Here are some things I’ve read in the past week:
- Facebook is withering away.
- Every company should have paid ads on Facebook.
- Link posts are most effective on Facebook.
- Text posts from businesses are not being shown any more at Facebook.
- 73% of Americans 12-17 are on Facebook.
- Teens no longer use Facebook.
Clearly, these things can’t all be true.
But the existence of all these varying claims demonstrates one of the problems with data-driven decisions: all data are not created equal.
Before you make decisions on the basis of information, consider these points:
- Who says? Pew Research says that 73% of Americans 12-17 have Facebook accounts. Pew is a reputable research firm that conducts research on a large scale on a regular basis and has done so for years. I trust their methodology and they have no axe to grind. I read that teens don’t use Facebook at Mashable, which is a good place for tech/trend news, but the author was a middle schooler who says she and her friends are kind of over Facebook even though they love it. This brings up not just the question of reliability of sources, but also the next point…
- What’s the sample size? Pew asked 1,445 internet users in phone interviews conducted in English and Spanish. The middle school author chatted with her friends. Gideon read a blog post here about Google’s importance in local search and questioned my acceptance of a sample size of 2,504. That, he said, is a tiny fraction of the entire population of the U.S. He’s right. However, the larger the sample, the more accurate the data is likely to be, given good research methodology. I’m comfortable with data based on thousands of responses, but not with data based on a handful of responses.
- What’s the chance of bias? If Facebook says that everyone should pay them for ads, I have to assume that there is some benefit for them in convincing businesses that this is true, so there is a strong possibility of bias. I should therefore find that claim less convincing if it comes from them.
- How close is the population to my target? Do you sell shoes to women? No? Then information on how Facebook works to sell shoes to women may not be relevant for you. For example, if you sell biowaste removal systems to hospitals, information on how shoe companies use Facebook to sell shoes to women is pretty much completely irrelevant to you. The data you collect about your own clients is the most useful and relevant (assuming your sample size is large enough and your methodology is good), then the data on people in your target market doing things similar to what you want them to do.
If you’re selling to hospitals, the number of 12 to 17 year olds using Facebook doesn’t matter to your marketing efforts. The information on link vs. text vs. graphic posts on Facebook doesn’t matter to your marketing efforts, either, unless it is based on good data from a large sample size of people who buy things for hospitals.
Check the data before you use it. Make sure that it is both valid and relevant to your populations. Then you’ll actually get the benefits of data-driven decision-making.