Can't find what you are looking for?

Email us at info@vertical-leap.uk or call us on 023 9283 0281. Alternatively, fill out the form on our contact us page.

Three ways to apply data science to your social media

Three ways to apply data science to your social media

Categories: Data science, Social media

Every day Instagram users publish almost 50,000 photos. Twitter users tweet 473,400 updates and 4.3 million users head to YouTube to watch a video. Some of those numbers and actions will matter for your brand. Others will have no impact at all.

Identifying important social media cues, trends and signals now requires extensive analytics capabilities. So how do you stop the incoming data flood and transform it into a steady stream of distilled insights?  By applying science to the most pressing problems at hand.

1. Use cluster analysis to improve micro-targeting

Social media networks allow you to reach everyone and anyone, wherever they are in the world. That’s a good thing if you want to have a lot of followers. But as a brand you are not necessarily interested in getting likes from Sydney-based Katie or Vincent from France. They may be great people, but hardly relevant prospects for a local travel agency in Leeds.

Social networks present us with a new challenge – how do I find my ‘tribe’ of customers and connect with them?

These two questions are extremely important if you are staging a new product launch and plan to test the waters with different marketing collateral. You need to understand how your target demographics will engage with your offers. Let’s see how data science can address this.

First of all, you can deploy algorithms to help you identify the most commonly discussed topics on social media in your niche. You can match the popularity of certain topics (e.g. crafts, food or beauty) to a specific platform (Pinterest, Facebook, Instagram). This way you find where your target audience hangs out. Next, you can classify the sentiment expressed by users in different conversations around your brand, or your competition. Are they fans of your other products? Did they sign up with your competitor?  What do they care about in general when it comes to ___?

At this point, you should have enough data to understand who you should target. But that’s not all – you can now multiply the number of likely customers by applying cluster analysis. Cluster analysis enables you to group users in specific communities:

  • Single girls, aged 19-25, still in college, interested in affordable hair care products and beauty hacks
  • Married women, aged 25-35, with an average income of £35K and above, interested in ‘best’ hair care products, supplements and salon treatments

What’s more intriguing though, is that an advanced K-Means clustering algorithm can help you establish even closer proximity between such user groups. You can estimate how frequently they access certain websites e.g. YouTube, Instagram or a certain store and match their preferences with additional demographics data. Your customer profile will then look as follows:

“20-25 year old Instagram fans, who access the platform three to six times per day, posting about #haircolour #hairstyles #Loreal. Expressed negative opinions about Dove. Has neutral opinion about supplement brands, hair vitamins. Often posts about health, beauty, family and travel planning.” Now that’s a customer profile you can work with!

Taco Bell used a similar approach when launching their mobile ordering app in the US.

Taco Bell mobile ordering app

To create a massive marketing push, the company used NetBase Audience 3D platform to identify 3.5 million people expressing positive thoughts about them on social media over the last three years. This audience was further broken down into 90 unique micro-segments, based on exactly what people said that they love, want, need, crave and eat at Taco Bell. Using this emotional and behavioural data, the company launched a series of micro-targeted ads amassing some impressive results:

  • The app was downloaded 3.7 million times shortly after the launch campaign
  • Taco Bell extended their reach 4X
  • Achieved a 2.5X higher retweet rate than other Twitter audiences

2. Learn what your customers care about (without explicitly asking them)

67% of the buyer’s decision is complete before they even reach the sales stage. Social media conversations (public and private) have a massive impact on what ends up in our baskets. Nearly 84% of millennials admit that user-generated content from strangers has some impact on whether they will buy a product. Pinterest inspires 72% of users to shop regardless of whether they are looking for something new or not.

Data science allows you to ‘hear’ what your audience is paying attention to when making their choice without presenting them with the dreaded customer satisfaction survey. For example, Crimson Hexagon used algorithms to identify what car shoppers talk about the most on social media:

Discussed topic analysis

Clearly, a lot of Toyota current and wannabe owners worry about fuel efficiency. Knowing this nugget, you can adjust your slogans and creative copy on social media to address just that.

NOTE: Similar social data visualisations can be a powerful standalone tool for storytelling  

3. Adopt data-driven influencer marketing

Finding the right match to promote your brand can be tough in the era when anyone with an Instagram profile can claim to be an influencer. Though 92% of marketers have found influencer marketing to be effective in 2017, most also admit that their practices of vetting the right candidates were ineffective, driven mainly by manual research and cursory assessment of the key engagement metrics.

Influential, a startup leveraging IBM’s Watson AI tech stack, plans to streamline this matchmaking process. First, the company uses data science to identify the brand’s audience, profile and personality based on their social media presence. Next, the algorithms run a similar assessment of the influencers listed in their database. They pull the last 22,000 words produced by an influencer and then analyse them based on 47 psychographic traits. The goal is to predict the likelihood of having successful ongoing partnerships between a particular brand and an influencer based on the contextual and personality overlap, along with past campaign results generated by a certain influencer.

You can employ data science and cluster analysis at a smaller scale to identify opinion leaders and influencers within your targeted communities and approach them with partnership deals. Your influencer marketing campaigns should not be based on vanity metrics alone; they can be backed up with solid data.

Data science helps connect the dots and capture the opinions and sentiments that matter in a scattered social media environment. It’s time to choose the best course of action based on data, not assumptions.

More about data science in marketing

How to use data science to create interesting brand stories

5 signs you should invest in data science

Expert interview: The importance of data science for today’s marketer

 

George is an SEO Specialist and Data Scientist in the Portsmouth office. He has worked in E-Commerce and Digital Marketing across many industries, for and with companies all over Europe. Before joining Vertical Leap, George worked as Marketing Director for his own company, for which he managed to expand the company’s activities to 5 European countries. George has been creating websites for more than 10 years and he has in depth experience in designing and bringing optimised E-Commerce websites to market.

More articles from
Back to top