Stannah · CRO · Decision support

DataPoints

We built a modular framework to scale websites. Then we used it to ask a harder question: which information actually moves a person towards becoming a customer?

Design Engineer & Design Manager
2021
CRO

Role

Product Design · Design Engineering · Product direction

System

WordPress and theFramework giving semantic information to Google Analytics.

Outcome

Context-aware experience, better conversion. Better data.

Tools & capabilities

WordPress · Javascript · HTML · Google Analytics

The Project

DataPoints didn’t start with analytics.

It started years earlier, with the architecture of the website itself.

The Framework was designed as a modular system for Stannah’s international websites. Instead of treating each country as a separate website, we created reusable structures, modules, layouts and calls to action that could be assembled in different ways for different markets.

The Framework separated page structure from page content. Markets could combine layouts, modules, product information, forms, FAQs, technical specifications, video, calls to action and other reusable elements without rebuilding the product each time.

Experimentation

We were already testing the information.

Across markets, we documented experiments around forms, social proof, product imagery, information above the fold and the amount of choice presented to users. The question was increasingly less about visual preference and more about what information helped people move through the funnel.

A/B testing was useful. But it still required us to decide what to test before the website could tell us anything.

Which information transforms a user into a customer?

The question that led to DataPoints.

DataPoints

Our pixels weren’t just pixels.

Every meaningful element could carry a semantic value.

Measurement rules

Seeing is not the same as loading.

A datapoint only counted when the user had a credible opportunity to consume it. The first version used simple rules to separate content that technically existed on the page from content that was actually exposed to the person.

Rule 01

Inside the viewport.

The information only became eligible after entering the visible area of the user's screen.

Rule 02

2 seconds in a row.

The element needed to remain continuously visible long enough to represent meaningful exposure.

Rule 03

Then, one point.

Only after those conditions were met did that semantic piece of information receive a DataPoint.

The important distinction

Correlation needed context.

A thousand views and a hundred conversions do not prove that one piece of information caused the conversion.

XZ information

Seen in 1,000 sessions.

100 conversions

Occurred in those sessions.

YZ information

Was also present during many of the same journeys.

Compare Patterns

The value came from reading combinations and relative exposure across sessions, pages and funnel moments.

DataPoints · product hypothesis

Instead of guessing the next A/B test, the website could start showing us where the interesting questions were.

What DataPoints unlocked

From reading pages to reading minds.

The first goal was better analysis. But the architecture pointed further: once the website understood what kinds of information a person consumed, it could eventually use those signals to shape the experience itself.

Global read.

Aggregate the same semantic information across pages and markets for different types of analysis.

Specific reads.

Understand which information matters at a particular page, product or funnel moment.

Dynamic content.

Use behavioural signals to infer what information a user is looking for and adapt future content accordingly.

Components carried meaning, not only styling.

A reusable block could have visual rules, behavioural rules and analytical meaning at the same time.

The system was designed to learn.

Build once, test repeatedly, collect behavioural evidence and feed the result back into the next product decision.

It anticipated context-aware experiences.

The long-term idea was not merely to report behaviour but to let the interface respond to what the user appeared to need.

The questions changed

The team could ask better questions.

Not “what content do we have?” but “what content is helping users decide?”

What do they read?

Identify the information users spend time with.

For how long?

Separate exposure from real attention.

Who converts?

Connect content behaviour with outcome.

What should change?

Use evidence to prioritise content and UX.

Driven by data.
Drawn by human nature.

Stannah · SMS Framework · DataPoints