Why I deleted 4 features before launch

I had 9 CMS integrations planned for my SaaS. I shipped 5. The other 4 got deleted before a single line of adapter code was written.

This saved roughly 6 weeks of build time and one likely refund cycle from disappointed early users.

The product is Articfly. It generates SEO blog articles and publishes them to your CMS. Multi-provider was the differentiator from day one. The original list: WordPress, Shopify, Ghost, Notion, Custom API, Webflow, Wix, Feather, Framer.

I spent two weeks doing API discovery before writing any adapter code. That two weeks killed four of the nine.

What I cut

Framer. No public CMS write API. You can read content. You cannot create or update pages. The integration would have been a "Coming Soon" button forever.

Feather. Notion-to-blog product. Public Blog API exists but is strictly read-only. The realistic path to Feather users is Notion (they use Feather to render what they wrote in Notion). So I built Notion instead.

Webflow. API exists. Build is 2-3 days. Zero customer signal asking for it. Articfly was three weeks from launch and I needed to ship what users were actually requesting.

Wix. Headless Blog API exists. Per-instance OAuth is complex. Same problem as Webflow. Nobody had asked. Held.

The 3 questions I now run before any integration

1. Does a public write API exist today? If "soon" or "private beta", it does not exist.

2. Has a real customer asked by name in the last 30 days? Not "would be cool". Asked by name.

3. Realistic build time including OAuth and edge cases, plus 50 percent? Still fits the sprint? Queue it. Otherwise hold.

The four cuts failed question 1 or 2. WordPress, Shopify, Ghost, Notion, Custom API passed all three.

The math I avoided

Framer: impossible = 0 hours, 1 dead page

Feather: impossible = 0 hours, 1 dead page

Webflow: 3 days build = ~25 hours

Wix: 6 days + OAuth = ~50 hours

Total avoided: = ~75 hours + 2 dead pages

Plus support cost on integrations that were secretly broken. Plus the trust hit when users noticed. Plus marketing copy I would have walked back later.

I built https://articfly.com on the principle that subtracting is easier than supporting. Five integrations cover roughly 80 percent of the stack I wanted to reach.

What is the feature you almost built that would have been a disaster?