Pick the tool for whoever changes the site every week.
We run two kinds of marketing site. Our own, kraftbase.com, is built in Framer: the pages, the blog you're reading and the hosting. For a client, a B2B company whose product is a mobile app, we built the marketing site in Next.js: long articles, a large FAQ, landing pages for each audience and one page that talks to the product.
Both work. They suit different teams, and the difference only showed once each site had been running for a while. This post covers what each site had to do, the decisions behind each one, how they're set up, what running them is like, and the signs that a site should move from one to the other. Where we measured something, we say how.

What each site needed
Our site, on Framer:
Designers change it every week, without waiting on a developer.
A blog that grows fast. 47 posts are live, and each one’s FAQ, call to action and structured data for search have to stay right as we edit.
A review step before anything goes live, because several people and scripts touch the content.
No hosting to run. We’re a design studio; our developers’ time goes to client work.
We’d built sites in Next.js before, and for our own we wanted the CMS built into the tool rather than bolted on. Framer also takes most of the difficulty out of design: our designers design straight into Framer, so there’s no handoff from a design file to code. Day to day, a designer makes most of the changes. When a change touches the CMS itself, our Framer developer does it.
The client’s site, on Next.js:
Mostly content. 24 long articles, a big FAQ and audience landing pages.
Product features, not just pages. Users sign in, sign out and delete their account, which means real login and calls to the product’s API.
A CMS to switch content on and off without a developer.
Analytics from three tools on every page.
Why code for this one? It isn’t only the amount of content. The site behaves like part of the product: visitors sign in, sign out and delete their account, and each of those calls the product’s API. Features like these are very hard to build in Framer, and some aren’t possible at all. On top of that, it had the reasons that have moved other clients of ours from Framer to code: content too nested and complex for Framer’s CMS, a Framer plan for that much CMS content that cost more than the site justified, and interactions too complex to manage in a Framer site.
Framer vs Next.js: the difference that matters
Framer is a visual builder that also hosts the site. You design pages on a canvas, content lives in Framer’s CMS, and publishing is one button. Next.js is a React framework. The site lives in a code repository, a developer changes it, and a host like Vercel deploys each change.
So the real question isn’t which one is faster or cheaper. It’s who changes the site every week, and how much the site has to behave like the product. Everything else follows from that.
| Framer (our site) | Next.js (client site) |
|---|---|---|
Who changes pages | Designers, on the canvas; a Framer developer for CMS changes | Developers, in code |
Review before live | A branch, merged by a person | A commit, with lint and a build check |
Where content lives | Framer’s CMS (we write it in a repo and push it in) | Mostly in code; the CMS holds on/off flags and legal pages |
Motion | Built into the canvas | Code (Motion and GSAP) |
Product logic | None needed | Sign-in, sign-out and account deletion, calling the product’s API |
Hosting | Framer | Vercel |
The decisions
Who edits, and how changes get reviewed
On Framer, our designers change layouts and templates directly. When we added a lead heading to every blog post, it was one new CMS field and one text layer on the post template, and all posts had it at once. For review, Framer gives us branches: every change we push from our content system lands on its own branch, someone checks it, and only then is it merged and published. Between 5 and 18 November 2026 we published the site 18 times.
On the client’s Next.js site, a change is a commit. Every push deploys through Vercel, and a check runs lint and a full build before each commit. That’s a solid review step for developers. It also means a marketer can’t move a section without one.
Where the content lives
Our posts live in Framer’s CMS, but we write them as Markdown in a repository and push them in with scripts. That gives us version history, checks and one place to review everything. It also showed us the edges of Framer’s rich text: our scripts can add and change parts of a post but can’t remove them, so a big rewrite means rebuilding the post on a branch. Diagrams, like the ones in this post, have to be images.
On the client’s site, most long content lives in the code itself, as data files. The CMS holds only an on/off switch per article and some legal pages. That suits developers, since everything is typed and reviewed. Editors can’t change an article’s text without a developer, and there’s no draft preview: depending on the page, a CMS change either goes live straight away or waits for the next deploy.
Design and motion
Framer’s interactions and animations are part of the canvas, so our designers ship motion without code. On the Next.js site, motion is code: a shared scroll-reveal component built with Motion, and GSAP for six heavier sections. That gives full control, and every animation is a developer’s job.
Connecting to the product
This is where Next.js earns its place. The client’s account-deletion page loads a real login and calls the product’s API, and that login code is scoped to that one page, so the rest of the site never loads it. Our Framer site has nothing like that and doesn’t need it: its job ends at “talk to us”.
Performance
We measured both with Lighthouse on mobile settings, three runs per page, and took the median.
Our Framer site (18 November 2026): the home page scored 74 and transferred 6.9 MB, more than half of it a single hero video. A blog post scored 90 and transferred under 1 MB.
