RESEARCH · WHAT WE BELIEVE
What we believe, and what would prove us wrong.
Four things Expanza is built on, how strong the evidence is for each, and the test that could change our minds.
1. Replying fast matters
The strongest evidence is old and American: companies that replied to a web enquiry within an hour were nearly seven times as likely to reach the decision-maker 1. Couples also decide quickly: 79% book within four weeks of getting in touch 2.
What we don’t have is a study showing that faster replies win more weddings at UK venues. The 2011 figure is about reaching people, not booking them. So we treat it as a strong direction, and the first thing we check in a client’s own records is whether reply time and bookings are actually linked there.
2. The answers are already in the records
Most businesses have years of enquiries, bookings and payments in systems that don’t talk to each other. We believe that joining them up shows where money and time are lost. The evidence is a mix of case studies and common sense, and data quality is often poor: one study found only 3% of companies’ data met a basic standard 3.
The test: in the first ten days of every pilot, we say plainly whether the records can answer the question. If they can’t, we stop.
3. A ranked list beats the simple rule
This is the belief we have least proof for. On our made-up test hotel, our ranking didn’t beat “reply to the newest first”, and a rule written from intuition did worse than both 4.
The honest offer today is: every enquiry answered fast and followed up properly, with the records used to point out the ones that deserve a call. Ranking is one part of that, not the headline, until real results say otherwise.
4. AI projects fail for human reasons
Across surveys and studies the pattern is consistent: projects fail because nobody uses the output, the data is poor, nothing was measured fairly, or one visible mistake destroyed trust 5. The model is rarely the problem.
That’s why we start with the problem and the people, put the output inside tools the team already uses, check every fact in code, and agree how we’ll measure before building anything.
What this means for a business
- Ask anyone selling you AI which of their beliefs they have least evidence for. If they can’t answer, be careful.
- Before any project, agree what would count as it not working.
What we don’t know
- Whether reply speed changes bookings at UK venues, and by how much.
- Whether owners value “more bookings” or “less chaos” more. Our first conversations will tell us.
Sources
- Fact Oldroyd, McElheran and Elkington, Harvard Business Review, March 2011 · US companies; reaching the decision-maker, not a sale
- Claim Bridebook Wedding Report 2026 · Survey of 7,000+ couples by a wedding platform
- Fact Harvard Business Review, 2017: only 3% of companies’ data meets basic quality standards
- Fact Expanza Labs: Wrenholt Hall results, 25 Sept 2026 · Our own test on made-up data
- Fact Why most AI projects fail (our explainer, with its sources)