Extracting phone numbers from Google Maps
Email extraction from Maps is a crawl with a 20-40% hit rate. Phone extraction is reading a structured field that almost every listing carries. This is the channel the data is actually good at, and the workflow that respects it.
Last reviewed
Why phone is the strong field
A Google Maps listing exists so customers can contact the business, and the phone number is the mechanism. It is a structured field, entered and maintained by the owner, displayed on every surface of the product. That gives extracted phone data properties no other contact channel in this dataset has: near-universal coverage on active listings, no crawling or guessing involved, and freshness tied to the listing itself.
Compare the alternative: email discovery means following the website link and crawling for a published address, with realistic coverage of 20-40% of rows. Phone-first is not a fallback strategy for local outreach; it is the strategy the data supports, and the one that reaches the half of local businesses that have no website at all.
Extraction at city scale
- Run a grid sweep of the category and metro, since a single search stops at roughly 120 results and the numbers you want are in the long tail past it.
- Export with every field, not just the phone: the category, rating and review count columns are what make the call list workable.
- De-duplicate on place ID, then on the normalised number, which catches one business under two listings.
- Drop permanently-closed listings before anything else touches the file.
- Normalise to E.164 in the same pass.
Working the numbers: routing beats volume
The export gives you three columns that turn a phone list into a call plan:
- Category predicts mobile versus landline, which predicts channel. Mobile-skewed categories (trades, tutors, solo services) take WhatsApp; landline-skewed ones (clinics, hotels) need a call and a front-desk script. The industry guides carry per-niche channel notes.
- Review count is the qualification signal: band it before dialling so openers match the business's maturity rather than a generic script.
- Opening hours, where published, are literally the answer-rate column. Calling a restaurant at noon wastes the dial; the field says so in advance.
Teams that route on these three columns before dialling convert the same list at a multiple of teams that work it top to bottom. The data was free with the extraction; the discipline is the differentiator, and the prospect-list workflow covers the rest of it.
Staleness, the honest weakness
Phone's failure mode is decay: businesses close, change providers, port numbers. On an unfiltered pull expect 5-15% dead numbers, worse in high-churn food and beauty categories. Three mitigations, in cost order: filter the closed flag at cleaning time (free), refresh the extraction quarterly so the list never ages past usefulness (cheap under a flat licence), and run a bulk line-validation pass before any large campaign (a fraction of a cent per number, cheaper than one rep-hour of dead dials). The data quality guide puts the field's reliability alongside every other column.
The compliance lines that matter for phone
Voice: screen against do-not-call registries where your market runs them; several apply to business numbers. Messaging: bulk WhatsApp and SMS carry both legal rules and platform enforcement, and account bans are the operationally binding constraint. Honour every opt-out permanently across channels. The legality guide covers collection and use in full; the summary for phone work is that enforcement follows dialling behaviour, so the routing discipline above is also the compliant posture.