The client’s Next.js site (5 November 2026): the live home page scored 58, an article page 65 and the FAQ 41. Every page loaded 3.2 MB of font files, and the home page shipped 557 KB of JavaScript (gzipped).
These are different sites, with different content, measured on different days, so this isn’t a benchmark. What both runs show is that the weight comes from assets: a video on one, fonts and images on the other. Neither tool is the reason a page is slow.

Where the home page weight came from, from one Lighthouse run per site. The Next.js breakdown is from the same build served locally, which transferred about the same as the live site.
How each is set up

Same job, two paths to production: a branch and a publish button, or a commit and a deploy.
Write. On Framer, designers edit the canvas and our scripts push post content. On Next.js, developers change code and content files.
Review. On Framer, each change sits on its own branch. On Next.js, a pre-commit check runs lint and a full build.
Merge. On Framer, a person merges the branch. On Next.js, the commit lands on the main branch.
Go live. On Framer, one publish updates the whole site. On Next.js, Vercel deploys every push.
Redirects and hosting. Framer keeps redirects in its settings and hosts everything. On Next.js, the host handles the domain and the code would hold any redirects.
What it’s like to run
Framer means almost no operations: no servers, no builds, no deploy pipeline. The surprises came from the editor, not from hosting:
Preview only picked up our scripted edits after a reload.
A form on our Guides page never submitted, so we replaced it with a direct download.
Redirect rules can’t contain parentheses.
Once, a publish was blocked by an error in an unrelated component until it cleared the next day.
The Next.js site has the usual developer chores: dependency upgrades, builds and deploys. It also has more ways to go quietly wrong. Its home page is built once at deploy time, so new events from the CMS only appear after a redeploy.
What we learned
Match the tool to the person who changes the site every week. That decides more than any feature list.
A visual builder still needs a review step. Branches gave us one on Framer; without them, several people and scripts writing to one live site would have been risky.
Code gives control and moves the work to developers. Content in code is clean, and every copy change becomes a ticket.
Fix assets before blaming the tool. Video, fonts and images were the heaviest part of both sites.
When to move (either way)
Signs a Framer site should become code:
It needs real login, payments or data from your product on more than a page or two.
You want it to share components and design tokens with your app.
Your team writes content in a repository and the CMS has become the bottleneck.
Your content is deeply nested (items inside items) and keeps fighting the CMS structure.
The plan your CMS content needs costs more than the site is worth to you.
Signs a coded site should move to Framer:
Marketers wait on developers for page and copy changes.
The site is pages and a blog, with no product logic.
Nobody wants to own builds and upgrades.
We’ve made the first move for clients, from Framer to code. Each time, the content had outgrown Framer’s CMS, the plan for that content no longer paid for itself, and the interactions were more than the canvas could manage. By then the clients were finding Framer hard to work with, not easy.
A move in either direction means rebuilding pages, exporting the CMS content, and keeping every old URL alive with redirects so you don’t lose search traffic. We’ve written up what that looks like in a Webflow to Framer migration for SaaS. If you’ve decided on code and are choosing a framework, read Astro vs Next.js for marketing sites. If you’re choosing between visual builders, see Framer vs Webflow for B2B SaaS.
We’ve left plan and seat costs out on purpose: they change often, so check Framer’s pricing page for current figures.
Related reading
Whichever tool you pick, the pages matter more. Start with our SaaS website structure and what drives SaaS website redesign cost.
FAQ
For most, yes. Our own site and its 47-post blog run on Framer, with structured data, FAQs and redirects. It fits when designers and marketers change the site and it doesn't need product logic. Plan for its rich-text limits if you edit content with scripts.
Only the parts that live in a CMS. On the site we built, most article text lives in code, so editors can switch articles on and off but not rewrite them. A headless CMS with preview helps, and someone still has to build and maintain that setup.
Not by itself. In our Lighthouse runs, the heaviest pages on both sites were slow because of assets: a hero video on our Framer home page, and fonts and images on the Next.js site. A Framer blog post of ours scored 90 on mobile. Fix assets first.
Yes. Rebuild the pages in code, export the CMS content, and redirect every old URL to its new address so search engines carry rankings over. Move when the site needs product logic or shared components, not just because the team grew.
It doesn't have to. Keep the same URLs where you can, add a redirect for every one that changes, and carry over titles, descriptions and structured data. Then check indexing in Search Console for a few weeks after launch.
Yes. Framer has branches: changes sit on a branch until someone merges them, and publishing is a separate step. Every change we push to our own site with scripts lands on its own branch first, so nothing reaches the live site without a person checking it.
Engineering
Framer or code for your marketing site?
Tell us who edits the site and what it has to do. We'll tell you how we'd build it and why.
