10 things we learned shipping Anserra to production

A miscellaneous list. In no particular order.

1. The free plan is the marketing team

We priced Anserra with a real free tier — 100 messages/month, 3 chatbots, no card required. Every conversation about this internally centred on “won’t this cannibalise sales?” The answer, empirically, is no. The free tier is our top acquisition channel. People who try free convert faster and cost less than any other segment.

2. Honesty is a growth strategy

We say when a plan is running in demo mode. We say what our free tier does and doesn’t include. We publish real limits. This isn’t ethics theatre — it’s a growth strategy. Customers arrive already trusting the product because they read the docs and saw that we didn’t lie.

3. Onboarding beats features

The first thing that scared trial users off: connecting the widget. Not the retrieval quality, not the model choice, not the pricing. Getting the script tag onto their Shopify or Next.js site.

We wrote per-platform snippets (script tag, iframe, React, Next.js, WordPress, Shopify, Webflow, Wix) and time-to-live dropped from ~2 hours to ~5 minutes. Trial-to-paid conversion doubled.

4. The mistake that cost three days

We built the widget as a real script tag from day one — no iframe. It ran on the host page’s origin. This meant every customer’s CSS could and did break our widget layout in creative ways.

Rebuilding as an iframe with a `postMessage` protocol took three days we didn’t budget for. Should have been the design from the start. Cross-origin iframes are how widgets live.

5. Nobody reads long homepage copy

We wrote a beautiful homepage with three paragraphs of copy. Traffic bounced on the third scroll. We rewrote to a bento-grid layout with per-feature mocks. Conversion went up 30%.

The lesson: on a landing page, show more, tell less.

6. The billing page decides upgrades

Our first billing page was functional. Plan cards with checkmarks. Convert rate from free-to-paid: 3%.

We rewrote it with per-metric usage bars, a monthly/yearly toggle with a “2 months free” chip, and honest explanations of what each plan includes. Convert rate: 7%.

The billing page is where the buying decision actually happens. Design it accordingly.

7. Slack for handoff is the killer integration

Every trial user wanted Slack integration for human handoff. We ranked it #3 on the roadmap; should have been #1. Once we shipped it, trial conversion moved measurably.

The lesson: listen to what trial users ask for, not what your prioritisation framework says.

8. Documentation is a product surface

The docs pages generate ~40% of our marketing traffic from developer-focused Google queries (“embed chatbot Next.js”, “how to build RAG with pgvector”). We write docs like we write product because they are product for the developer segment.

9. Customers value speed of response more than perfection

We reply to trial-user emails within an hour, on average. This is not scalable indefinitely, but at our scale (small team, engaged trial base) it’s the single most valuable thing we do. Customers who chat with a human within an hour of signing up convert at 4× the rate of those who don’t.

10. The product is the marketing

Every time we’ve been tempted to spend on paid acquisition, we’ve reinvested in the product instead. So far it’s worked: word of mouth, developer share, and Google traffic from long-tail queries have carried us. There will come a point where we need paid; we haven’t hit it yet.

The through-line

Ship what customers actually need. Say what your product actually does. Fix onboarding before you fix anything else. Reply to emails fast.

There is no clever hack. This is what “product-led growth” means when you strip the marketing jargon.

If any of this resonates: try Anserra free. No credit card. Live in about five minutes.