Address intelligence — office-first lists and the remote-worker filter.

TL;DR

• Never ship to a home address. A researched home address reads as surveillance, not delight. Office addresses only, without exception.
• Never email asking for a shipping address. Cold recipients don’t reply, the ask kills the surprise, and the campaign’s economics die on the response rate.
• Build the list offices-first, people-second. Enumerate every corporate location per account, then find the most senior person within ~200 miles of one. The intuitive inverse — find execs, then match addresses — collapses the list.
• Optimize for the most senior person who plausibly works in person — sometimes the COO, not the CEO.
• Filter remote workers by distance: claimed location more than ~50 miles from the nearest office means don’t ship, or pick a different person at the account.
• Build the re-send path before launch. The filter is a prediction; misses become recovery conversations, not losses — but only if the re-send workflow already exists.

The premise

Every other outbound channel fails silently. A cold email to a dead inbox bounces unread; a display ad to the wrong account wastes cents. Nobody at the target ever knows the attempt happened. Direct mail inverts this. Every failure is physical and visible: a package to an office the person left two years ago, a gift sitting in a mailroom for a recipient who works from home three states away, a courier at the door of a registered-agent address that is a filing cabinet in a law office. Some fraction of those failures are witnessed — by the mailroom, by a colleague, occasionally by the intended recipient hearing about it secondhand.

This is why address intelligence is the hidden hard part of the channel. Gift selection is a taste problem and fulfillment is a vendor problem, but the address is a research problem, and it’s the one input the operator cannot buy off the shelf: contact-data vendors sell emails and phone numbers at high match rates, but nobody sells a reliable mapping of person → the physical desk they will actually sit at next Tuesday. The teams that run this channel well — Clay documented their own version publicly in their “How Clay Uses Clay” teardown, which several operating numbers here draw on — treat the address as a first-class research output with its own pipeline, verification pass, and error budget.

The stakes are asymmetric. A correct address buys the seller ten seconds of manufactured warmth and an opener. An incorrect one wastes the ~$20 unit cost, which is survivable. But a creepilysourced one — a home address, a “we know where you live” signal — burns the account, and Tier 1 accounts are finite. There are only 500 Fortune 500s. You cannot un-creep a CFO.

The two hard rules

Rule one: never ship to a home address. This holds even when the home address is trivially findable, even when the recipient is fully remote, even when the gift is thoughtful and cheap. The reason is what the delivery says. A package at the office says: a vendor did clever research about my company. A package at my front door says: a stranger researched where I live. The identical gift flips from delight to surveillance on nothing but the delivery point, because the office is a professional surface the recipient expects to be reachable at and the home is not. No note card survives the second reading. Recipients who work from home are excluded from the shippable list — that’s what the remote-worker filter below is for — or represented by a different person at the same account. They are never reached at home.

Rule two: never email asking for a shipping address.The “we’d love to send you something, where should we ship it?” email is the intuitive compliance-safe move, and it fails three ways. First, arithmetic: cold recipients don’t reply to vendors, and a campaign gated on a reply inherits cold email’s single-digit response rate before a single package moves — the 300-account campaign becomes a 12-account campaign and the economics die on the spot. Second, the ask destroys the mechanism: the gift works because it is a surprise, and an announced gift is a transaction the recipient must now evaluate as a bribe. Third, the reply-gate selects for the wrong people — those most willing to hand an address to an unknown vendor are rarely the senior, skeptical buyers the campaign exists to reach. The correct posture is the uncomfortable-sounding one: research the office and surprise them there.

Invert the lookup: offices first, people second

The intuitive pipeline runs person-first: pull the senior titles at each target account, look up where each person sits, ship. It collapses quietly in practice. A large fraction of executives have a claimed location — the city on their LinkedIn profile, the metro in the contact-data record — that doesn’t match any shippable office: they’re remote, they’re in a satellite city the data vendor doesn’t know about, or their profile location is three roles stale. Each mismatch deletes a row. Start with 300 accounts and one hand-picked executive each, apply an honest “can we actually ship to this person” test, and the surviving list is often less than half the campaign you budgeted — with the misses concentrated in the largest, most distributed, most valuable accounts.

The fix is to invert the order of operations:

• Step one — enumerate every corporate location per account. An AI research agent takes the account domain and returns a structured list of allphysical locations — headquarters, satellite offices, branch offices, regional hubs — each with a full street address, held as structured data on the account row. Not just the HQ from the data vendor’s firmographic record: the careers page, the contact page, office-opening press, the Google Maps listings. For a mid-market SaaS company this is two or three addresses; for a bank or insurer it can be hundreds, and that breadth is precisely the asset.
• Step two — find the most senior person based near one of those locations. A second research pass per account: given the office list, find the most senior relevant person whose claimed location falls within roughly 200 miles of any office on it. The radius is a search parameter, not a shipping decision — the stricter shipping filter comes later.

