Why is direct mail attribution harder than digital?
Every digital channel leaves a data trail by default. A banner ad fires a pixel when it loads. An email registers an open when a 1x1 tracking image downloads. A paid search click carries a GCLID that follows the session into the CRM. Physical mail does none of this. Once the Royal Mail delivers your pack, nothing is recorded unless the recipient takes a deliberate action.
The problem is compounded by the time gap. A well-designed direct mail piece might sit on a kitchen counter for two weeks before prompting a purchase. By then, the recipient may have been retargeted on Facebook, received a follow-up email, and clicked a Google Shopping ad. Last-click attribution, still the default in most GA4 setups, assigns 100 per cent credit to whichever digital touchpoint came immediately before checkout. The mail piece, which may have planted the initial intent, disappears from the attribution report entirely.
That invisibility problem causes direct mail to be chronically undervalued in blended channel reporting. The solution is not to pretend direct mail works like digital. The solution is to use measurement methods designed specifically for offline channels.
Method 1: Unique landing URLs and QR codes
How it works
Assign a distinct landing page URL or QR code to each mailing variant, list segment, or creative version. When recipients visit that URL, you know the traffic originated from that specific mail piece. GA4 picks up the session, you tag the URL with UTM parameters (source: directmail, medium: post, campaign: spring26-segment-a), and conversion events are attributed to the mail channel in your reports.
QR codes lower the friction barrier substantially. On a smartphone, scanning is a two-second action. Typing a URL such as sortediq.com/spring26 into a browser takes longer but still works for recipients reading at a desk. Most campaigns now include both: the QR code as the primary mechanism and a short vanity URL as a fallback. The vanity URL also matters for call-to-action copy clarity on the piece itself.
What URL-based attribution misses
Only recipients who respond digitally are counted. Someone who received your pack, rang your sales number, and bought over the phone will show in your inbound call log but not in your URL attribution report. Similarly, a retail customer who walked into a branch because of a mailed voucher will not appear in any digital dataset. In sectors with strong phone or in-store response rates, URL-based methods can undercount mail-driven revenue by 40 to 60 per cent.
Variant-specific phone numbers (one number per mailing version, all routing to the same team) close part of this gap. They are more expensive to set up than landing page variants but capture the phone-in audience that QR methods miss.
Practical setup checklist for URL attribution
Before your mailing drops, confirm these four things are in place:
- Each mailing variant has its own landing page URL with correct UTM parameters already appended.
- GA4 (or your analytics platform) has a conversion event defined for the purchase or lead action on the landing page.
- The QR codes have been tested on both iOS and Android before the print file is signed off.
- The vanity URL redirects correctly and the redirect preserves UTM parameters in the final URL.
Method 2: Hold-out test cells
What a hold-out cell is
A hold-out cell (also called a suppression group or control cell) is a randomly selected portion of your mailing universe that you deliberately withhold from the mailing. Both the mailed group and the unmailed hold-out come from the same list, selected by the same targeting criteria. Over the measurement window (typically the campaign period plus a post-mail tail of two to four weeks), you compare the conversion or revenue rate of each group.
The difference is the incremental lift attributable to the mail piece. If the mailed group converts at 3.2 per cent and the hold-out converts at 1.8 per cent, the incremental conversion rate is 1.4 percentage points. Multiply by the number of pieces mailed and by average order value, and you have an attributable revenue figure that is causally defensible rather than correlational.
How large should the hold-out be?
For most UK direct mail campaigns, a hold-out of 10 to 20 per cent of the mailing universe is sufficient, provided the expected response rate is above 0.5 per cent. For campaigns below 5,000 names, a hold-out of 30 to 40 per cent may be needed to achieve 95 per cent statistical confidence. The calculation is a standard two-proportion z-test: if the z-score exceeds 1.96, the observed lift is statistically significant at the 95 per cent level.
A quick worked example: you have a list of 20,000 B2C consumer records. You mail 16,000 (80 per cent) and hold out 4,000 (20 per cent). Expected base conversion rate is 2 per cent, expected lift is 1 percentage point. With those figures, the minimum detectable effect at 95 per cent confidence, 80 per cent power, is approximately 0.7 percentage points. Your expected lift of 1 point clears that threshold comfortably, so the test design is sound.
Matched control design as an alternative
Pure random hold-outs are cleanest but not always possible. If your list was built by geographic territory and you cannot withhold a random 20 per cent without operational problems, a matched-control design offers a workable alternative. You select a control territory that is comparable to the mailed territory on the demographic and behavioural dimensions most likely to drive purchasing, then use regression adjustment to control for any residual differences. This approach requires more analytical effort but is legitimate when randomisation is impractical.
In our experience, marketing teams that present hold-out lift data to finance directors get direct mail budgets approved far more readily than those presenting URL-click attribution alone. The causal logic is straightforward and finance understands it.
Method 3: Post-purchase surveys
Why surveys recover what experiments miss
Hold-out tests measure aggregate lift. They cannot tell you which individual buyers were influenced by the mail piece versus those who would have bought anyway. Post-purchase surveys recover some of that individual-level signal by asking buyers directly. The key design point is to use a prompted channel list rather than an open-text question.
An unprompted question such as "How did you hear about us?" consistently under-reports direct mail because respondents do not think of a leaflet received ten days ago as a referral source. A prompted list that explicitly names "Post or direct mail" as one of six or seven options recovers 30 to 50 per cent more postal attributions than unprompted formats, according to DMA UK research on offline channel recall.
