By Ryan Richardson · Published 8 October 2026
This ran as one coordinated test on a $19-$47 digital guide sold to cold Meta traffic: 12 copy variants, 3 design systems, 3 price points and 6 CTA wordings, allocated in real time by a Thompson-sampling bandit, plus 5 entirely different page architectures tested in parallel against a multi-section control. Three full rounds of iteration happened inside one week, including a complete CRO rebuild between round 1 and round 2 that re-scored the page from 4.6-7.1 out of 10 up to 8.1-8.6 out of 10 across the review's own rubrics.
91% of visitors scrolled past the halfway point of the page. Meta ad click-through rate ran 7.95-10.00% across every pattern and price tested, against a 5% industry benchmark. Click-to-landing-page retention (the share of ad clicks that actually turned into a tracked page view) ran 50-81% depending on pattern and price. None of these numbers are marginal; they're the signature of an ad and a page that are doing their job of holding attention.
Against roughly 800 landing page views and roughly 600 link clicks, the CTA (the click that leads toward checkout) was clicked about 4 times, for a CTA click rate of about 0.7%. Zero of those reached a completed checkout. A healthy direct-response page at this price range should see a 5-15% CTA click rate and a 1-3% purchase rate on cold paid traffic; this ran at roughly a tenth of the low end of that CTA range, and nowhere close to the purchase range.
Across 282 tracked visits spread over 12 copy variants x 3 designs x 3 prices x 6 CTA wordings, the probability-of-best-variant scores stayed close to uniform in every dimension: 0.06-0.12 across copy variants, 0.32-0.35 across designs, 0.24-0.43 across prices. The bandit math was running correctly; it simply had no purchases to use as a learning signal, so every arm looked statistically identical to it. A bandit can't differentiate winners from losers if nothing downstream ever counts as a win.
Three separate levers were already ruled out by the time this test ran: tracking (the pipeline was verified end to end), page design (three CRO rounds and a full rebuild, no change), and price inside the $19-$47 band (3 prices, no differentiation). That leaves offer shape, audience and channel as the remaining candidates, and offer shape was judged the most likely: a flat, single-item digital guide at this price doesn't give cold Meta traffic a reason to act immediately, regardless of how the page around it is built.
| Claim | Value | Source |
|---|---|---|
| Total ad spend across the week | ~$200-250 AUD | Sixty Steps bandit test logs |
| Total landing page views | 800+ | Sixty Steps bandit test logs |
| Scroll past halfway | 91% | Sixty Steps bandit test logs |
| Meta ad CTR range across patterns/prices | 7.95%-10.00% | Sixty Steps bandit test logs |
| Click-to-landing-page retention range | 50%-81% | Sixty Steps bandit test logs |
| Total link clicks | ~600+ | Sixty Steps bandit test logs |
| CTA clicks | ~4 | Sixty Steps bandit test logs |
| CTA click rate | ~0.7% | Sixty Steps bandit test logs |
| Purchases | 0 | Sixty Steps bandit test logs |
| Page CRO re-audit scores after rebuild | 8.4 / 8.1 / 8.6 out of 10 | Sixty Steps bandit test logs |
| Page CRO scores before rebuild | 4.6-7.1 out of 10 | Sixty Steps bandit test logs |
| Tracked bandit visits across the 12x3x3x6 combination test | 282 | Sixty Steps bandit test logs |
| P(best) range across copy variants | 0.06-0.12 | Sixty Steps bandit test logs |
| P(best) range across designs | 0.32-0.35 | Sixty Steps bandit test logs |
| P(best) range across prices | 0.24-0.43 | Sixty Steps bandit test logs |
| Benchmark CTA click rate for a healthy DR page in this price range | 5-15% | Sixty Steps bandit test logs |
| Benchmark purchase rate for cold paid traffic in this price range | 1-3% | Sixty Steps bandit test logs |