I ran duo.ca on Webflow with the Finsweet suite for three years. Moving off was a call I resisted, then resisted, then made. The short version lives in the going to market pillar.
What Webflow was good for in 2023 and 2024
The Webflow build was a fast yes. I had been building in Webflow since 2018, the Finsweet suite covered attributes, CMS filters, and interactions, and I could ship a fully designed page in an afternoon without touching code. For a solo operator standing up a new brand, that independence was worth a lot.
The early site looked good. Homepage, about, first version of the insights blog, all shipped with clean design and a CMS I could maintain myself. Change a testimonial, add a service page, rework the nav, done in an hour in the browser.
For a brochure site with moderate content, Webflow is still one of the best tools on the market. That is not the argument. The argument is what happened when the strategy changed underneath it.
The first crack: SEO on blog CMS items
The friction showed up when I committed to a Wednesday cadence on the insights blog. Four posts a month, each with structured data, an OG image, a canonical URL pointing at the final duo.ca slug, and FAQ schema rendered cleanly for AI tools to pick up.
Webflow technically supports all of this. In practice, doing it reliably across dozens of CMS items involves workarounds. Meta descriptions on CMS items bind to a field, which works until you want to override it on a specific post. Canonical URLs require a custom field plus a head embed, fine at ten posts and fragile at a hundred. OG images work, but the title and description fields do not always sync cleanly with the underlying metadata.
FAQ schema was the harder one. Rendering JSON-LD inside a CMS item, pulling question and answer pairs from a referenced collection, and validating that Google and the AI crawlers saw it the same way, took more head-embed duct tape than I wanted to maintain. Every new post was a small chance for a schema mistake to ship silently.
None of this is Webflow being broken. It is Webflow being built for marketing pages, not for an editorial machine. Once my strategy depended on schema and metadata being perfect across a growing library, the tool was working against me.
The second crack: cost per page as the content strategy scaled
The insights blog grew from twelve posts to a plan for thirty-plus articles across two pillars, two pillar guides, and eventually programmatic SEO pages. Webflow's pricing scales with CMS items, site plan tiers, and workspace seats. At the volumes I was planning, the economics started looking wrong.
Not broken. Wrong relative to what I was getting. The per-page cost was paying for a visual editor I barely used anymore and a CMS that was actively slowing me down. On Next.js with MDX, the per-page marginal cost is effectively zero once the infrastructure is in place.
At a hundred pages, the annualized Webflow cost plus my time tax on every schema workaround was higher than the one-time cost of rebuilding. At two hundred pages, it was not close.
What made it tractable: Claude Code and AI-assisted development
I would not have made this call two years ago. Hiring an agency to rebuild duo.ca would have been $15K to $30K for a team that needed detailed specs and three months. I did not have that budget, and the ROI was hard to justify against a Webflow site that was technically working.
Claude Code changed the math. I am not a developer. I can read code, I have shipped small things with Cursor and Copilot, but I have never held an engineering role. With Claude Code as the primary IDE, I could describe what I wanted in plain English, get working Next.js code, iterate in real time, and ship to Vercel without writing component libraries by hand.
The six-week rebuild cost under $2,000 in API and tooling spend, plus my time. The AI tooling is the reason this decision is available to founders who could not have considered it a year ago.
What the rebuild actually looked like
Six weeks, running in parallel with client work and LinkedIn content.
Weeks one and two went to scaffolding the Next.js project, App Router, the MDX pipeline, and the design system. I built in a preview route so the old Webflow site stayed live. Weeks three to five were content porting. All twelve posts moved to MDX with updated frontmatter, FAQ arrays, cluster tagging, and canonical fields. Week six closed out schema rendering, FAQ components, sticky TOC, internal linking, and Vercel deployment. The production domain still points at Webflow until cutover, so I run the new site in preview without losing SEO equity on the old URLs.
The tradeoff I accepted is that I maintain the code. In exchange, the site renders exactly the way I want, ships a new page type in an afternoon, and emits clean structured data on every route. For AI citation and topical authority, that tradeoff is the right one.
What I would tell a founder weighing the same call
Webflow is still the right answer for most B2B companies. If your site is mostly positioning pages that rarely change, if you want marketing shipping without engineering, if content is not your primary channel, stay. The cost of rebuilding is real, and the business case is thin at that shape.
The honest signal that you should consider moving is specific. You publish more than thirty pages a year. Your SEO and AEO strategy depends on structured data being clean across the whole library. You want authoring in MDX, not a visual editor. You have AI coding tools you trust, or a development partner who uses them. When three of those four are true, the math starts working.
Do not switch because Webflow is uncool or because a developer told you it is not real code. Switch because the strategy has outgrown the tool. That is the only reason that holds up a year later.
For the deeper framing, see Next.js vs. Webflow for a B2B marketing site. For the related service, see AI-built web presence.
A short coda
I am glad I moved. I am also glad it took three years, because the case only became clear once the strategy demanded it. The lesson is not that Webflow was wrong. The tool has to match the work, and the work changes faster than most founders update their stack.
Common questions.
Was Webflow actually causing SEO problems for DUO?
Yes, though not in the way most SEO audits would flag. The issue was not that Webflow cannot do SEO. It was that schema, canonical URLs, and OG tags on CMS items required workarounds that piled up across a growing content library. For a site whose strategy depends on AI citation and structured data, those workarounds were a tax I no longer wanted to pay.
How much did the rebuild cost?
Direct build cost was under $2,000 in Claude API and tooling spend, plus my own time. Compared to a Webflow rebuild at agency rates, which would have run $15K to $30K for equivalent scope, the economics were not close. The real cost was the decision itself, not the execution.
How long did the migration take?
About six weeks, working on it in parallel with client work and LinkedIn content. Roughly two weeks to build the new Next.js foundation, three weeks to port content and hit visual parity, one week to finalize schema, FAQ rendering, and internal linking. A full-time focused sprint could have done it in three.
Would you recommend Webflow for other B2B companies?
For most of them, yes. Webflow is the right call if your site is a brochure, if your team ships five pages a year, and if you want marketing independence from engineering. I would not recommend it if content is a primary channel, if you are shipping dozens of SEO pages a quarter, or if AI citation is central to your distribution strategy.
