Case Study
Turning a “launch & learn” test into a repeatable performance-TV engine.
Early in 2025 we set out with Finanzguru to prove that linear spots can be tightly measured, ruthlessly optimised and fast to improve. Two back-to-back flights —an initial December-to-January run and a February follow-up— gave us an ideal testing ground – essentially a controlled benchmark environment where every variable (channel, day-part, creative length) could be isolated and judged on hard performance data.
Here’s the story of what we tweaked, why we tweaked it, and how those tweaks changed the numbers.
Where we started
Our first flight ran with a healthy six-figure budget and produced a respectable volume of installs and registrations. Acquisition costs were still high, largely because the average cost-per-response topped €2.60 and CPI remained in triple digits.
That opening sprint was never meant to be perfect—it was our data-gathering lap.
What changed in February
Armed with four weeks of granular log-level data, we re-deployed for 1–28 February. Spend came down, but output didn’t: installs doubled and registrations doubled again.
Most telling were the economics:
- Cost-per-response fell by ~70 % (from €2-plus to just €0.80)
- Cost-per-install dropped about two-thirds (down to ~€50)
- Cost-per-registration was cut by more than half
Same channels, same creative assets—radically different economics.
Four levers that moved the needle
- Temporal micro-bidding
- Mondays came out cheapest (≈ €0.59 CpR) while Sundays proved most expensive.
- The 06:00-09:00 slot beat every other day-part on response and install metrics, whereas late-night 23:00-01:00 was the money pit.
That insight let us shift impressions into low-noise, high-intent hours without paying prime-time premiums.
- Channel-stack efficiency: Nineteen channels entered the plan; three—CrimeIInvestigation, Universal TV, and History Channel—did the heavy lifting, each delivering sub-€0.30 CpR. TLC and WELT lagged badly and were throttled back. Concentrating spend where the data said “go” took friction out of every metric further down-funnel.
- Creative compression: We leaned into a 70 / 30 rotation favouring ten-second cut-downs. Shorter spots didn’t just save media cost; they consistently beat the 20-second versions on CpR, CPI and registration efficiency.
- Week-by-week pivoting: Our analysis showed CW 6 (early February) as a sweet-spot—CpR bottomed at €0.33—while CW 5 spiked hard and triggered an immediate bid pull-back. This real-time throttle kept ineffective weeks from poisoning the month’s average.
How the workflow looks now
- Plan light, measure heavy. We go in with a broad media footprint, but every line-item is on a short leash.
- Automate date-level alerts. Any day or channel drifting 2× above the flight average lights up a dashboard signal.
- Treat creative like a variable, not a fixture. If the ten-second edit wins, it keeps the weight; the 20-second gets benched until it can justify its cost.
- Document everything. Each decision—bid change, channel pause, creative swap—feeds a living playbook for the next flight.
What’s next
The February flight proved that LTV can scale when every input is judged on bottom-funnel value. Our next sprint will test a few fresh channels with high target-group affinity while locking in the 06-13 time-block and the proven 70 / 30 creative mix.
Take-home thoughts for fellow planners
- Hour of day often beats headline reach. Prime-time looks glamorous; 06:00-09:00 prints conversions. Read our latest take on which time people are watching TV.
- Channel lists should stay liquid. Cutting laggards funds more volume where it matters.
- Short-form isn’t just cheaper—it persuades faster.
If you’re wrestling with similar trade-offs—budget versus depth, reach versus registrations—let’s talk. Continuous knowledge-sharing keeps the whole discipline sharper, and maybe GLADTOBE has something to offer.