The ordering matters because of what each version does to list size. Person-first fixes the person and hopes an address matches — every failed match is a lost account. Office-first fixes the much larger set of shippable destinations and searches the org chart for someone near one — every account with a real office and a senior person near it stays in. The list stays large, and every row is shippable by construction: the address was verified before the person was chosen, not reverse-engineered afterward. Same accounts, same research budget, roughly double the shippable rows.

Seniority versus presence

Office-first changes the question you ask about people. Person-first asks: who is the most senior buyer? Office-first asks: who is the most senior buyer who plausibly walks into one of these buildings? Those are different people surprisingly often, and the second is the correct target — a package on the desk of a present COO beats a package in the mailroom pile of an absent CEO every time, because the entire value of the channel is the moment of physical receipt.

The instructive case from Clay’s own campaign was Bank of America scale: the CEO of an institution that size is perfectly findable, but matching him to a specific shippable desk is a low-confidence guess. The COO matched cleanly to a Charlotte office. Marginally less senior, dramatically more shippable — and at that seniority the distinction is irrelevant to outcomes: either title can cause the deal, and neither will sign the contract off the back of a $20 gift. What the gift needs is a real desk under a real person with real influence. Take the seniority discount; it costs nothing and buys certainty of receipt.

Operationally: rank the relevant titles per account, walk down the list, and take the first person who clears the presence bar. Do not fall in love with the org chart’s apex.

The remote-worker filter

Step two’s 200-mile search radius finds candidates; it does not clear them to ship. The shipping decision runs a stricter computation: distance in miles between the person’s claimed location and the nearest office on the account’s list.Compute it as a literal number per row — a number supports a threshold and an audit trail, where “seems close enough” supports neither.

Beyond the threshold, don’t ship — or pick a different person at the account. The threshold itself is a risk-tolerance dial, not a law of nature:

A person 400 miles from the nearest office is remote with near-certainty; a person 8 miles away is plausibly at a desk twice a week. The threshold draws the line through the ambiguous middle, and where to draw it depends on the cost of a miss. What no threshold does is reach certainty — which is the next section.

The filter is a prediction — budget the misses

Distance-to-office is a proxy for in-office presence, and proxies miss. Hybrid workers who badge in twice a month, stale profile locations, the VP who moved and never updated anything — some packages will arrive at offices their recipients no longer visit. Accept this at design time: the filter’s job is to push the miss rate down to where the economics work, not to zero. A campaign that demands a guaranteed address before shipping never ships.

What converts the residual misses from waste into assets is the recovery conversation. The follow-up sequence surfaces them naturally — “we sent a package to your desk at the Charlotte office, did it land?” draws the reply “I never got it” or “I actually work from home.” That reply is one of the easiest openers in sales: the recipient has initiated contact, over a low-stakes logistics topic, with a mild social pull toward continuing. The play is mechanical — apologize, offer to re-send to whatever address they prefer (they are now volunteering it, which rule two forbade you from requesting cold), and keep the thread alive. The failed delivery bought the same ten seconds the successful one would have.

The operational requirement is that the re-send path exists before launch: inventory reserved for re-sends, a fulfillment flow that accepts a one-off corrected address, and a named owner for the reply. A recovery conversation that ends in “let me check whether we can send another one” and a four-day silence squanders the opener. Build the path first; the misses will use it in week one.

Verification tactics

Between the automated pipeline and the shipping label sits a verification pass. Four checks earn their cost:

Cross-check the claimed location against the office list.The person’s LinkedIn location is the primary input to the distance filter, so audit it: does the profile actually say a city near the matched office, and is the role recent enough that the location is plausibly current? A profile last touched three roles ago is weak evidence. Where the profile and the data-vendor record disagree, trust neither and prefer a person at the account whose signals agree.

Confirm the office is a real, occupied location.Corporate-registry data is polluted with registered-agent addresses — legal service points at a law firm where no employee has ever set foot — and with offices vacated eighteen months ago whose listings outlived the lease. Signals of a real, occupied address: a Google Maps listing with photos and reviews, current job postings citing the location, the address on the company’s own careers or contact page, recent employee posts geotagged there. An address that appears only in registry filings ships nothing.

