At TweetReach, we’re often asked about how to measure share of voice (SOV). Measuring share of voice involves comparing one brand’s metrics to the total conversation about that brand’s category. Historically, this has been a difficult exercise because high quality data is hard to come by. But Twitter is an abundant and accessible source of real conversational data, allowing us to easily track mentions across a variety of brands. You can now determine the size of conversation for an entire category and compare your own brand to the overall conversation.
You can measure share of voice for any set of similar topics - competitive brands, products, companies or people. You can even compare share of voice across political candidates. Political candidates are the perfect example for a share of voice comparison. There are usually several people in the race, with a few frontrunners and a few hangers-on, just like most any product or business category. And people are talking about them on Twitter, providing a remarkable dataset for analysis.
Earlier this year, we tracked tweets about the U.S. Republican presidential candidates (see our interactive visualization and analysis). Now that Mitt Romney has emerged as the presumptive GOP nominee, we’re tracking the candidates for the vice presidential slot on the Republican ticket. VP candidates are not elected separately, but we can still use Twitter to gauge popular opinion and awareness on these candidates. Plus, they make a great example for a blog post about share of voice.
So, here are four steps to using Twitter data to measure share of voice.
1. Decide who you want to compare.
Before you start measuring, you’ll need to determine which competitors to compare to your own brand. What are the brands that make up the category you’re interested in measuring? Pick two to ten to compare. It’s probably easy to pick out your one or two most direct competitors, but also consider other less obvious choices you should add, as well as any large brands that make up your category. It’s possible that what your customers perceive as related might not even be on your radar, so think about this carefully.
In our Republican vice presidential candidate tracking, picking who to track was not that difficult. There are a set of people who have publicly made some indication that they’re interested in the job, and others that analysts and others who pay attention to these kinds of matters think could be chosen. So after a little research, we narrowed our field of possible candidates to 10 people:
- Kelly Ayotte
- Jeb Bush
- Chris Christie
- Bobby Jindal
- Bob McDonnell
- Tim Pawlenty
- Rob Portman
- Marco Rubio
- Paul Ryan
- John Thune
There are probably a few others we could include (or remove), but this is a solid, representative list for our needs. However you choose, pick 2-10 related brands to monitor in addition to your own.
2. Set up appropriate keywords for tracking.
Next, you need to track comparable terms for all brands. Most Twitter measurement tools (TweetReach included) will require a set of queries or keywords to begin tracking tweets. In this step, your goal is to make sure that your metrics aren’t later impacted by a data quality issue. If you monitor one brand’s Twitter account, then monitor all brands’ Twitter accounts. You probably know all the keywords you’d want to track for your brand, so think as carefully about the others as you did your own. Are you using common misspellings or nicknames? Are there other languages to consider? Multiple official Twitter handles or hashtags?
In this GOP VP case, we’re tracking full names (“Marco Rubio”) and Twitter handles (@marcorubio) for all candidates. We opted not to add last name-only keywords since candidates like Jeb Bush and Paul Ryan have fairly common last names and that would result in more noise than useful data. Since we can’t track their last names, we won’t track any other last names either. You can decide what makes sense given your goals, but just be consistent across all brands.
3. Collect enough data.
The next step is to start collecting data. Some tools do this in real time, and others have historical data you can mine. Either way, collect enough data that it’s representative of the full spectrum of conversation about your brands. Conversations can be spiky over short periods of time, so it’s best if you have weeks or preferably months to balance out those spikes across all brands. A longer data collection period also allows you to notice trends in SOV changes. The more data, the better. The longer you’ve been collecting data, the better.
GOP VP candidates see jumps in Twitter mentions when they’re featured in the news or after a public appearance. Some will just see more total tweets over time. We want to track long enough that we can differentiate between a legitimately higher metric and a one-time spike. In our specific case, we’ve only been tracking these candidates since early May (so just over two weeks) and the data is still pretty immature. Some of the candidates have been added even more recently than that, so their data is newer still. This means we shouldn’t take any of these metrics too seriously yet. But they will improve over time, so when we check in next month, we’ll have a much more representative picture of the true conversation.
4. Compare several metrics.
Finally, it’s important to compare brands across several different metrics to truly understand what’s going on. You may have a favorite stat or a particular KPI you’re targeting, but try to compare a few different metrics before deciding which to use moving forward. One brand might have a high reach, while another could have a lot of tweets. Use several metrics to compare, to see where the patterns are and what metrics make most sense in your industry or category.
Let’s look at a few metrics for the current top three Republican VP candidates (at least according to Twitter): Chris Christie, Marco Rubio, and Paul Ryan. This list will likely change as our data matures, but it’s fine for an early analysis.
One of our favorite metrics to start with is simple tweet volume. Tweet counts are useful in understanding the size of the conversation about a candidate. Below are graphs for both tweets per day and cumulative tweets so far this month for the three candidates.
You can see that Ryan (yellow) is slightly ahead of Christie (blue) in cumulative tweets right now, but both are increasing steadily. Christie has had two large spikes in daily tweet volume, while Ryan has had one. Both of these metrics will stabilize after a few more weeks, and we’ll have a clearer picture of who’s on top. Right now, I’d say Ryan has the slight edge on Christie, but it’s close.
And if we’re actually going to look at share of voice, let’s compare each candidate’s tweet volume to overall tweet total. In the past two weeks, there have been 46K total GOP VP candidate tweets. 35.2% of those mentioned Ryan, with Christie close behind at 32.9%. Track SOV over time, as changes in a brand’s share could indicate important perception shifts. For example, when we started tracking GOP presidential candidates in early January, Ron Paul dominated that conversation’s share of voice, and was mentioned in more than 40% of all tweets. But by April, that share had dropped off almost entirely, leaving the rest to Mitt Romney.
It’s also helpful to look at several metrics side-by-side. In this case, let’s compare reach, tweet volume and number of unique contributors.
Looking across these three metrics, Christie appears to be the frontrunner. His reach is currently more than 15 million, with 10 million for Rubio and 8 million for Ryan. Looking at reach and tweet volume in conjunction with contributors – the number of unique people talking about a candidate on Twitter – it seems like a lot of different people are talking about Christie and Ryan, while Rubio has a smaller group of vocal supporters. To achieve a 50% higher reach when compared to the other candidates, Christie was probably mentioned by a celebrity, typically the only people to have follower counts over a few million. (In this case, it turns out @jimmyfallon, who has 5.5 million followers, tweeted publicly to the governor.)
Reach is an excellent metric for share of voice, because it tells you about the size of the potential audience for a brand. The bigger the reach, the larger the variety of people who are spreading the message. A high reach indicates a diversity in contributors and audience, as well as some potentially influential and high-follower contributors.
We also recommend unique contributors as a share of voice metric. Which brand has more different people talking about it? One caveat about both reach and contributors is that since these are metrics based on counting uniques, you can’t compare one brand’s metric to an overall sum, since you can add up reach or contributor numbers to get overall reach or contributors. You can only compare reach to another brand’s reach. That’s still useful, but may not be a traditional share of voice metric.
Twitter and share of voice
Twitter is a incredibly rich source of share of voice data. If you’re tracking similar brands, products or people and one has an audience on Twitter, it’s likely they all will. Due to the real-time, public and archivable nature of Twitter, we can access this data for all kinds of useful analyses. People can and do talk about their favorite – and least favorite – brands on Twitter. For all these reasons, Twitter is perfect for SOV analysis, if you do it right. Doing share of voice right means selecting the appropriate brands to compare, ensuring consistency in search queries, aiming for long-lived data collection, and embracing diversity in data analysis.