The number listed on a company's website is almost never the number you want. That's the main switchboard, the line a gatekeeper answers, the one that routes you to a voicemail tree. A direct dial or a mobile number reaches the actual person, and the gap between the two is the difference between 18 dials to a conversation and fewer than 2.
The catch is that direct dials and mobiles are the hardest contact data to source: the providers that get them right cover under half of any given list, and the ones that cover most of your list get a lot of them wrong. No single tool solves both problems. This is how to find a work phone number by chaining several so you connect on the first dial.
Step 1: Get the inputs a phone provider actually needs
A phone lookup is only as good as what you feed it. Phone providers do not work from a name alone; they match against an identity, and the more anchored that identity is, the higher the hit rate.
Before you run a single lookup, assemble three things for each contact: full name, company domain, and a professional profile URL. The name disambiguates, the domain ties the person to a specific employer (there are a lot of John Smiths, but far fewer at acme.com), and the profile URL is the strongest single anchor a provider can match against. In Clay, these are your corner pieces. If you start with just a name and company, run a professional-profile finder first to fill in the profile URL, then run phone enrichment off the completed record. Clay's own provider testing shows match rates swing hard on input type: the same provider that hits 81% coverage starting from an email drops well below that starting from a bare name.
Skip this step and your phone coverage collapses before you've spent a credit. Spend two minutes completing the identity first.
Step 2: Choose between a direct dial and a mobile number
These are not the same number, and the right one depends on who you're calling and where they sit.
A direct dial is a desk line that belongs to one person inside the company, so it bypasses the switchboard but only reaches them at their desk. A mobile number reaches the person anywhere, which is why mobiles connect at higher rates for senior decision-makers who are rarely at a desk. The main switchboard reaches a gatekeeper whose job is to keep you away from both.
What each kind of business number actually reaches
Tap each card to flip it and see what that line actually reaches.
Target the mobile when you're calling executives or field roles. Target the direct dial for desk-bound roles like inside sales or support. Both beat the switchboard.
A switchboard reaches a gatekeeper, a direct dial reaches a desk, and a mobile reaches the person, which is why mobiles win for senior decision-makers.
Step 3: Pick a phone provider, and understand why one is never enough
Phone data is the most punishing contact category there is, and the tradeoff between accuracy and reach is starker than anything you've seen with email.
Clay ran a controlled test across eight mobile-phone providers, scoring each on quality (how often the number is correct) and coverage (how much of a list it matches). The result is a wall: the providers that nail accuracy barely cover the list, and the providers that cover the list give back more wrong numbers.
Mobile-phone providers: quality vs. coverage (Clay data test, 2025)
Tap any dot to see its exact quality and coverage. The dots sit along a tradeoff line, not in one corner.
No single mobile-phone provider gives you both accuracy and coverage. At 90%+ quality only Forager and Prospeo remain, and they cover under 45% of the list, so any one tool forces a choice between a clean list and a complete one.
This is worse than email, where top providers clear 90% quality at 90% coverage. It is also far better than it used to be. In Clay's 2024 mobile test, most providers landed between 15% and 35% quality, and the US leader topped out near 59%. The category is maturing fast, but the structural tradeoff has not gone away, and it is the entire reason the next step exists.
Step 4: Chain providers so the first valid number wins
Picking the single best provider is the wrong move. The right move is to ask several in sequence and stop the moment one returns a number that passes validation.
This is how enrichment works in Clay. You set an ordered list of phone providers, cheapest-first. Clay queries the first one; if it returns a valid number for that row, the row is done and no other provider is billed. If it comes back empty or fails validation, Clay falls through to the next provider, and the next, until a number lands or the list is exhausted. You pay per successful lookup, not per provider, so adding a fourth provider costs nothing on the rows the first three already solved.
One contact running through a cheapest-first phone waterfall
Dana Lin · VP Sales · acme.comUsable coverage
44%
One provider alone: 44%
Cost so far
$0.000
Cheap providers run first; you only pay the chain until one lands.
Querying providers in sequence and stopping at the first validated number turns several mediocre coverage rates into one high one, and you only pay for the gates actually queried.
Teams that run phone enrichment this way describe the change in plain terms.
“Not having Clay would hugely reduce our ability to run good outbound campaigns. We wouldn't be calling people as much, because it's hard to get good phone numbers. We wouldn't be emailing as much, because we'd be likely to bounce or go to spam. With Clay, we have a reliable source.”
Order matters less than completeness here. Lead with a high-coverage provider to clear the bulk of the list cheaply, then let the high-accuracy specialists clean up the remainder.
Step 5: Validate every number before it reaches a dialer
A returned number is a claim, not a fact. Phone data decays as people change jobs and carriers, and an unvalidated list quietly fills your dialer with dead and reassigned numbers.
Build validation into the chain rather than bolting it on after. In Clay, a phone waterfall can require each candidate number to pass a validation check before the row is marked done; a number that fails sends the row back into the waterfall to try the next provider, exactly as a missing number would. This is the same logic Clay applies to email, where catch-all addresses can be excluded so only safe-to-send results survive. The effect is that your final phone column is not every number a provider offered but the first number that was both found and verified, which is a much shorter and much more dialable list.
The failure mode to avoid: treating coverage as the finish line. A 90% coverage rate built on unvalidated numbers is worse than 60% coverage you can trust, because every bad dial burns rep time and trains the team to stop using the phone at all.
Step 6: Confirm the number belongs to the right person and role
The last risk is subtle. A number can be accurate, validated, and current, and still belong to someone who left the company or moved into a role you don't sell to.
For high-value accounts, add an AI research step that verifies the person still holds the role you're targeting before a rep dials. Claygent, Clay's AI web research agent, reads public sources and returns a structured verdict you can branch on. Use a prompt like this.
Confirm whether {{full_name}} currently works at {{company_name}}in a role matching "{{target_title}}".Check the company website, recent press, and public professionalprofiles. Return JSON only:{"still_employed": true/false, "current_title": "...", "confidence": "high/medium/low", "evidence_url": "..."}Do not guess. If you cannot confirm employment, setstill_employed to false and confidence to low.
Route only the high-confidence, still-employed contacts to your dialer. The rest go back for re-enrichment or get dropped. This is the step that keeps a clean phone list from quietly going stale between the day you built it and the day a rep actually calls.
Common failure modes when finding work phone numbers
Most phone-finding problems trace back to a handful of repeatable mistakes.
- Starting from a name alone: Without a domain and a profile URL, providers can't disambiguate, and coverage craters. Complete the identity first (Step 1).
- Trusting one provider: Any single tool forces a choice between a clean list and a complete one. Chain them (Step 4).
- Skipping validation: Unvalidated numbers fill the dialer with reassigned and dead lines, and reps stop trusting the data. Validate inside the chain (Step 5).
- Confusing the switchboard for a direct line: The number on the website reaches a gatekeeper, not the buyer. Source the direct dial or mobile (Step 2).
- Letting the list age: People change jobs constantly; a number that was right last quarter may reach the wrong company today. Re-verify role before dialing high-value accounts (Step 6).