Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Brand Marketing › Expert

Planning Evidence for a Radio and Podcast Burst for a Bed Linen Brand

You plan, before any money is spent, how an invented online bed linen brand will know whether its first radio and podcast burst worked: the response routes, the reads around airings, the search and direct lift measures and a regional comparison, with what only a mix model could show later. You then design a regional test of brand cues in its performance ads, with its metrics, guardrails, timing and decision rule, placed so the two tests never blur into each other.
ExerciseExpert levelModule 3: Brand evidence the board can act on5 minUpdated

Why this matters

Once the advanced brand marketing level has won the money for brand media, the harder question is what will count as proof. That evidence has to be fixed before the spend, so nobody picks the flattering reading later, and radio and podcasts leave no click for reports to credit.

The linen brand and its first broadcast spend

Fremquaskel Linen is an invented online bed linen brand. It sells only through its website, advertises with paid social video and paid search, and has never bought broadcast media. Its packaging and ads share an indigo-and-cream stripe and a folded-corner logo, but its performance videos show them only in the end frame.

Illustrative example — not a real result

Fremquaskel Linen: the plan, the baseline and the proposals

LinePlanCan be bought by region?
Regional radio£36,000 over 8 weeks, breakfast and drive time, logged spot timesYes, in any of its 8 sales regions
Podcast host reads£18,000 across 3 national sleep and home showsNo
Performance ads, running£22,000 a month, paid social video and paid searchYes, in 16 areas, 2 inside each sales region

Last year: 2,600 orders a week, spread evenly across regions, at £46 contribution each. Regional orders and branded searches go back 2 years; the regions form 4 matched pairs whose orders stayed within about 3% of each other over any 12-week window. Checkout never asks how buyers heard of the brand, and a test redirect dropped its campaign parameters.

The agency proposes one code, LINEN30, for 30% off on every station and show, with podcasts judged on show-note clicks. The marketing manager wants radio in all 8 regions and a verdict after week 2.

Evidence for media that leaves no click

The guide to measuring TV, radio, podcasts and print sets out a ladder: response mechanisms that count a floor, response around airings, search and direct lift against a Baseline, asking people, regional tests and mix models, chosen in proportion to the money and the medium. The brand cue ad test uses the regional rung to ask whether performance ads should carry the brand's cues early, read on two clocks: cost per result now, branded search and orders later.

Task

  1. Choose the response mechanisms for radio and each show, and answer the agency's code proposal.
  2. Say how airings will be read, and why that read cannot settle whether the burst paid.
  3. Set the search and direct lift measures, their baseline and what they are followed through to.
  4. Design a regional test where the medium allows one, and say what effect it must see.
  5. Say what only a mix model could show later, and what this burst gives one.
  6. Design the brand cue test: arms, metrics, guardrails, duration and decision rule, clear of the radio test.

Worked answer

Response mechanisms. Radio and each show get their own short address and code, with one modest offer everywhere so placements can be ranked. Redirects are retested until they keep their parameters, as the offline media flight launch checklist requires, and checkout gains the question. The agency's code is refused: it ranks nothing, measures the discount and leaks to deal sites.

Airings. Station logs give real spot times; sessions and branded searches by the minute are compared with the same slots on days without a spot, in a window fixed before analysis. That ranks stations and dayparts but misses next week's buyers.

Search and direct lift. Weekly Branded Search, paid and organic together, and direct visits by region, against last year's weeks adjusted for trend and against regions without radio, with promotions annotated. A rise is attention until it is followed through to orders.

Regional tests. Radio becomes a Geo Experiment: it runs in one region of each matched pair, chosen at random, and the other joins the Holdout Group, compared on total orders over the burst and a cool-down, never at week two. Radio everywhere leaves no comparison. Podcasts reach every region, so they run in planned on and off blocks, read in the holdout regions. The arithmetic below shows the test can see an effect that pays the radio back on first orders, and little less, so smaller effects are reported as unproven.

The mix model. Only a model puts every channel on one scale, with carry-over and diminishing returns, to set next year's split. This burst's regional split and on and off blocks give it variation and a calibration point.

The brand cue test. It starts after the radio cool-down, each ad area paired only with one of the same radio status, so any radio legacy lands on both. One area per pair, chosen at random, shows the stripe and logo in the first frames; the other keeps the end frame; both versions are built fresh and nothing else differs. Primary metric: orders per area from Fremquaskel's own records over the test and a follow-on window. Guardrails: cost per result by group, branded search by area from brand campaigns and site analytics, and frequency per person. The rule is the experiment's: adopt early cues if orders rise reliably with branded search, keep light cues if only cost rises, stop testing if nothing moves, rerun if frequency differs. Both results enter the quarterly brand health report.

Illustrative example — not a real result

Fremquaskel Linen: can the radio test see what matters?

  • Read period: 8 weeks of radio plus a 4-week cool-down, 12 weeks.
  • Expected orders in the 4 radio regions without radio: 1,300 a week × 12 = 15,600.
  • Extra orders needed to pay back £36,000 at £46 each: about 783, or 5.0% of 15,600.
  • Matched pairs have stayed within about 3% of each other, so a 5% lift should stand out; a 2% lift would not.
  • Brand cue test: 8 pairs of areas, matched on 26 weeks of history, 8 weeks of test and a 6-week follow-on window.

Common mistakes

  • One generous code for everything. It ranks nothing and measures the discount.
  • Judging podcasts on show-note clicks. Listeners search instead.
  • Radio in every region. The clean comparison is gone before the first spot.
  • A verdict at week two. Late buyers carry much of the effect.
  • Cue pairs straddling radio and non-radio regions. Radio then shows up as a cue effect.
  • Reading the cue test's branded search nationally. Only a reading by area belongs to an arm.

Pages this lesson applies

Terms used on this page

Also related

Related