The Point of Diminishing Postage: Auditing Door-Drop Frequency in UK Campaigns

Most UK door-drop budgets are decided in two numbers: how many households, and how many times. The first number gets argued over. The second is usually settled by a sentence in a planning meeting about 'reinforcement' or 'staying front of mind'. Those effects are real enough, but they are not measurable at the point where a purchase order is signed. The commercial question is narrower and answerable: does the second, third or fourth leaflet to the same household generate enough additional response to pay for its own print and distribution? Answering it requires a campaign designed to answer it before the first van leaves the depot.

Frequency is an assumption wearing a media plan

Unaddressed distribution in the UK is sold overwhelmingly on coverage. Royal Mail Door to Door, Whistl and the independent leaflet networks all price per thousand households in selected postcode sectors, with minimum volumes that discourage small experiments. That commercial shape pushes advertisers towards one large, well-covered drop rather than a set of frequency cells, and towards repeating a successful drop without testing whether the repetition itself did any work. The result is a familiar pattern: a campaign that worked once is run three times, response per drop falls, and the fall is blamed on creative or on the offer rather than on the marginal value of the fourth leaflet.

Building cells that can be defended

Household-level randomisation is not available in unaddressed distribution. Every household in a selected sector receives the leaflet, so the only workable randomisation unit is the postcode sector. That is a constraint, not a barrier, but it has consequences for design.

Sectors should be assigned to frequency arms — one drop, two drops, three drops — after matching on geodemographic profile, dwelling type, tenure and, where a customer file exists, historical response by sector. Arms should be matched rather than simply equal in number, because a frequency effect of two or three tenths of a percentage point is easily swamped by a mismatch in affluence or in the proportion of flats.

Cell sizing is where most plans fail. If expected response is around 0.8 per cent and the smallest commercially interesting increment is 0.15 percentage points, the required sample per arm runs into the tens of thousands of households once intra-cluster correlation between sectors is allowed for. With a 50,000-piece minimum order, that frequently leaves room for only two or three arms in a single campaign. The realistic answer is to run the frequency test across two campaign cycles and pool the results, rather than to squeeze six arms into one buy and report noise.

The attribution trap

Suppose three drops go out with the same URL and the same QR code. Responses arriving after drop three will include households that saw drop one and simply took their time. Credit those to drop three and the third leaflet appears far stronger than it is. This is the single most common error in door-drop frequency reporting, and it flatters exactly the drop that should be under the most scrutiny.

There are two defensible fixes. The first is a distinct response key per wave — separate short URLs, separate QR destinations, a dedicated Freephone number or a match-back key on a coupon — combined with a declared lag window per wave, for example fourteen days of attributed response with a further twenty-eight-day tail reported separately. The second, and stronger, is a held-back final drop: deliver drops one and two to every sector in the test, then withhold drop three from half the matched sectors and compare cumulative response at day fifty-six. The difference is the increment, and it is measured rather than inferred.

JICMAIL data on item-level attention is useful context here, and it is often misused as a substitute for this work. It describes how many times a piece was handled, not how many extra sales a repeat delivery produced. Attention and increment are different currencies.

The arithmetic of the third drop

Take an illustrative case. Distribution at £48 per thousand and print at £22 per thousand gives a cost of £70 per thousand delivered. Contribution per response, after fulfilment and discount, is £26. Break-even incremental response is therefore 70 divided by 26, or 2.69 responses per thousand — 0.269 per cent.

Now the observed increments from a properly keyed three-wave test: drop one at 0.82 per cent, drop two adding 0.31 percentage points, drop three adding 0.09. The second drop clears break-even comfortably. The third does not. Nine tenths of a response per thousand returns £23.40 against £70 of cost — a loss of £46.60 per thousand, or £11,650 across a quarter of a million households.

Cumulative response in that example rises from 0.82 to 1.13 to 1.22 per cent. The second drop adds a 38 per cent relative lift; the third adds 8 per cent. That shape is not a law, but it is common enough in considered-purchase categories that any plan assuming a flat third drop should be asked for its evidence.

Wear-out, wear-in and the creative confound

Frequency and creative are usually tested together and then reported as one finding. If the second drop carries new artwork, the increment cannot be attributed to repetition rather than freshness. A useful test carries four arms: same creative twice, fresh creative twice, same creative three times, and fresh creative three times. The cost is real, but so is the confusion it removes.

There is also a genuine wear-in effect in some categories, where a second identical leaflet outperforms a fresh one because the household recognises it and has already formed an intention. It is common with low-value FMCG coupons and rare with high-consideration purchases, where a repeated identical message tends to read as a reminder that the household has already declined.

Turning the answer into a decision

A frequency audit should end in one of three actions. Cap frequency at the point where marginal contribution turns negative and reallocate the marginal budget to reach extension in fresh sectors, which almost always beats a fourth drop. Or hold frequency and change the offer, since an exhausted audience responds better to a new proposition than to the same proposition more often. Or renegotiate with the distribution partner: repeat drops into sectors already covered are frequently cheaper per thousand than fresh coverage, which moves the break-even and can justify a third delivery that would otherwise fail.

None of this requires an elaborate research programme. It requires distinct response keys, matched sector arms, a declared lag window and the discipline to report marginal rather than cumulative response. The third leaflet is not a question of persistence. It is a marginal-cost question, and it has an answer.