Make a fair comparison.
Start with the same artwork, copy, offer and destination. Create an SDR control and an HDR variation. Use the platform’s experiment tools where available to reduce audience overlap and uneven delivery.
Keep audience, placement, timing and bidding conditions as consistent as the experiment allows. If those change too, you will have a harder time knowing what caused the result.
Check that you are testing HDR.
Upload the original export and inspect the final placement. A platform may alter the image or show different versions in different contexts. If both variants end up as ordinary SDR images, the test cannot tell you much about HDR.
Your audience will also contain different devices and rendering paths. A campaign-level result measures that real mix, not an effect seen equally by every viewer.
Decide what success means.
Choose a primary outcome before starting: a qualified visit, signup or sale, for example. Click-through rate can help explain attention, but it should not stand in for the outcome your campaign needs.
Set a practical test duration and evaluation method before inspecting the winner. Small samples and repeated early checks can make noise look persuasive. If the evidence is weak, call the result inconclusive.
Keep the claim as small as the evidence.
A result from one creative, audience and placement does not establish that HDR always performs better. Record the conditions and repeat the experiment before generalizing.
Treat third-party success stories as ideas for experiments. Without the test conditions, audience mix and underlying results, a reported improvement cannot tell you what to expect from your own campaign.