Purple

PurpleWiFi analytics and guest engagement

Rebuilt to get found

Purple's website had grown too big to change and its enquiries were scattered across tools. We rebuilt it so the site publishes its own content in 13 languages, gets found in search and by AI assistants, and every enquiry lands in one place with its source attached.

Site migrationCustom buildContent automationMarketing automation
purple.ai
The purple.ai homepage after the rebuild
6,500+
pages moved to the new site
13
languages published automatically
3.4M
Google search appearances in one 28-day window
pages moved to the new site
6,500+
languages published automatically
13
Google search appearances in one 28-day window
3.4M

Purple provides WiFi analytics and guest engagement for more than 80,000 venues worldwide, from Harrods to McDonald's Belgium. The product was enterprise-grade. The website was working against it: thousands of pages nobody could change without a developer, a content operation that had stalled, and enquiries landing in different tools depending on which form someone happened to fill in.

This is a story about traffic. Not about the technology underneath, which gets one short section near the end, but about what the rebuild was for: making sure that when someone goes looking for what Purple sells, they find Purple.

The problem

Purple lives on being found. IT buyers search for answers about guest WiFi, network access and visitor analytics, and the company that shows up with the best answer gets the conversation. That only works if the site can keep growing, and Purple's could not.

The site had swollen to thousands of pages in 13 languages on a platform it had outgrown. Every structural change was made by hand. Publishing new content meant a content team Purple did not have, multiplied by 13 languages. A cluster of separate mini-sites ran on a different platform again and shared nothing with the main site. And when a prospect did enquire, the lead went into one of several tools with patchy source information, so "which marketing is actually working" was an argument rather than a question with an answer.

A site that cannot grow cannot be found more. That was the real cost.

What we built

We moved all 6,500+ pages onto a platform built for getting found, where content is data rather than hand-built pages. Four things matter about it:

  • Content that publishes itself. A pipeline drafts in-depth technical guides on the topics Purple's buyers search for, illustrates them, records audio versions, translates them into the other 12 languages and publishes the lot. Adding to the site no longer needs a content team; it happens on a schedule.
  • Pages built to be read. Every page is served fast and complete, readable by Google and by AI assistants. As more buyers ask ChatGPT instead of searching, the same pages answer both.
  • One front door for every enquiry. Every form on every Purple property feeds a single database. Each lead arrives within seconds with its source attached, gets enriched and classified, and lands with the right salesperson. Nobody argues about the numbers because there is only one set of numbers.
  • One source of truth for what Purple says. We compiled Purple's positioning from contributor interviews and 339 internal case studies into a single queryable messaging source, so every new page and campaign starts from approved facts rather than someone's memory.
Content autopilotPages in 13languagesFound in searchFound by AIassistantsEnquiries inone database
The site drafts and translates its own content, the pages get found in search and by AI assistants, and every enquiry lands in one database with its source attached.

What it means

More pages, on more of the topics buyers search for, in more languages, each one findable, and the count goes up every week without anyone lifting a finger. The scale this operates at: in one 28-day window in spring 2026, purple.ai appeared in Google search results 3.4 million times and was clicked through more than 10,000 times. Every page the autopilot adds is another way in.

<!-- [METRIC - confirm] organic traffic and lead-volume movement since cutover: no before/after series documented; do not publish a growth claim without it. --> <!-- Do not add Purple-internal figures here (SmartLead cleanup, Ads spend restructure) without Purple's explicit OK. -->

And when those visits turn into enquiries, none of them leak. Form fill to named salesperson, with the source intact, in seconds.

The technical detail, briefly

For the people who want it: the site is a Next.js monorepo backed by Supabase, server-rendered, with content held in Postgres and hreflang clusters wired per page across the 13 locales. Lead capture is one Supabase edge function behind every form, writing the raw payload before validation so nothing is silently lost, with database-level dedupe and an n8n workflow doing enrichment, classification and routing. The separate mini-sites are being consolidated into the same monorepo as apps sharing one schema. That is the whole section; the point of the system is the paragraph above this one.

The pattern

This is a productised pattern, not a one-off. For any business with a content-heavy site and a fragmented funnel: move the site onto a platform where content is data, point an autopilot at it so publishing compounds without headcount, route every form into one database so attribution survives, and compile the company's positioning into one canonical source so every asset starts from the same facts. Purple was the first tenant of this system. The same architecture deploys for the next client with their brand, their languages and their CRM swapped in.

How we did it

Related product

The lead capture and attribution system in this study, productised. Purple was tenant number one.

See Funnelwell

Next case study

Vets at Home

From The Pet Euthanasia People to Vets at Home: a rebrand and the platform behind it

Read it

Want a system like this?

Book a call. 30 minutes, no pitch deck. We will look at where AI can do real work in your business and tell you exactly what we would build first.

Book a call. 30 minutes, no pitch deck.