Working with Vercel?

Vercel bill alarming, deploys red or Edge Functions timing out? We sort it out.

Most Vercel accounts that reach us were set up by somebody who moved on, and the last person to look at the usage page shut the tab quickly. The build worked once, a few of the deploys lately have not, and the invoice keeps growing on its own. We trace where the function invocations and the bandwidth are actually going, cut the parts that do not earn their keep, and leave the account working the way the dashboard says it should.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

deploy.log

still failing

What it is

Vercel at a glance

Vercel, formerly known as ZEIT, was founded in 2015 by Guillermo Rauch and is headquartered in San Francisco. The company is the commercial home of Next.js, Turborepo and Turbopack, and runs the hosting platform most teams reach for when they ship a modern React application.

The platform deploys static sites and framework applications such as Next.js, Nuxt, SvelteKit and Astro to a global edge network, with a fresh preview URL on every git push, built-in image optimisation, serverless functions, edge functions, and managed primitives for KV, Blob storage and Postgres (Neon-backed) alongside an Observability dashboard.

Type
Front-end cloud and edge platform
First released
2015 as ZEIT Now
License
Proprietary SaaS; the CLI and Next.js are open source
Maintained by
Vercel, Inc.
Typically used for
Next.js apps, marketing sites, SaaS front ends, AI-builder output, headless-CMS frontends

The honest version

Vercel pros and cons

No technology is all upside. This is where Vercel earns its place and where it tends to cause trouble.

Where Vercel is strong

  • Zero-config deploys for most frameworksA git connection is enough for Next.js, Nuxt, SvelteKit, Astro and a dozen others to build and ship without a hand-written pipeline behind them.
  • A global edge network is includedStatic assets and cached responses are served from locations near the visitor, so a small team gets CDN behaviour without negotiating a separate contract.
  • Preview deployments on every pull requestReviewers click a link that shows the branch running on real infrastructure, which catches far more problems than screenshots pasted into a ticket.
  • Image and script optimisation built inThe platform resizes images, converts formats and hoists third-party scripts without a loader to configure, handling two of the usual performance wins for free.
  • Developer experience most teams actually likeBuild logs read well, the CLI stays out of the way, and the gap between pushing a commit and seeing it live is short enough to encourage small frequent changes.
  • A generous Hobby tier for small projectsPersonal sites and side projects run under the free plan comfortably, which is why so many prototypes ever leave somebody's laptop in the first place.
  • Serverless and edge functions in the same repoAPI endpoints and middleware live next to the pages that call them, so the backend for a small product does not need a second service running alongside the frontend.
  • KV, Blob and Postgres primitives are there when you need themA key-value store, object storage and a managed Neon-backed Postgres are available from inside the dashboard, which is enough to stand up a product without a second provider.
  • Dashboards that are worth looking atThe Observability view ties deployments, logs, function invocations and web vitals together, which means slow routes and expensive functions get noticed before the invoice arrives.

Where Vercel bites

  • Pricing climbs sharply beyond the Pro tierThe jump from Hobby to Pro is easy, the jump from Pro to Enterprise is a conversation with sales, and bandwidth overage at forty dollars per hundred gigabytes punishes a viral post.
  • Function size limits bite on heavy dependenciesThe fifty-megabyte serverless bundle sounds roomy until a headless browser, a PDF toolkit or a machine-learning helper is pulled in, at which point the build refuses to ship.
  • ISR caching has shifted between Next.js versionsBehaviour that was implicit in Next.js 13 became explicit in 14 and changed again in 15, so an app that was quietly cached last year is quietly uncached this year and nobody notices until the bill arrives.
  • Cold starts on edge and serverlessA function that has not been hit in a while takes longer on its first invocation, which is fine for a form submission and painful for a user-facing API on a low-traffic route.
  • AI-builder output ships patterns the platform bills forCode that Lovable, Bolt or v0 generates often calls an uncached fetch on every request, runs image optimisation against remote URLs it shouldn't, and turns a cheap page into a line item.
  • Vendor lock-in through specific primitivesKV, Blob, the image loader and the cache tag API are proprietary enough that moving to another host later means replacing each one, not just redeploying.
  • Not suited for long-running background workFunction timeouts measured in seconds rule out anything resembling a worker queue, a long export or a slow third-party call, which has to live on a Node host or a queue provider instead.

Sound familiar?

Common Vercel problems we fix

The Vercel problems that bring people to us, and what you have once each one is fixed.

Deployment overage: additional bandwidth charged at $40/100GB

The monthly bill is suddenly thousands of dollars

Uncached fetches on every request, image optimisation pointed at an uncontrolled origin, and function invocations where a static page would do, usually all three at once.

