Research
The find-rate study
How often does LocalContact find the people who run a local business, and how often are the names right? We're going to measure it on real businesses, checked by hand, and we've written down exactly how before we start.
Updated 16 Nov 2026
Results coming
The study hasn't started yet: no business has been picked or searched. When it's done, the results will be published on this page.
What we'll measure
- Found: for how many businesses LocalContact returns at least one person with a name and a job title.
- Decision maker found: for how many it returns someone in charge, such as the owner, founder, managing director, general manager or practice manager.
- Right names and titles: how often the people it returns really work there, in the role it says.
- Checked decision maker: for how many it returns a decision maker that a person then confirmed. This is the main result, because it's closest to what a customer gets.
- Differences: all of the above for businesses with and without a website, by country and by kind of business. The main comparison is businesses with no website against businesses with one, and we'll report it whichever way it comes out.
Which businesses
- 500 real local businesses in the United States, the United Kingdom and India, exactly half of them with no website.
- Five kinds of business: restaurants and cafés, heating and air conditioning, plumbers, dental practices, and hair and beauty salons.
- Picked at random from open data and public registers, such as OpenStreetMap, food hygiene ratings and company registers. Never from Google Maps, Yelp or similar sites, whose terms don't allow copying their listings.
- Chains and franchises with more than 10 locations, businesses that have closed and duplicates are left out. Two people check each business before the study starts, and whether it has a website is decided by hand: a page on a directory or a social network doesn't count.
How each business is searched
Once, with the same live search our customers use, from the business name and address only, with the same job titles for every business. Nothing is tuned for the study, and no customer account is used or charged.
How the names are checked
- Every name, by hand. A person looks each one up in official registers and licence lists first, then the business's own website and pages, news and directories, for up to 10 minutes per name. The short extract the search used doesn't count as proof: the page itself has to say it.
- One grade per name. Confirmed (an official register names them in a current role there, or two separate sources do), likely (one other source does), wrong (they aren't at this business now) or can't tell. The job title is graded too: right, nearly right, wrong or unknown.
- A second, blind check. A random 30% of the businesses are checked again by a different person who can't see the first answers. A third person settles every disagreement, and we publish how often the first two agreed.
- A phone check on a sample. For a random set of businesses, a person calls the business's public phone number and asks who owns or runs it, with a fixed script. We never sell anything on these calls and never call the people the search found.
How we keep it honest
- The method is fixed before a single business is picked. We'll register it publicly (on OSF Registries or as a dated public commit) before the study starts and link it from this page.
- Every random choice comes from a seed that is the registration's own id, so nobody could choose it, and anyone with the same sources can draw the same sample.
- Any change made after registration is listed with its date and the reason.
- A search that fails is tried once more; if it fails again, it counts as nothing found.
- Groups of fewer than 30 businesses are shown with their size but no rates.
- We publish the results whatever they are.
Privacy
- We never publish a real person's or a real business's name. Public files hold totals only, plus a made-up sample that shows the format.
- The study's data is kept apart from customer data and linked to no account. Names and other personal details are deleted 12 months after the results are published.
- We never email, message or call the people the search finds. Anyone who asks to be removed from the study is removed within 30 days: write to [email protected].
How the results will be published
- On this page: each result overall, by country, by kind of business and with or without a website, each with its margin of error (a 95% interval), and how often the two checkers agreed.
- A CSV file of the totals to download, and a made-up sample in the same format as the results.
- The same files in a public research archive (Zenodo) under a free licence (CC BY 4.0), with a permanent link (a DOI), plus a description of the dataset for search engines.
- The limits, in plain words: three countries and five kinds of business only, public sources can be out of date, and each business is searched once, on stated dates.
- A new edition about every six months, on a fresh sample, with older editions kept online.