Social media marketing customers asked recently, “I’d love to know how to look at analytics on social media platforms like Instagram. What SMM analytics tools or other software programs can I look at for more metrics?” The data we’re given usually isn’t sufficient to answer the questions we have, particularly in social media data. And we all know the high value of data these days, it’s the new oil, so to speak.
Often, social selling platforms give us very basic data – take a look, for example, at what comes out of the Twitter API: created_at, screen_name, source, favorite_count, retweet_count, description, url, followers_count, friends_count, listed_count, text, status_id Now, there’s a fair amount of data there, but most of it is descriptive analytics – what happened. A tweet happened – how many likes, retweets, etc. happened? There’s almost nothing in the data to explain why audiences did what they did with that particular tweet. You get a similar amount of social media data from Facebook as you do from Twitter SMM tweets.
We get even less data out of Instagram: Account, User Name, Followers at Posting, Created, Type, Likes, Comments, Views, URL, Link, Photo, Title, Description, Image Text, Sponsor Id, Sponsor Name, Total Interactions So what’s a marketer to do? We have to create the data we need to answer questions. This is a process known as feature engineering. Feature engineering is fancy for taking data we have and creating new data from it.
Here’s a simple example of Instagram data. Suppose you have this timestamp: 2021-02-06-18:00:41-ET What information is encoded in this piece of data? From this we could extract: Year Month Day Day of Year Day of Month Day of Week Hour Minute Second Timezone
Could some of that Instagram data information be useful? Sure thing – everyone wanting to know when the best time to post on Instagram would find that information embedded in a timestamp. We’d just need to extract it from the data. Here’s another Insta example. “It was a yawn but it sure looks like he’s yelling that he’s hungry. #dogsofinstagram” This is a caption on an Instagram picture I put up earlier. What could you derive from this piece of IG data?
You will also end up getting less social media data from platforms like Snapchat, TikTok, Clubhouse, Hive and a medium amount of analytics information from LinkedIn, Pinterest, Quora, Reddit, and YouTube.
Length of text in characters is also key. Length of text in words Number of hashtags Handles mentioned (if any) Grade level of writing Nouns Verbs Other parts of speech Sentiment and tone Topics and keywords When we do feature engineering, we take what we have and squeeze every last bit of value out of our data so that we can use advanced statistics, data science, and AI to understand better what happened and perhaps to start teasing out why. For example, I’m working on a slide deck about the top 10% of Instagram brand posts out of 2 million posts, and one of the things I’m looking at are the hashtags unique to the top posts, hashtags that don’t occur in other posts. That little tidbit might be helpful to explain why some posts do better than others with business social media marketing.
Some feature engineering does require advanced technology, but a lot of it can be done right inside a spreadsheet. When you’re struggling to answer complex questions about your analytics, the reason you might be struggling is that you haven’t extracted all the possible data from what you have on hand. Dig into the power of feature engineering and see what treasures await you in your data. It’s all crucial to marketing your business correct socially, tracking key social media analytics, and improving your overall business reputation on social media.