Spend back to where it should be

Error: Serverless Function size limit exceeded (50MB)

The function bundle is too big to deploy

A headless browser, a PDF toolkit or a machine-learning package pulled into a route that only needs a slice of it, with no tree shaking because the import is dynamic.

Bundle trimmed, deploy green

Build failed: Command npm run build exited with 1

The build goes red on Vercel but green locally

A dependency that is only installed on a developer's laptop, an environment variable missing from the project settings, or a case-sensitive import that macOS was forgiving about.

The same build in every environment

Edge Function exceeded CPU time limit (50ms)

An Edge Function times out in production

Heavy work hidden inside middleware or an edge route, or a dependency that calls synchronous crypto on every request, both of which blow past the edge CPU budget.

Edge code stays inside its budget

ISR revalidation failed: fetch returned 500

Content updates but the live site keeps serving stale pages

A revalidate call that reaches a broken upstream, a cache tag that was never attached, or a webhook from the CMS that never got wired to the revalidation endpoint.

Content changes show on the next request

Error: Image Optimization failed with 403 from upstream

next/image is broken in production

The optimiser cannot fetch the source because the remote host is not in images.remotePatterns, the bucket blocks the optimiser's user agent, or the CDN strips the signed query string.

Images served through the optimiser

Preview deployment 404: branch not synced

Previews stopped appearing on pull requests

A disconnected git integration, a monorepo project filter that no longer matches, or a protected branch rule that keeps Vercel from checking out the ref.

Every pull request gets a working preview

admin seat: unknown

The person who set it up has left

Billing on an unknown card, domains pointing at a project in somebody else's team, and nobody with the permissions to change either.

The account in your team, in your name

What we do

Vercel help, from rescue to upkeep

Call us for one failing deploy, or hand us the whole project.

Deploy rescue
A red build on Friday afternoon turned green, with the cause written down so the next engineer does not meet the same wall.
Cost review
Reading the Observability dashboard honestly, finding the invocations and the bandwidth you are actually paying for, and cutting the ones that earn nothing.
Caching and ISR
Tags, revalidation and the right rendering mode per route, so a change in the CMS shows up on the live site without doubling the invocation count.
Edge and serverless
Keeping edge functions inside their CPU budget, trimming serverless bundles under the size limit, and moving long-running work off Vercel to somewhere it can actually finish.
AI-builder cleanup
Taking a Lovable, Bolt or v0 project that works for ten users and getting it ready for ten thousand without rewriting the parts that are already fine.
Handover and takeover
Moving an account out of a former developer's workspace, straightening out domains, env vars and billing, and leaving you with the keys.
Managed Tech
Monitoring, framework upgrades and a team to call, on an ongoing basis.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running Vercel alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

Vercel questions

Something else? Call+1 737 727 7327

Why is our Vercel bill suddenly thousands of dollars?

Something in the application started doing work on every request that used to be cached, or a page that used to be static is now rendering dynamically, or image optimisation is pointing at a remote host that returns huge originals. We read the Observability dashboard with you, find which routes and functions are responsible for the invocations and the bandwidth, and put caching, static rendering and sensible image sources back in place. In most cases the bill drops back to where it was within a billing cycle.

Can you take over a Vercel project from a previous developer?

Yes, and we do this regularly. We move the project into a team you control, straighten out the git integration, the domains, the environment variables and the billing, and document how the deploys actually work so you are not reliant on the next person remembering. If the previous developer is uncooperative we work around it using the domain registrar, the git host and your payment records.

Why does our Next.js build keep failing on Vercel?

Usually a difference between the build on a developer's laptop and the one running on Vercel's clean checkout: a missing environment variable, a case-sensitive import that macOS let slide, a dependency installed globally but not listed in package.json, or a Node version mismatch. We read the build log properly, reproduce the failure in a clean container, and fix the cause rather than pinning around it.

Can we deploy our Next.js app somewhere other than Vercel?

Yes, and for some workloads it costs noticeably less. Next.js runs anywhere that can run Node, and the standalone output target produces a slim container suitable for AWS, Cloud Run, Fly or a plain Node host. Image optimisation, ISR and the Edge runtime need a little more thought off the platform, so we swap the image loader, put a CDN in front for caching, and keep middleware simple. For most apps the extra configuration is worth the saving and the removed lock-in.

How do we fix ISR revalidation failures on Vercel?

First we work out whether the fetch inside the revalidation path is actually succeeding; a 500 from the upstream blocks the whole thing. Then we check that the fetch is tagged and that revalidateTag is called from the webhook endpoint the CMS actually hits. Finally we confirm the cache is behaving in production by watching the Observability log while a known change goes through. Most ISR failures come from one of those three links being quietly broken.

How much does Vercel help cost?

Estimates are free. Describe the project and the invoice and you get a price before any work begins.

Free estimate

Vercel project giving you trouble?

Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.