1. The quality of the social media you collect is defined by where you search and the search terms you use – you need to be able to specify online locations and you need to be able to use Boolean search terms. Free and simple tools are great for learning, but will not deliver a commercial outcome.
  2. In most cases a majority of the data you collect will NOT relate to your investigation. Budget plenty of time/resources to clean up the information. For example, if you search for Apple you will want to remove fruit and cider, as well as a large quantity of spam.
  3. Automated coding is not as good as manual coding, manual coding is not perfect, and a wise trade-off is often to start with manual coding and then automate it. Do not trust automated options straight out of the box.
  4. Social media research is great for some things, for example finding out whether people have noticed your new campaign and what people think about it, and poor for other research problems, such as which of four as yet unreleased pack designs will be most successful if launched.
  5. Social Media Research results have relative values, not absolute values, so benchmarks need to be established to help assess if a result is weak, average or strong. The benchmarks should be designed at the start, not created afterwards.
  6. Lots of things you can do with social media data are not ethical, and some are not legal, so be careful. You do not want to be the next PatientsLikeUs and BuzzMetrics outcry. If you have not heard about this case click here to read it.

If you are in North America and would like to learn more about Social Media Research, join one of our workshops in October. Click here to find out more about them and the other No Nonsense workshops.