Data should be every media buyer’s absolute best friend. Decisions based on rich and meaningful data support continuous improvement and ultimately promote enhanced revenue.
Critical caution: Your data needs to be statistically significant, evidencing the likelihood that a relationship between two or more variables is caused by something other than random chance. For example, if we see a cat cross a particular street on a Monday, we really can’t assign any confidence to this event occurring again the following day. If that same cat crossed the same street every day for a full week, we’d have enough data to confidently assume he’s going to cross again every day of the following week. With enough observation, we’d also be able to collect enough data to learn there’s a catnip factory across the street.
Same goes for your campaign targeting. Observe long enough and you’ll identify solid traffic behavior. With further analysis, you’ll be able to pinpoint specific influencers (catnip for humans) driving the behavior. You can then use those findings to optimize.
How Long to Wait Before Optimizing
Never make major changes or media buys based on just a few hours of performance. There are too many variables that can influence data within such a short timeframe. For example, you could end up cutting a would-be successful campaign because you looked at data from non-peak traffic hours, taking that inaccurate representation of performance as an overall indicator. Always let a campaign run for at least a day to ensure you collect enough data to secure valuable, meaningful data.
There will be exceptions to the rest runtime. Some campaigns are intended to be high impact for a significantly shorter period of time. Keep the total runtime of your campaign in mind when deciding what your initial data-gathering period should be.
It’s understandable that you want to control costs, and cutting ill-performing campaigns is a fast way to protect your budget. But cut too many campaigns too early and you will lose more money than you make in the long run. Give your campaigns enough time to demonstrate whether they’re targeting properly.
How Many Tests and Samples is Significant
Let’s look at banner performance as another example of statistical significance in your media-buying operations. Make sure you have enough samples or tests in play. When integrating banners within your campaigns, employ three to five in the same campaign. If you use just one or two, you won’t be able adequately compare performances. It’s hard to accurately identify performance influencers or relationships with such a limited sample size.
The caveat here is if you employ too many banners. The time and money required to manage and analyze their performance statistics will be too great. Establishing the exact, best number of banners for your specific operations might require some trial and error, so make sure you carefully document your results from each attempt. Always remember to reference your notes and previous trials each time you begin a new campaign. This approach should provide a springboard to help you get an increasingly larger head start with each subsequent campaign.