Tweet and Meet

Avatar

Midwest Minnesota Twitter Meetup

Measuring Online Conversation – Social Media ROI

I’m going to be honest.  I haven’t completely cracked this nut.  My gut tells me I’m going down the correct path, but I don’t have all the answers.  I’ll give you what I have thus far, and a glimpse of where I’m going with this problem.

My motivation for answering the Social Media ROI equation, stems from our clients’ needs.  They want proof, and I’m going to do my best to give them that proof.

In the most simple terms, if you believe that positive word of mouth generates an increases sales, then increasing word of mouth online should logically have a similar effect.  This is my starting point.  It sounds simple, but it has become a lot more complex, as I started peeling back the layers.

My next step was to look for correlations of month to month sales data, with increases or decreases of online mentions of brand x.  We had some initiatives to spur an increase in online conversation of brand x.  You could see the increase, it was very apparent, and there was a lag effect from the inception of the campaign.  It generated initial momentem for the span of about 1 month, then it hit a plateau, but sustained for about 3 additional months, then settled down to a day-to-day volume that was still 100% increase from before the initiative, where it still sits today.

When looking at sales data, from whatever source, you have to take into account numerous data points which affect this overall number.  Distribution changes, mechandising changes, TV, Print, Radio, FSI, Consumer Confidence, the wind blowing a butterfly in Africa, but seriously, there’s a lot that effects that number.  This is where a good analyst can sift through those data points, and give you a clearer picture of influence.  This is why it’s much easier for small business, who don’t always put money into those areas, see a much more dramatic increase in sales from an increase in positive online coversations.

So what does something like that look like?  Well, Mark Addicks – CMO of General Mills, showed the audience at Blogwell exactly what it looks like (via @josephrueter).  It maps online conversations/mentions of Fiber One to sales data.  According to General Mills, online conversation was the second leading driver of sales, while distribution was number one.

And I rest my case.

London School of Economics study finding an increase of online share of voice of 7% increased a business growth by 1%.

  • taulpaul

    I realize I didn't include cost saving measures into this post, as this is sometimes much more difficult to quantify, but I do believe it should in some way be included.

  • http://metricsman.wordpress.com/ Don Bartholomew

    Hi Paul,
    Good post, Paul. This kind of correlation model is challenging for the reasons you mentioned (e.g. isolating SM impact from all other ways it could happen) as well as factors like seasonality. Previous efforts have shown that correlations will improve if you factor the online conversation data to account for sentiment (you don't want to count negative mentions do you?) as well as competitive activity (at minimum). So share of positive discussion is preferable to just conversation volume for starters. Another challenge with correlations is that you can achieve a relatively high correlation but have a relatively low confidence level due to the large volume of data required (one to two years worth is not too much). One model you might consider is correlating online conversations with something like net promoter index or purchase consideration, and then correlating this with sales. This two-stage model should yield tighter correlations and more diagnostic capability. -Don B @donbart

  • taulpaul

    Thanks Don,

    I appreciate the insight and feedback. We do take sentiment into consideration. The brands I have worked with, to date, have seen moderate increased in overall positive sentiment (i.e. 5% increase in the span of 3 months). It was also interesting to see a competitor of this brand drop 15% in positive sentiment, and watched sales decline sharply, a month later. Have you seen any examples of using this as a regular forecasting tool?

  • http://metricsman.wordpress.com/ Don Bartholomew

    Not yet in social media specifically, but yes in general PR measurement situations I have used and seen others use models for forecasting. One reason to go to the trouble and expense of developing a model is that it may be used as a predictive tool going forward so long as fundamental assumptions within the model do not change. -Don B

blog comments powered by Disqus