Can Social Influence Be Distilled Into A Score? Part I, The Potential
If you’re planning a campaign to stimulate word of mouth discussion of a product or a brand, one question likely keeps arising — how influential is a particular person online? Is it worth reaching out to them? If you’ve been pondering how to measure or benchmark social media influence, several companies, including Klout, PeerIndex and Kred, […]
If you’re planning a campaign to stimulate word of mouth discussion of a product or a brand, one question likely keeps arising — how influential is a particular person online? Is it worth reaching out to them?
If you’ve been pondering how to measure or benchmark social media influence, several companies, including Klout, PeerIndex and Kred, are clamoring to assist you in identifying and connecting with influencers with high social capital ranking scores.
Even if connecting with influencers isn’t an immediate need, the boss might be asking for explanations of your brand’s Klout score, so now is a great time to get up to speed!
Wait, What Do We Mean By Online Influence?
What is online influence, also called social capital, and can these companies really reduce it to a number? Many readers will remember that it wasn’t so long ago that many would cite their site’s Alexa rank as a badge of honor, blissfully unaware that the score had no scientific basis nor business value.
Can marketers, HR professionals and others rely on these social capital scoring services to identify influencers or are the scores so flawed to make them meaningless?
For the purposes of social media marketing, influence might be defined as:
the capacity to have an effect on the behavior of someone in a way which results, directly or indirectly, in a business outcome
Marketers identify who is influential in their target market and cultivate these influencers to encourage brand advocacy and word of mouth recommendations, recommendations which multiple studies have identified as more credible and effective than traditional advertising messages.
Moreover, under the right conditions, word of mouth advocacy can spread virally through social networks.
Marketers active in social media for some time may well know who the key influencers are in their target market. But even then, its nice to be able to have a third party to help confirm assumptions. Companies like Klout, PeerIndex and Kred say they can do just that.
How Does Online Influence Work?
For a person’s online musings to impact others, several things need to come together.
The person must have an audience, i.e. their messages have the potential to reach others. The potential reach of a message in a social network depends on several factors including how many real “followers” (friends, fans, connections) a person has and when the message is posted. PR professionals will know that sending a press release out on Friday evening is the best way to insure bad news gets buried.
Several social networks, such as Facebook and Google+, prioritize the display of posts for followers based on recent interaction between the follower and the post’s author.
(I mention real followers, as many fake followers exist. Some brands and politicians, suffering from size anxiety, buy fake followers by the kilo. Another category of fake followers are spam bots which automatically follow users, often with an ultimate goal of distributing spammy links to dubious websites.)
As if that wasn’t enough, the posted message must resonate with a recipient. Resonate is itself a nuanced concept.
Is the post on a topic of interest to the recipient? Is the author perceived as an authority on the topic or does she otherwise have credibility due to a personal relationship with the recipient? Are the contents of the post timely? Something the recipient was thinking too? Is the message original? Is the message appreciated for its tone or style (use of humor, sarcasm, irony)?
If the message resonates to any degree, then it will have succeeded in capturing a recipient’s attention, no small feat in a world saturated with messages. What is wonderful about Internet marketing is that we know a message resonates if it provokes a recipient to interact with it. Typical online social interactions include:
- commenting on a post / update (comment, reply)
- endorsing a post / update (like, +1, favorite)
- sharing a post / update (forward, re-tweet)
- clicking on a shared link
- playing a shared video
In the case of messages which originated from a friend of a friend, a user may also choose to follow the original author (circle a person, become their friend, etc.).
Business Outcomes Count (And Can Be Counted)
For businesses and professionals using social media as a marketing tool, engagement should lead to business outcomes.
With my business model, I’m happy when someone publishes a link to an article I’ve written or to my website in general. I’m even happier when someone clicks on one of these links (bringing me more reputation and contacts).
To capture click data, we need to use URL shorteners, Google Analytics or similar tools: influence scoring tools aren’t enough.
How Do Influence Score Systems Work?
Social scoring services may have their place in the planning and execution of a social media word of mouth campaign. But to understand both their strengths and limitations, a prerequisite for their correct use, and to explain a brand’s score to the boss, it is important to know how they work. This is also true for HR professionals who want to consider influence scores in hiring decisions. Fasten your seat belts, the ride is going to be a bit rough.
