Social Media is Big Data Panel Notes
Today I represented Three Deep as a panelist for a discussion about social media at the #bigdatamn conference hosted by Medtronic. The panel was entitled Social Media IS Big Data and was moderated by the great Ward Tongen. The experience was positive and we had a good turn out for our panel, even though it was near the end of the day. The audience asked many great questions, and the panel provided solid answers. In fact, so many good questions were asked that we didn't get a chance to delve on any of the pre-formulated questions that came up.
Not wanting these notes to go to waste, I thought it would make a good blog post. Do you find these to be helpful?
What's important to do to get actionable data?
- Collect it, understand it, spot check it, analyze it and present it. Repeat.
- Most stakeholders only pay attention to the presentation layer, so this piece is ultimately the most important focus for end users. However, you must do extensive data collection and analysis in order to be able to present something actionable.
- This is where most programs fall short. They take aggregates and averages and use them as a report claiming insight. This is not insight, but rather reporting. In order to do a true analysis, it is important to dig several levels deep into the data to tell a story.
- The biggest problem we run into is that it takes 10 times longer to get to a groundbreaking level of insight and we don't always have time to get it done. This is where we can use software to make us more efficient.
How do you take Data and turn it into insights that are usable across the organization?
There are many levels of insights that can be drawn from data, and many tools to get at these insights. A mix of things like word clouds, pattern analysis, n-grams and hard excerpts of responses can be used to find macro and micro insights, and these will often be the difference between a report that goes into the recycling bin and one that turns into action.
The other thing to consider is the amount of time needed to take action. The bigger the company, the more difficult this is to achieve. The key is to surface insights early and often in the process to gain short term support and allies while waiting on actions to be taken.
It might even be worthwhile to call out short term, mid term and long term goals and milestones that are needed for the overall success of the program and surface them in the dashboards and reporting that we create.
What philosophies and strategies do you have for tackling the problem?
It's all about finding a needle in a haystack without going insane. At a high level, this involves manual data review to understand the landscape of information you have available. This is a daunting, but necessary proposition that we need to understand before trying to bite off a big set of data.
We recommend digging deep into the data to find true insights that are valuable to the organization, and then determining a strategy to bring this level of insight to the entire business. Conducting this exercise is not meant to be a replacement for the tools of the trade, but rather a method of fine tuning the tools to get at what you really need.
Tools are often developed with a one size fits all approach, and are rarely configured for specific industries, and almost never configured for a specific business. Teaching the tool what is important to you will lead to long term sustainability and time savings in the long run.
How do you filter the good stuff from the noise?
The key to filtering is to not trust automatic filters. Start the filtration process by spot checking the data that comes in to understand insights. This could involve manually reviewing many of the entries collected to better understand the landscape. In addition, try to measure the influence of the source of the comment to understand if they have an audience.
This information will lead to a strong understanding of the information available and how it might need to be filtered for the future. No filters are perfect, especially the automatic ones provided by a tool. In fact, these automatic filters often provide a false sense of security for those using the reports. Filter manually on a subset of data and you will be much more insightful.
Are dashboards the best way to get my stakeholders to visualize what the data is saying?
Dashboards can be tricky, because they start off being highly interesting to stakeholders, but they quickly lose interest if they regurgitate the same numbers over time. What we started doing was adding a mixture of data, charts and insights to our dashboards to appeal to multiple learning types. We also do long term comparisons (year over year, month to month and 13 month rolling).
Another important thing to do is add insights to the body of the email that goes out with the dashboard. In addition, you may decide to change up the delivery method, printing out dashboards for those in the same office. It's hard to ignore paper.
What are the most common business use cases for social data?
- Customer service
- Customer profiling
- Product development
- Proactive communication
- Sales and lead generation
- List development
- Relationship development
For companies that have social listening programs, what is the most typical area for improvement?
Taking action in a rapid and impactful way. There is often a problem with social data being regarded as too small to take action, and companies are often not equipped to respond ublicly to inquiries due to legal and regulatory issues.
Up until the social revolution, it was much easier for companies to control the flow of the conversation. This seems to be the biggest hurdle to overcome, because it's unlike anything they have had to address in the past at such an enormous and open ended scale. Companies can improve their status by rethinking the value provided by social data and investing in the process, tools and resources to maximize the value of social media.