Methodology
How we read the evidence.
Evidence before opinion. Here is how we interpret what customers and businesses tell us, and what we will not claim.
How conversation evidence is interpreted.
We look for patterns across conversations: questions that repeat, topics customers ask about before deciding, answers they cannot find, and reasons they hesitate. A single conversation is a data point; a repeated pattern is a signal worth acting on.
What data is used.
The questions visitors ask, the topics of those conversations, the details visitors choose to share, and the business information a business approves. In general web analytics we use IDs and categories, never message content or payment data.
Observation is not inference.
“Visitors keep asking whether you serve Burnaby” is an observation. “Your service-area information may be unclear” is an inference. We label which is which, and we treat a recommendation as something to test, not a guaranteed outcome.
What we do not claim.
- rankings or traffic outcomes
- statistics from samples too small to support them
- results we did not measure
- that a change caused a result without evidence
Evidence limitations.
Conversations show what people ask, not everyone who left without asking. Results depend on many things outside a website. Where evidence is partial, we say so.
Privacy and anonymization.
Aggregated findings are anonymized before they are published, and only used with permission or a contractual basis.