February 26, 2012

The ROI of Social Media

Social media ROI has been the elephant in the room for many years now, as marketers and analysts struggle to address it and calculate it, even though we know how important it is to drive corporate social media investment forward.

Fuseware has developed some great algorithms for detecting how many eyeballs have seen social media conversations related to a brand, topic or event. For example, the recent Deloitte fun run in Pretoria garnered a social media response of close to 20,000 views from the mentions of only 50 people. In social media, even a few influential people can have a tremendous market impact when they speak about the same topic.

The question now is, how does the number of views translate into hard and fast ROI that a marketing director can present to justify the social media investment. I have been asked this question several times by my clients, and would like to give an official response.

The advertising industry has a metric called the “advertising equivalent value”, which calculates an approximate ROI based on the number of views and demographics of a piece of content. Many companies take the easy route and use this metric in social media, but I don’t think it quite fits due to the vast differences in conversations and demographics. Firstly, I don’t think social media ROI can easily be automatically calculated. The value from 20000 people talking about a fun run is different to the value of 20000 people talking about a 20% discount sale by a large retailer. The ROI needs to be determined strictly from the market’s conversion potential, that is the rate at which people who view the content end up converting to what the company may be offering, and the downstream revenue that they eventually may give to the company. For example, 20000 people who convert at 1% for a R100 offer are more valuable than 20000 people who convert at 1% for a R10 offer – so even if the social media impact is the same, the ROI would be different.

The first metric to find is the probability that a person on social media will convert, if they have been exposed to your social media content. This then boils down to a problem of statistics, for which we can use Bayes Theorem to solve.

Let’s define A as a conversion. Essentially, this is the desired action you want your customer to take.
Let’s define B as a social media user, who has been exposed to your brand related content.

P(A|B) is the probability of converting a customer, if they have seen your social media content. The variables we need to solve are as follows:

P(B|A), which is the probability of a social media user having seen your social media content, once they have converted. This can be found by surveying which of your converted customers have done so because of your specific social media efforts. If you have an online store, you can build the metric into your e-commerce system by tracking users through cookies as they click through your social media links onto your site.
P(A), which is the probability of a conversion. This is taken from the company’s existing conversion rate data that exists for its past advertising campaigns, and is independent of social media related conversions. This can also be taken from existing advertising initiatives, such as PPC campaigns.
P(B), which is the probability of a social media user seeing your content on social media, whether or not they have been converted by your current offer. This is the tricky part, one which a company like Fuseware can help you solve. It requires calculating how many users are in the brand’s market in social media, and how many have been exposed to specific brand related content.

An important note to make is that the more accurate you make P(B), the more accurate the final ROI calculation will be. For example, if you are monitoring  your entire brand’s mentions as social media content, you will find that there are a combination of positive and negative mentions, as well as many mentions unrelated to your conversion – many of this will be irrelevant to the ROI calculation of your campaign. This is why one needs to be able to track a campaign using a specific hashtag on Twitter, so the exact exposure of that hashtag can be determined, and the content involved with the hashtag is guaranteed to be campaign-related.

With a probability value calculated, and the number of social media content views determined via analytics, it is now the company’s task to put a value on a single conversion. A conversion could be a product sale, in which the value of a conversion is the revenue from that product, or it could be more intangible, such as getting the person to sign up for a mailing list, which may have a downstream value of R100 per year, on average. In any case, this will need to be defined by the company and its unique conversion targets, for which they should already have the value.

The final ROI then becomes fairly trivial:

The investment gain is calculated from P(A|B), multiplied by the number of views on social media for the specific campaign, multiplied by the value of each conversion. The cost of the investment is calculated from the expenditure on social media by the company, including time billed, agency rates and the cost of producing social media related content such as viral images and media.

Some points to note

  1. Engagement has an indirect effect on ROI and is very important, but the social media exposure metric indirectly absorbs it. More engagement also drives intangible brand loyalty, which has not been considered in our equation.
  2. The quality of content related to calculating P(B) is critical to the calculation. 5000 views of “i love vodacom” will have a vastly different response to 5000 views of “i hate vodacom”, even if both have the same exposure. However, due to the calculation of P(B|A), there is a direct relationship between exposure to this content and converting a customer.
  3. In the end, viral laser-focused brand campaign content is what drives the ROI equation.
  4. This equation calculates ROI from a purely monetary perspective. Intangible things like reputation, PR value, trust building, and market differentiation are not taken into account, which of course should be.

I would love to hear your opinion of this in the comments below. If you would like more information or elaboration, please contact me at mike [at] fuseware [dot] net.

Be Sociable, Share!

One Comment

  1. Jeff December 28, 2012 at 6:06 am

    follow me on twitter @bayestheoremqed.


Leave a Comment