Multichannel Attribution When One Touch Is Physical: A Practical Setup

The standard objection to adding direct mail to a multichannel motion is that you can't measure it. It's a strange objection, because the measurement most teams trust — digital touch data — is softer than it looks. Email opens have been unreliable since inbox privacy features started pre-fetching pixels. Click data is polluted by security scanners. View-through attribution on ads is a modeling exercise wearing a metric's clothing.

A physical touch, run properly, produces one piece of evidence harder than any of those: a FedEx delivery scan. A specific envelope reached a specific named person's office at a specific timestamp, backed by signature. That's not an inference — it's an event. The attribution question isn't whether the physical touch can be measured; it's how to join that event to the rest of your funnel data. Here is a practical three-layer setup, followed by the honest limits.

What you actually know, per channel

Attribution setups go wrong when they treat all touch data as equally solid, so start by ranking what you actually know. For email: that you sent it, and (noisily) whether it was opened. For calls: that you dialed, and whether a conversation happened. For LinkedIn: that you viewed or messaged. For a physical piece shipped through B2BMail: that it printed, that it shipped FedEx Priority to a named contact at an address verified before printing, and the delivery scan — date and time it arrived. That last stream comes from per-piece tracking: each envelope carries its own tracking ID, visible in the mail tracking dashboard in real time.

Notice what verification adds to the data quality story: because unverifiable addresses never print, the mailed population contains no phantom sends. Every row in your mail data represents an envelope that genuinely traveled to a genuine desk. Compare that to an email campaign where some unknowable fraction of the list was dead on arrival, and the physical channel starts looking like the clean end of your dataset — a point worth internalizing before diving into direct mail attribution models generally.

Layer 1: delivery-timestamp correlation

The first layer is the simplest and the one to build first. Take the delivery dates from the tracking dashboard and line them up, contact by contact, against the response events in your CRM: replies, connected calls, meetings booked, deal-stage changes. Then ask one question — what happened at each mailed contact in the 7 to 14 days after their envelope's delivery scan? Because delivery is timestamped per piece, the correlation is at the contact level, not the campaign level: you're not asking 'did response rates rise in March,' you're asking 'did Dana Chen, whose envelope landed March 4th, take a meeting March 9th.'

Operationally, this is a spreadsheet join, and it's worth being plain about the mechanics: B2BMail doesn't push data into your CRM — there's no integration. You export delivery dates from the dashboard, export contact activity from your CRM, and match on contact. A RevOps analyst can build the first version in an afternoon and refresh it weekly. Unsexy, fully within reach, and it produces the single most persuasive artifact in this whole exercise: a contact-level timeline showing envelope landings followed by responses.

Layer 2: matchback against the unmailed

Correlation timelines tell you what followed the touch; they can't tell you what would have happened anyway. That's the job of matchback attribution — comparing outcomes for the mailed population against a comparable unmailed one. The clean version: when selecting contacts for a physical touch, hold out a randomly chosen slice of otherwise-identical contacts who get every other touch in the motion except the envelope. After a quarter, compare meeting rates, opportunity rates, and progression between the two groups. The gap is the most defensible estimate of the physical touch's incremental effect you can produce.

If a deliberate holdout isn't politically feasible — nobody enjoys explaining why fifty good accounts were assigned to the control group — you can approximate with a natural comparison: similar-tier accounts worked in the same period without mail. It's weaker, because the accounts that got mail were usually chosen for a reason, and that selection bias flatters the mailed group. Say so in the readout. A slightly humbler number that survives scrutiny beats an impressive one that collapses under the first sharp question, which is the core discipline behind measuring direct mail ROI honestly.

Layer 3: direct response mechanisms on the piece

The third layer builds attribution into the mail piece itself. A QR code on the piece pointing at a UTM-tagged landing page turns a scan into a digital event your analytics already understand. A personalized URL does the same per contact. A named reply channel — 'email me directly at…' — makes inbound responses self-identifying.

Treat this layer as a floor, not a ceiling. Direct response mechanisms only capture the people who respond through the mechanism, and in B2B that's the minority path: the more common sequence is that the envelope lands, the rep calls the day it lands, and the meeting gets booked on the phone — a flow the QR code never sees. Count the scans, but never present scan volume as the measure of the channel. Layers 1 and 2 exist precisely because layer 3 undercounts.

The weekly routine that makes it real

• Monday: export last week's delivery scans from the tracking dashboard; append to the mail-events sheet.
• Same morning: export CRM activity (replies, connects, meetings, stage changes) for all mailed and holdout contacts; refresh the join.
• Flag every response event landing within 14 days of that contact's delivery scan; tag the deal so the touch survives into pipeline reporting.
• Monthly: compare mailed vs. holdout on meetings and opportunities; update the running incremental estimate.
• Quarterly: fold the numbers into channel planning alongside the KPIs you already track, and recompute cost per meeting.

An hour a week, most weeks. B2BMail's performance reporting gives you the delivery-side rollup — what shipped, what landed, when — and your CRM holds the response side; this routine is simply the join between them, maintained on a schedule instead of reconstructed in a panic before the QBR.

The honest limits

Now the part most attribution content omits. First: none of this proves causation on any individual deal. A meeting that followed an envelope by three days may have been coming anyway; only the matchback comparison, aggregated over enough contacts, separates lift from coincidence — and on small volumes, a quarter's gap can be noise. Report ranges and directions, not decimals. Second: in a multichannel motion, credit-splitting is a modeling choice, not a discovery. If a contact got four emails, two calls, and an envelope before booking, no mathematics can tell you what 'share' the envelope earned. Pick a consistent convention, disclose it, and resist the urge to defend the decimal places.

Third: delivery is not attention. The scan proves the envelope reached the desk — in B2BMail's experience FedEx envelopes essentially always get opened, but 'opened' and 'acted on' are different events, and the gap between them is exactly what your response data measures. Fourth: matchback quality is bounded by your CRM hygiene; if reps log meetings inconsistently, the join inherits every gap. The attribution setup is a mirror of your data discipline, not a substitute for it.

What to actually report

Resist the dashboard sprawl. Three numbers carry the argument: contact-level response rate within 14 days of delivery, incremental lift versus the comparison group, and cost per meeting for the mailed motion — computable directly in the cost-per-meeting math once you know pieces sent and meetings attributed. Presented together with their caveats attached, they answer the only question leadership is really asking: is the physical touch earning its cost? And because the physical channel's delivery data is per-piece, timestamped, and free of phantom sends, you'll often find it's the channel whose numbers you can defend most confidently in the room.

The short version

• A delivery scan is harder evidence than an email open — start your attribution thinking from what each channel actually proves.
• Layer 1: join per-piece delivery timestamps to CRM response events at the contact level; it's a weekly spreadsheet join, not an integration.
• Layer 2: matchback against a holdout or comparable unmailed group for the incremental story.
• Layer 3: QR codes and PURLs capture the direct-response minority; never mistake scans for the full effect.
• Report ranges, disclose your credit-splitting convention, and let the three headline numbers — response rate, lift, cost per meeting — do the talking.