Troy Janisch: Social media benchmarking: Beyond sentiment and share of voice — Live from the Brands-Only Summit

Coverage of this session by Kristen Platt of Connect with her by following her on Twitter.

2:45 —'s Kurt Vanderah introduces U.S. Bank's Troy Janisch.

2:46 — Troy: Data can be eye-opening… but it can also be a snoozer (especially for executives). I hope to provide you with three three actionable insights: To find them, to gather them, and to bind them. Or, just one to rule them all (which I'm about to explain).

2:48 — Troy: The evolution of social insights: As you've become more data-informed, you can create more actionable reports. However, those aren't always helpful because they are report of what happened last month, instead of what's happening now.

2:50 — The goal is to be a “data-driven” company, as opposed to “data-informed.”

2:51 — Troy: The problems we have with social insights: strategy, tactics, and creative. Statistics problems are also present: not enough data, bad experiment design, video, and correlation vs cause. The biggest data problem: When you're wrong, you don't know it.

2:53 — Troy: Social data is all about estimating proportions. As long as you look at it at a high-level, you'll have better results. Accept raw data: it's easier and more affordable. You can always adjust the data once you have it.

2:54 — Troy explains that there is danger in the hidden third factor: lurking variables. The tendency for items that are correlated (best time to post on Facebook) to appear as causes (quality of content). You have to take into account the unpredictable nature of the users.

2:56 — Troy: Sentiment doesn't work without a net. Using Net Sentiment: Calculate sentiment on a scale of -5. to +5, excluding mentions with neutral sentiment. This improves “readability” and comparability over time. Some calculations:

  • difference = positive – negative
  • sum = positive + negative
  • Therefore, net sentiment = (difference / sum)*5 (without the *5, you get a raw percentage, but the *5 is more clear to present to your executives)

2:59 — Troy: This is the social listening pyramid that you are automatically participating in when you're using social listening tools. It goes from being data aware, to data informed, to finally being data driven.

3:00 — Troy: You can optimize your listening by auditing considerations. Auditing = accuracy. You need to audit (verify and adjust) sentiment mentions for your brand.

3:01 — Troy shares some examples of how not all mentions are equal. You need to eliminate clutter, create meaningful categories, and add other dimensions.

3:02 — Troy: Social weight varies by network. Facebook is semi-private, vs. Twitter that is very public.

3:04 — Troy: Create various campaign scorecards so that you can compare different metrics. They also have a space to write the objectives of the campaign and to highlight the best metrics.

3:04 — Troy: Make business decisions with social data. Social data is business data. If you ONLY use social data to measure social performance YOU ARE MISSING THE MOST POWERFUL INSIGHTS. Social Data is an inventory of conversations of every conceivable topic, product, or idea. Inventory varies. The more specific the topic is, the more further backwards in time you’ll need to look.

3:04 — Troy ends by sharing advantages to this approach:

  • Meaningful brand metrics for all social campaigns and content.
  • Provides benchmarking against key competitors
  • Tactic can be adapted to leverage other target markets
  • Tactics can be expanded for all digital marketing tactics
  • Limited/no additional funds required to implement.


Q: How would you recommend reaching out to executives since they're not familiar with the data sets we're measuring against?

A: Troy: Sure, we've tried to looking at the number of touches someone has had with our brand and when they get to the purchase point.

Q: How do you automate benchmarking with share of voice?

A: We average 10,000 mentions a month, so every mention gets audited by someone on our team, (in addition to our sentiment rules), and then we sort the data sets. When we benchmark, we undo all our rules so that we're looking at apples to apples.

Q: What do you do with the data that falls into different categories based on the rules you've set?

A: Yes, it's more work, but a lot of our charts have a ton of different data points and since it's not ideal for quick insights, we pull the insights out of these charts and that's what we show to our executives.

Q: Does net sentiment account for normalizing the data?

A: Net sentiment does help that: think of it as a 7-layer salad: This is our entire influence for the month, but this layer was from our influencer program last week, or this layer was created after our promotional campaign two weeks ago. It really allows you to look for what is driving the overall number.