Assume mailroom interception at large offices. At a 2,000-person building, packages do not walk from the loading dock to a named desk unmediated; reception, mailrooms, and executive assistants intercept, screen, and sometimes open them. This is fine — plan for it. It is why the gift and note must be self-explanatory to whoever opens the box, and it is the logistical argument for advertising to the whole buying committee rather than sniping the one recipient: whoever intercepts the package should already have heard of you. The ads chapter in this cluster covers the mechanism; interception is the default at scale, not the exception.

Spot-check a sample by hand. Before a 300-package run, manually verify twenty random rows end-to-end — person, presence signals, office, occupancy. The sample error rate says whether the pipeline is ready or a prompt or threshold needs another pass. Shipping is the one step you cannot un-run.

Address data hygiene

The pipeline above produces data that must live somewhere with structure, because every downstream trigger keys off it.

Office lists are structured data on the account: one row carries the full enumerated location set — street address, city, source, occupancy evidence — not a comma-mangled blob in a notes field. The list outlives the campaign; the next campaign into the same account starts from a verified location set instead of from zero.

The shippable unit is the person-plus-office pair, one row each: the chosen person, the matched office, the computed distance, the threshold it cleared, the verification evidence. Not “person, with an address column” — the pair, as a first-class record. When the recovery conversation later reveals the person works from home, you correct that row’s pairing without corrupting the account’s office list, and the audit trail shows which prediction failed and why.

Delivery status is first-class campaign state. Queued, shipped, delivered, confirmed, missed, re-sent— tracked per pair, updated from the fulfillment vendor’s events, and treated as the spine of the campaign, because the orchestration layer keys everything off it: shipment triggers the ads, delivery triggers the follow-up email, a “never got it” reply flips the row to missed and opens the re-send path. A campaign that cannot answer “which packages arrived yesterday?” from its own data cannot time its follow-ups, and “did it land on your desk?” sent three days before it lands burns the exact credibility the package was meant to buy.

Common operator failures

• Shipping to a home address “just this once” because the target is high-value and fully remote. The gift reads as surveillance, and Tier 1 accounts do not regenerate.
• Emailing for the address. Reach collapses to the cold-reply rate, and the surprise — the entire mechanism — is spent before anything ships.
• Building the list person-first. Half the hand-picked executives fail the address match, and the misses concentrate in the biggest accounts.
• Insisting on the CEO. The apex title with an unshippable address beats out the COO with a verified desk; the package joins a mailroom pile addressed to someone never in the building.
• Eyeballing the distance filter. No computed miles, no threshold, no audit trail — per-row vibes, applied inconsistently.
• Treating the filter as a guarantee.No re-send inventory, no corrected-address flow, no owner for the “I never got it” reply — the easiest conversation in the campaign dies in the inbox.
• Shipping to registered-agent addresses.The registry said it was the company’s address; it was a law office. The package is signed for by a paralegal and recycled.
• Keeping addresses as free text. No structured office list, no person-office pairing, no delivery states — nothing downstream can trigger, and the follow-up asks about a package still sitting in a warehouse.

Pre-launch checklist

• Office enumeration run per account; every location structured, with source and occupancy evidence
• Registered-agent and stale addresses purged from the office lists
• Person selection run office-first: most senior relevant person within ~200 miles of a listed office
• Distance-in-miles computed per pair; threshold chosen deliberately (50 standard / 25 strict / 10 strictest) and applied uniformly
• No home addresses in the shippable list; no address-request emails in the sequence
• Twenty-row manual spot-check passed at an acceptable error rate
• Re-send path built and owned: reserved inventory, corrected-address flow, named owner for recovery replies
• Delivery status schema live (queued / shipped / delivered / confirmed / missed / re-sent), wired to the fulfillment vendor’s events
• Buying-committee awareness layer scheduled around delivery, on the assumption of mailroom interception

Where this fits

Address intelligence sits between gift selection and fulfillment. The gift chapter decided what ships; this chapter decides where and to whom, and produces the two artifacts everything downstream consumes: the shippable person-office list and the delivery-status spine. Fulfillment executes against the list; the orchestration and follow-up layers key their triggers off the status field.

It is also the layer that determines whether the channel is safe to run at all. Every other failure in a gifting campaign wastes money; an address failure of the wrong kind — the home delivery, the surveillance read — spends account trust that took quarters to build. The teams that get durable results from direct mail are not the ones with the cleverest gifts. They are the ones whose address layer is boring, structured, verified, and built to recover from its own misses.

Related chapters

• Gift Selection — what ships: cheap, on-brand, and experience-first beats expensive and generic.
• Fulfillment and Vendors — executing the shippable list through gifting platforms and delivery-event APIs.
• Buying-Committee Ads — the awareness layer that makes mailroom interception work in your favor.