselling stopped being a sales technique and became a product design question. There has to be a reason inside the product to move up, and creating that reason means deciding where the line between tiers falls rather than coaching anyone on how to ask.

Build the reason to move up into the product first, then pick the customers for whom that reason has actually appeared. Five steps.

1. Decide where the tier line falls. Split on numbers alone and customers reduce their usage instead of upgrading. Put above the line the things that become a problem as an account grows: splitting permissions by department, a record of who opened what, an approval step before something goes out.
2. Define the signal that says it's time. Write it concretely — more than 20 seats, more than 30 external shares a week, and one inbound question about dividing permissions. Without a written signal, two people looking at the same account reach different conclusions, and the reason a deal closed stops being repeatable.
3. Choose where the paywall sits. It's the point that blocks someone reaching for a higher-tier feature. Placed before the need exists it reads as an ad; placed at the exact moment things got awkward it reads as a reason to change plans. Blocking permission splitting for a team of three is the early version. Blocking it for twenty people editing the same document is the late one.
4. Split the approach by amount. A customer paying $25 a month belongs to an in-product prompt; a customer paying $2,300 a month belongs to Sofia calling them. Put a person on the first and the hours cost more than the revenue. Leave the second to a prompt and nobody who has to clear an approval process can act on it.
5. Count the increase separately. Measure it as expansion MRR — the monthly recurring revenue added by customers who were already paying — by freezing the customer list as of the first day of the period and counting only within that list. Accounts that arrived mid-period don't go in the number no matter how large the deal was.

It fits when the tiers genuinely differ and when customer accounts grow over time. Permissions and audit records that nobody asks about at twenty people become contract conditions at 300, so the reason to move up appears from the customer growing rather than from anything you do. Products whose customers stay the same size have to manufacture that reason instead.

The two neighboring levers differ in what gets sold. Seat expansion is seat count rising on the same tier, and cross-sell adds a different product on top. In a per-seat pricing model seat expansion usually happens first, and the friction it creates becomes the upsell signal.

Four limits.

1. It doesn't work when the tier difference is only numbers. Facing a plan that differs only in storage and seat caps, customers delete things or share logins.
2. Aimed at a customer with no need yet, it backfires. Someone who hits a blocked screen every week concludes the product only wants money rather than deciding to upgrade.
3. Moving an existing feature up to create the line drives current customers away. Adding something new above the line and taking something away from below it are completely different events from where the customer sits.
4. Counting only the increase produces an illusion. Downgrades and departures get counted in other columns, and the place both sides get combined is net revenue retention (NRR). Strong upsell numbers alongside NRR at 90 percent means more was lost than gained.

"This account is at twenty-three seats now, and they've asked about permission management. It's time to bring up the business plan."

Every pre-agreed signal has fired, so the conversation can start.

"The paywall is sitting too early. Blocking permission splitting for a team of three is just an ad."

The block lands before the need appears rather than at it.

"Of this month's $13,500 in expansion MRR, $3,000 is upsell and the rest is seat growth."

Separating one column by lever to see which one actually worked.

When Pagedeck built its business plan, the first decision was what to put above the line. Mia dropped the proposal to split on storage and chose permissions by department plus a record of who opened what, both of them newly built. Sofia took the weekly list of accounts past twenty seats and contacted only the ones that had asked about permissions. Three months later expansion MRR was up, but most of it was seat growth and upsell accounted for eight deals. Mia put moving the paywall one step later on the next quarter's list.

The most common reading treats this as persuasion performed by sales. The picture that comes to mind is a conversation that raises an amount. But a customer who upgrades on a call is a customer for whom the reason already existed. Leave it as a sales problem and the target goes up while the tier line stays where it was, and six months later what remains is a worn-out account manager and a flat number.

The other misreading is that pricing the higher tier aggressively makes upselling work better. Each upgrade moves more money, so the math looks favorable. But once the gap runs past about three times, customers choose to endure instead of moving. They patch the friction by hand, share logins, and spend that time pricing other products. Charging more and making it possible to move up pull against each other, so adjusting one without the other jams the second.