Pick the tool for whoever changes the site every week.
We run two kinds of marketing site. Our own, kraftbase.com, is built in Framer: the pages, the blog you're reading and the hosting. For a client, a B2B company whose product is a mobile app, we built the marketing site in Next.js: long articles, a large FAQ, landing pages for each audience and one page that talks to the product.
Both work. They suit different teams, and the difference only showed once each site had been running for a while. This post covers what each site had to do, the decisions behind each one, how they're set up, what running them is like, and the signs that a site should move from one to the other. Where we measured something, we say how.

What each site needed
Our site, on Framer:
Designers change it every week, without waiting on a developer.
A blog that grows fast. 47 posts are live, and each one’s FAQ, call to action and structured data for search have to stay right as we edit.
A review step before anything goes live, because several people and scripts touch the content.
No hosting to run. We’re a design studio; our developers’ time goes to client work.
We’d built sites in Next.js before, and for our own we wanted the CMS built into the tool rather than bolted on. Framer also takes most of the difficulty out of design: our designers design straight into Framer, so there’s no handoff from a design file to code. Day to day, a designer makes most of the changes. When a change touches the CMS itself, our Framer developer does it.
The client’s site, on Next.js:
Mostly content. 24 long articles, a big FAQ and audience landing pages.
Product features, not just pages. Users sign in, sign out and delete their account, which means real login and calls to the product’s API.
A CMS to switch content on and off without a developer.
Analytics from three tools on every page.
Why code for this one? It isn’t only the amount of content. The site behaves like part of the product: visitors sign in, sign out and delete their account, and each of those calls the product’s API. Features like these are very hard to build in Framer, and some aren’t possible at all. On top of that, it had the reasons that have moved other clients of ours from Framer to code: content too nested and complex for Framer’s CMS, a Framer plan for that much CMS content that cost more than the site justified, and interactions too complex to manage in a Framer site.
Framer vs Next.js: the difference that matters
Framer is a visual builder that also hosts the site. You design pages on a canvas, content lives in Framer’s CMS, and publishing is one button. Next.js is a React framework. The site lives in a code repository, a developer changes it, and a host like Vercel deploys each change.
So the real question isn’t which one is faster or cheaper. It’s who changes the site every week, and how much the site has to behave like the product. Everything else follows from that.
| Framer (our site) | Next.js (client site) |
|---|---|---|
Who changes pages | Designers, on the canvas; a Framer developer for CMS changes | Developers, in code |
Review before live | A branch, merged by a person | A commit, with lint and a build check |
Where content lives | Framer’s CMS (we write it in a repo and push it in) | Mostly in code; the CMS holds on/off flags and legal pages |
Motion | Built into the canvas | Code (Motion and GSAP) |
Product logic | None needed | Sign-in, sign-out and account deletion, calling the product’s API |
Hosting | Framer | Vercel |
The decisions
Who edits, and how changes get reviewed
On Framer, our designers change layouts and templates directly. When we added a lead heading to every blog post, it was one new CMS field and one text layer on the post template, and all posts had it at once. For review, Framer gives us branches: every change we push from our content system lands on its own branch, someone checks it, and only then is it merged and published. Between 5 and 18 November 2026 we published the site 18 times.
On the client’s Next.js site, a change is a commit. Every push deploys through Vercel, and a check runs lint and a full build before each commit. That’s a solid review step for developers. It also means a marketer can’t move a section without one.
Where the content lives
Our posts live in Framer’s CMS, but we write them as Markdown in a repository and push them in with scripts. That gives us version history, checks and one place to review everything. It also showed us the edges of Framer’s rich text: our scripts can add and change parts of a post but can’t remove them, so a big rewrite means rebuilding the post on a branch. Diagrams, like the ones in this post, have to be images.
On the client’s site, most long content lives in the code itself, as data files. The CMS holds only an on/off switch per article and some legal pages. That suits developers, since everything is typed and reviewed. Editors can’t change an article’s text without a developer, and there’s no draft preview: depending on the page, a CMS change either goes live straight away or waits for the next deploy.
Design and motion
Framer’s interactions and animations are part of the canvas, so our designers ship motion without code. On the Next.js site, motion is code: a shared scroll-reveal component built with Motion, and GSAP for six heavier sections. That gives full control, and every animation is a developer’s job.
Connecting to the product
This is where Next.js earns its place. The client’s account-deletion page loads a real login and calls the product’s API, and that login code is scoped to that one page, so the rest of the site never loads it. Our Framer site has nothing like that and doesn’t need it: its job ends at “talk to us”.
Performance
We measured both with Lighthouse on mobile settings, three runs per page, and took the median.
Our Framer site (18 November 2026): the home page scored 74 and transferred 6.9 MB, more than half of it a single hero video. A blog post scored 90 and transferred under 1 MB.
