Vercel · unsolicited concept · 2026

What will it cost? Or save?

Showing a Hobby developer their own last thirty days, priced as Pro — before they decide.

Company: Vercel
Role: Senior Product Designer, Growth
Surface: The Hobby → Pro upgrade modal
Problem: $20.20 is quoted to the cent. The metered tail behind it is one clause of prose.
Move: Price their own last 30 days as Pro, and set the ceiling inside the checkout
Borrowed from the posting: Success and guardrail measures — applied as formulas and a decision rule, not described

Sven Manzanarez — an unsolicited concept, made for my application to the Senior Product Designer, Growth role. Not affiliated with Vercel.

Every usage figure and dollar amount in the prototype is invented.¹ The screenshots are from my own account.²

The decision nobody can price

Hobby is a bounded allotment. You cannot be billed. Pro is a meter — around forty billable resources, charged in arrears. The upgrade modal is the moment a developer swaps one for the other, and it quotes a single number.

$20.20

Quoted to the cent, in a table, with estimated tax broken out.

— — —

Every other meter, in one subordinate clause at the bottom of the legal paragraph.

“Other metered features bill in arrears per pricing.”⁴

Vercel's upgrade modal, verbatim. It is accurate, it is disclosed, and it carries the entire unpriced risk of the decision.

The developer's real question is not what does Pro cost. It is what will Pro cost me — and Vercel is the only party in the transaction who can already answer it. The usage data is theirs. It is sitting in the account being upgraded.

They have already solved this once

One line in that same breakdown does exactly what the rest of the modal does not:

Vercel's upgrade breakdown: Pro $20, Flat Rate CDN $0 with the line '1M CDN requests included · Tiers from 10M (+$20/mo.) up to 150M (+$300/mo)', 1 member $0, Estimated tax $0.20, Total $20.20 per month. — Vercel's own upgrade modal, 30 Sep 2026.²

Flat Rate CDN is a meter that was turned into a visible ladder — what's included, where the next step is, what the top costs. It is in the checkout, in plain type, before the card is charged. Vercel built that pattern and shipped it.

Everything below extends that line to the other thirty-nine meters. It is not a new idea in their product. It is one they have only applied once.

Principles

Price the past, not the future.A forecast is a promise you will be held to. Last month is a fact you already own.
For most accounts the answer is reassuring.Silence reads as a warning. Staying quiet costs the upgrade that would have been easy.
One number is a guess. Three are a decision.Charged today, a typical month, and a hard ceiling.
The ceiling is set before the card is charged.Not as step three of a welcome tour the user is already clicking past.
A cap must say what it costs you.If pausing takes the site offline, that sentence goes on the control, not in the docs.

The flow

Two accounts, the same four steps, opposite conclusions. Switch accounts or jump between steps at any point.

Steps 1 and 2 are new surfaces. Step 0 is a redraw of Vercel's live modal, kept to the same copy and layout so the additions can be judged against it.³

Why the cap is the second move and not the first

Spend Management already exists. It is listed as a Pro feature on the upgrade card, it renders on Hobby as an empty state you can look at but not use, and after you pay it appears as Next step 3 of 6, between claiming a domain and enforcing MFA.

Vercel's Welcome to Pro panel listing Spend management — caps, alerts, and full visibility into costs — followed by a Next steps list in which Configure Spend Management is the third item. — *Caps, alerts, and full visibility into costs* — offered after the commitment it would have informed.²

Vercel's downgrade modal, showing a billing period progress bar reading 20 of 30 days left and the line: You'll receive a refund of $13.46 for unused time on your plan. — The *downgrade* modal already does arithmetic in front of the user — progress bar, days left, refund to the cent. Leaving is better instrumented than arriving.²

So the second half of this concept is not a feature. It is a sequencing change: the same control, two screens earlier, with a number already suggested for it. That is the cheapest thing on this page to build and probably the first thing I would ship.

How I would know it worked

The posting asks for someone who can form useful hypotheses and choose success and guardrail measures. That is the practice I took from it, so here it is applied rather than described.⁶

Hypothesis. What stalls the Hobby → Pro decision is not the $20. It is the unpriced tail behind “other metered features bill in arrears.”

Each measure is attached to the specific thing a change produces, not to the change itself. Removing ① would remove S and A; removing ② would remove B.

The measures, written out

Primary

Day-60 Pro retention

R₆₀= teams still on Pro at t + 60dteams that upgraded Hobby → Pro at t

Monthly cohort, against a holdout. Ship gate: ΔR₆₀ ≥ +3 pp

Conversion on its own can be bought by saying less, so it is the wrong thing to optimise against. Retention cannot be bought that way: if the retrospective is honest, the people it converts should still be there two months later.