Where to place the survey question
The highest response rates come from embedding the question in the checkout flow itself, immediately before the confirmation button. A single-question insert at this point adds almost no friction and achieves completion rates of 60 to 70 per cent in tested e-commerce funnels. The order confirmation email is the second-best placement, with typical completion rates of 15 to 25 per cent. Standalone post-purchase survey emails sent more than 48 hours after purchase rarely exceed 8 per cent completion and should be used only as a supplement.
Calibrating survey data against hold-out results
Survey self-report and hold-out lift are measuring different things, so they will not give identical numbers. The hold-out measures incremental buyers created by the campaign. The survey captures which buyers recall the mail piece as influential. The right way to use them together is to treat the hold-out as your revenue attribution figure (the number you put in the board report) and the survey as a directional quality signal. If the hold-out shows strong lift but the survey shows low postal recall, that is worth investigating: it may signal that your mail piece is creating awareness that converts through other channels rather than directly.
Attribution methods compared
| Method | What it measures | Causal strength | Setup complexity | Best used when |
|---|---|---|---|---|
| Unique URLs / QR codes | Digital responses directly from mail piece | Medium (misses phone and in-store response) | Low (add to print file and GA4) | Your audience predominantly responds online; you need per-variant reporting |
| Hold-out test cell | Incremental lift across all response channels | High (causal if randomised correctly) | Medium (requires CRM or sales data per cell) | You need a revenue figure finance will accept; campaign is large enough for significance |
| Matched control design | Incremental lift where random hold-out is not feasible | Medium-high (depends on match quality) | High (requires analyst time for regression adjustment) | Geographic or operational constraints prevent random suppression |
| Post-purchase survey (prompted) | Individual-level channel recall among buyers | Low-medium (self-report bias; recall gap) | Low (add one question to checkout) | You want a channel breakdown by buyer segment; complementing hold-out with qualitative signal |
| Variant-specific phone numbers | Phone-in responses from mail piece | Medium (captures phone channel only) | Medium (telephony setup cost; routing configuration) | Significant share of your audience responds by telephone rather than online |
How to combine methods into a single campaign report
The standard reporting stack
For most UK direct mail campaigns, the practical attribution stack looks like this. First, assign variant-specific URLs and QR codes to capture the digital response baseline. Second, structure the list so that 10 to 20 per cent is randomly held out before despatch, and ensure that hold-out group is identifiable in your CRM or order management system. Third, add a prompted survey question to checkout or order confirmation. Fourth, once the measurement window closes (usually four to six weeks after mail drop), run the hold-out z-test to confirm statistical significance, then build your report around three numbers: total incremental conversions (from hold-out lift), digital-attributed conversions (from URL tracking), and survey-recalled postal influence (from survey data).
These three numbers will not add up to the same figure, and that is fine. Each is answering a different question. The hold-out lift answers "how many incremental purchases did this campaign generate?" The URL data answers "how many people clicked through from the mail piece specifically?" The survey answers "among all buyers in this period, what proportion say the mail influenced their decision?"
Integrating with digital attribution models
If you run GA4 with data-driven attribution, the URL-tagged mail sessions will already appear in your multi-touch credit allocation. The hold-out lift number is harder to feed back into GA4 directly, but you can record it as an offline conversion import. GA4's offline conversion import accepts a CSV of order IDs or client IDs with a conversion event and value, allowing the direct mail channel to receive credit in the same data-driven model as your digital channels.
For blended campaigns that combine postal with email, see also integrated direct mail and email campaigns for how to sequence the channels and set up a shared CRM identifier across both.
What a board-ready attribution slide looks like
A finance director reviewing your direct mail spend wants three things on one slide: the cost of the campaign, the incremental revenue attributed to it, and the return on ad spend (ROAS). From your hold-out test, you can give them all three with a confidence interval. "We mailed 80,000 consumers at a total cost of £48,000. The hold-out test shows 1,240 incremental purchases at an average order value of £62, generating £76,880 in incremental revenue. ROAS is 1.6x at 95 per cent confidence (z = 2.31)." That framing is far more persuasive than a URL click-rate, because it speaks in revenue rather than response mechanics.
For context on the typical revenue uplifts you should be targeting with UK direct mail, the benchmarks in UK direct mail ROI benchmarks 2026 give sector-level response rate and cost-per-acquisition ranges you can use to set realistic pre-campaign expectations.
Common attribution mistakes to avoid
Several errors recur in direct mail attribution work. The most expensive is using the mail date as the start of your measurement window rather than the estimated delivery date. Royal Mail business mail typically arrives two to four working days after posting. If you start your measurement clock on the day you hand packs to the mailing house, you will measure two to four days of background conversion rate before any recipient has seen the piece, which inflates your hold-out baseline and deflates your apparent lift.
A second common mistake is failing to suppress the hold-out from concurrent email or digital retargeting campaigns. If your hold-out group receives the same email sequence as the mailed group, you are no longer measuring the incremental effect of direct mail: you are measuring the difference between (mail + email) and (email alone). That is still a useful number, but it is not the question most campaign briefs are asking. Before despatch, confirm with your email and paid media teams that the hold-out CRM segment is excluded from all concurrent activity for the measurement window.
Third: do not report URL response rate as your primary attribution metric if your audience has significant phone or in-store behaviour. A 0.9 per cent QR scan rate looks weak against a 3.5 per cent email click rate, but if 60 per cent of your actual postal-influenced buyers rang in or walked in rather than clicking, the QR rate is a 40 per cent sample of true response. Report all three data points together, with the caveat that URL data represents the digital-response subset only.