How do you know that? Testing.
If you are marketing automobiles through personalized direct mailers, for example, it may not matter whether the marketing image shows a man or a woman behind the wheel. Time and energy spent swapping out such images might have little or no impact on results.
More useful might be pairing an image of a minivan with recipients known to have children or hybrid compacts with recipients in areas known for environmental sensitivity.
It’s about relevance, not necessarily demographic matching.
Test each variable separately.
Always test using within the same program rather than comparing to previous programs. This way, you are doing an apples-to-apples comparison. Otherwise, there could be other variables (timing, economic conditions, seasonal variation, database shift) that could impact the results. Say you change up your messaging and increase the size of your postcards from the standard to oversized and your response rate goes up 8%. What caused the increase? The change-up in messaging or the increase in postcard size? Without testing each variable separately, there is no way to know.
Split the list using a random data generator so there is no unseen influence in the recipient base. This will tell you exactly which elements had what impact on response, as well as how the combination of elements may have amplified the message over any single element alone.
Split testing is a critical aspect of any marketing program. Make it consistent, make it strategic, and make it intentional.