Guardrail

Upgrade start rate

S= upgrades confirmedupgrade modals opened

Hold: S ≥ 0.93 · S₀ (no worse than a 7% relative drop)

A real number can frighten someone off, and for the spiky account it probably should. But a large drop means the number is being presented badly, not that it should be hidden — so breaching this triggers a re-frame, not a rollback.

Guardrail

Billing-surprise rate

B= contacts tagged unexpected-charge within 45dnew Pro teams in cohort × 100

Hold: B ≤ B₀

If this rises, the estimate is being read as a promise. That is a copy failure with a real cost attached, and it is the failure mode I would expect this design to produce.

Falsifier

Band accuracy

A= teams where │invoice₁ − estimate│ ≤ max($5, 0.25 · estimate)new Pro teams in cohort

Hold: A ≥ 0.80 · Kill: A < 0.60

The number that can end the feature. If last month does not predict next month for most accounts, the retrospective is theatre and it should come out rather than be tuned until it looks better.

SHIP ΔR₆₀ ≥ +3pp ∧ S ≥ 0.93·S₀ ∧ B ≤ B₀ ∧ A ≥ 0.80 RE-FRAME ΔR₆₀ ≥ +3pp ∧ 0.85·S₀ ≤ S < 0.93·S₀ → the number is right, the presentation is not. Rewrite and re-run. KILL A < 0.60 ∨ B > 1.25·B₀

The trade I would accept, stated in advance so it cannot be rationalised afterwards: fewer upgrade starts in exchange for higher day-60 retention is a win. Fewer starts and flat retention is a loss, and I would roll it back.

What is theirs and what is mine

Separated so nothing on this page can be mistaken for shipped work.

Vercel's productCaptured from my own team account, 30 Sep 2026.²

Upgrade modal

The surface this concept changes. Source for every layout and copy decision in the prototype.

Vercel's Hobby plan billing page, showing the Upgrade to Pro card and a Spend Management section with a No team budget empty state.

Hobby billing

Spend Management is visible here with a “No team budget” row and a greyed Add Budget. You can see what you don't have.

Welcome to Pro

Where Spend Management is offered today — third in a six-item checklist, after the upgrade.

Cost breakdown

Contains the Flat Rate CDN tier line — the precedent the whole concept extends.

The clause

The sentence the concept exists to answer.

Vercel's Downgrade from Pro modal with a billing period progress bar and a stated refund of $13.46.

Downgrade modal

Proof the pattern is already in the design language. It was simply applied to the exit.

InventedDrawn by me. Nothing here exists in Vercel's product.

The retrospective blockYour last 30 days of Hobby usage, re-priced on Pro's rate card, itemised by resource, with the included credit applied. New surface, new copy.
The ceiling row inside checkoutSpend Management's existing control, moved into the upgrade modal with a suggested figure and the pause consequence written on the toggle.
The three-number totalReplaces the single Total line with charged today, a typical month, and a ceiling.
Both accounts, every figure, and all four measuresInvented to illustrate the two cases and to show the practice applied.¹ Real per-unit prices are not quoted.³

What I could not work out from outside

Does the Flat Rate CDN tier come back down?The modal says one spike moves you to the matching tier at the next billing cycle. It does not say whether a quiet month moves you back. If it does not, a single bad day changes the floor permanently — and that is the most important thing the upgrade modal is not saying.⁵
How many Hobby accounts have enough history to price?A brand-new account has nothing to re-price, and that is probably a large share of the upgrade population. The empty state may matter more than the populated one. I would want the distribution before building either.
Is the pause consequence survivable?“Your production deployments go offline” is the honest sentence. It may also be the sentence that makes nobody turn it on — in which case the cap is alerts only, and this design is half of what it claims.
Where does the agent usage land?Vercel Agent is on the Pro feature list. Agent-driven consumption is the least predictable kind there is, which makes a ceiling more valuable and a retrospective less so. I do not know enough about how it meters to design for it.

Notes

Every usage quantity, dollar amount, account and persona in the prototype is invented. Neither account is a real Vercel customer, and the four measures are my own proposal, not Vercel's. ↩
Screenshots were captured from my own Vercel team account on 30 September 2026. The plan structure, pricing copy and interface are Vercel's. ↩
Per-unit prices are not quoted. The prototype uses plausible round figures so that it does not misrepresent a rate card that changes. The arithmetic is internally consistent; the rates are not Vercel's. ↩
“Other metered features bill in arrears per pricing” is quoted verbatim from the upgrade modal's legal paragraph. ↩
Stated as an open question rather than a finding. I could not determine the behaviour from the product surfaces available to a Hobby or Pro account. ↩
The posting asks for a designer who can “form useful hypotheses, choose success and guardrail measures.” I took that as the practice to borrow and applied it to this concept rather than writing about it. ↩

Sven Manzanarez · productdesign.digital · Concept work, not affiliated with Vercel. All trademarks belong to their owners.