The client’s Next.js site (5 November 2026): the live home page scored 58, an article page 65 and the FAQ 41. Every page loaded 3.2 MB of font files, and the home page shipped 557 KB of JavaScript (gzipped).
These are different sites, with different content, measured on different days, so this isn’t a benchmark. What both runs show is that the weight comes from assets: a video on one, fonts and images on the other. Neither tool is the reason a page is slow.

Where the home page weight came from, from one Lighthouse run per site. The Next.js breakdown is from the same build served locally, which transferred about the same as the live site.
How each is set up

Same job, two paths to production: a branch and a publish button, or a commit and a deploy.
Write. On Framer, designers edit the canvas and our scripts push post content. On Next.js, developers change code and content files.
Review. On Framer, each change sits on its own branch. On Next.js, a pre-commit check runs lint and a full build.
Merge. On Framer, a person merges the branch. On Next.js, the commit lands on the main branch.
Go live. On Framer, one publish updates the whole site. On Next.js, Vercel deploys every push.
Redirects and hosting. Framer keeps redirects in its settings and hosts everything. On Next.js, the host handles the domain and the code would hold any redirects.
What it’s like to run
Framer means almost no operations: no servers, no builds, no deploy pipeline. The surprises came from the editor, not from hosting:
Preview only picked up our scripted edits after a reload.
A form on our Guides page never submitted, so we replaced it with a direct download.
Redirect rules can’t contain parentheses.
Once, a publish was blocked by an error in an unrelated component until it cleared the next day.
The Next.js site has the usual developer chores: dependency upgrades, builds and deploys. It also has more ways to go quietly wrong. Its home page is built once at deploy time, so new events from the CMS only appear after a redeploy.
What we learned
Match the tool to the person who changes the site every week. That decides more than any feature list.
A visual builder still needs a review step. Branches gave us one on Framer; without them, several people and scripts writing to one live site would have been risky.
Code gives control and moves the work to developers. Content in code is clean, and every copy change becomes a ticket.
Fix assets before blaming the tool. Video, fonts and images were the heaviest part of both sites.
When to move (either way)
Signs a Framer site should become code:
It needs real login, payments or data from your product on more than a page or two.
You want it to share components and design tokens with your app.
Your team writes content in a repository and the CMS has become the bottleneck.
Your content is deeply nested (items inside items) and keeps fighting the CMS structure.
The plan your CMS content needs costs more than the site is worth to you.
Signs a coded site should move to Framer:
Marketers wait on developers for page and copy changes.
The site is pages and a blog, with no product logic.
Nobody wants to own builds and upgrades.
We’ve made the first move for clients, from Framer to code. Each time, the content had outgrown Framer’s CMS, the plan for that content no longer paid for itself, and the interactions were more than the canvas could manage. By then the clients were finding Framer hard to work with, not easy.
A move in either direction means rebuilding pages, exporting the CMS content, and keeping every old URL alive with redirects so you don’t lose search traffic. We’ve written up what that looks like in a Webflow to Framer migration for SaaS. If you’ve decided on code and are choosing a framework, read Astro vs Next.js for marketing sites. If you’re choosing between visual builders, see Framer vs Webflow for B2B SaaS.
We’ve left plan and seat costs out on purpose: they change often, so check Framer’s pricing page for current figures.
Related reading
Whichever tool you pick, the pages matter more. Start with our SaaS website structure and what drives SaaS website redesign cost.
FAQ
For most, yes. Our own site and its 47-post blog run on Framer, with structured data, FAQs and redirects. It fits when designers and marketers change the site and it doesn't need product logic. Plan for its rich-text limits if you edit content with scripts.
Only the parts that live in a CMS. On the site we built, most article text lives in code, so editors can switch articles on and off but not rewrite them. A headless CMS with preview helps, and someone still has to build and maintain that setup.
Not by itself. In our Lighthouse runs, the heaviest pages on both sites were slow because of assets: a hero video on our Framer home page, and fonts and images on the Next.js site. A Framer blog post of ours scored 90 on mobile. Fix assets first.
Yes. Rebuild the pages in code, export the CMS content, and redirect every old URL to its new address so search engines carry rankings over. Move when the site needs product logic or shared components, not just because the team grew.
It doesn't have to. Keep the same URLs where you can, add a redirect for every one that changes, and carry over titles, descriptions and structured data. Then check indexing in Search Console for a few weeks after launch.
Yes. Framer has branches: changes sit on a branch until someone merges them, and publishing is a separate step. Every change we push to our own site with scripts lands on its own branch first, so nothing reaches the live site without a person checking it.
Engineering
Framer or code for your marketing site?
Tell us who edits the site and what it has to do. We'll tell you how we'd build it and why.