1. Gather Social Data
The first step a social influence ranking service has to perform is much like that of a search engine: collect raw data from the world wide web. This task is much harder than one might think to do right.
Ideally the service would scour (crawl) all major social networking sites, blogs, forums, question and answer sites, review sites and other sites where people add user generated content (UGC). In reality most services have taken the path of least resistance: they limit their scoring for most profiles to data from a single source, Twitter, as it is already nicely structured and is available from distributors GNIP and DataSift.
Much of Facebook data is off-limits unless a user explicitly grants permission to an influence measurement service. Information from blogs hosted on known blog hosting domains, such as WordPress.com and Blogger, should be rather easy to include; self-hosted blogs, often maintained by more savvy netizens, can be found by blogroll analysis and blog directories.
As the web is dynamic, the data collection process needs to be continuous and needs to identify variations from the last data collection cycle.
2. Extract And Organize Collected Data
Profiles, followers, posts and interactions need to be extracted from collected data. An attempt has to be made to connect a specific person’s profile across different social sites — not an easy task. Google closed their Google Social Graph API, which did just that, citing poor adoption.
3. Apply Scoring Algorithms To Data
Each company has its own method to decide what weight to give social media endorsements, comments, sharing and the like. Should a comment count more, the same or less than a share? Since an endorsement just requires a click, are they weighed less than comments and shares? Should a share by noted blogger Robert Scoble count more than a share by a friend?
The basic elements available to scoring services aren’t complicated but how to combine them to define a user’s social capital isn’t so simple. Klout changed their scoring methodology in October 2011, only to change it again in August 2012.
Influence score providers are generally reluctant to share this information in detail as their scoring algorithm is one of the few differentiators they have from competitors. Too much transparency, they say, would help those actively working to increase their scores, known as gaming the system.
Klout, PeerIndex and Kred each provide methodology overviews, with Kred touting that it is more transparent than the others as Kred shows the points attributed to each element which contributes to the overall score (visible by clicking the Total influence points panel in the story layout). Klout’s introduction of moments in their August 2012 update should close some of the perceived transparency gap.
In describing PeerIndex’s view of transparency, CEO Azeem Azhar uses credit scoring as an analogy. He says that is important to know what the inputs are and the data gathering process yet you don’t necessarily need to know the weight attributed to each input, nor understand the mathematical models used, to feel comfortable with the resultant score.
Kred Story, a visual presentation of the events contributing to a person’s Kred score, is Kred’s latest iteration on demonstrating transparency. Detail on each item which contributes to a score is available with a click, what Kred calls zooming in. I found the presentation to be a bit overwhelming (as did my browser which strained while the visuals loaded) compared to the older, albeit more limited, table view.
Additionally, the distinction between the various types of data display on the visual dashboard isn’t immediately clear. My posts and “friends” posts are mixed together with lists of top tags, locations, communities and the like. While Kred Story may aid serendipitous information discovery, the confused visual flow doesn’t lend itself to information at a glance analysis.
4. Display Scores
Each service prominently displays an individual’s composite score in a badge form. In reality it appears that the badge serves primarily to promote the service.
A single context-free metric doesn’t really say enough about a potential influencer to be useful. There’s no information about their authority on a particular subject domain nor does a score convey information on the demographics of their community of followers.
A badge score is much like the simplistic answer too many politicians give to a complex problem: people want to believe, many do, but they probably shouldn’t.
Each of the major services will also break out a user’s composite score in a series of more specific subscores. Klout provides, as described on their website:
- True Reach − Your True Reach is the number of people you influence. We filter out spam and bots and focus on the people who are acting on your content. When you post a message, these people tend to respond or share it.
- Amplification − Your Amplification is how much you influence people. When you post a message, how many people respond to it or spread it further? If people often act upon your content you have a high Amplification score.
- Network Impact − Your Network indicates the influence of the people in your True Reach. How often do top Influencers share and respond to your content? When they do so, they are increasing your Network score.
PeerIndex describes its subscores as
- Authority, the measure of trust; calculating how much others rely on your recommendations and opinion in general and on particular topics.
- Audience, a normalised indication of your reach taking into account the relative size of your audience to the size of the audiences of others. In calculating your Audience Score, we do not simply use the number of people who follow you, but instead generate from the number of people who are impacted by your actions and are receptive to what you are saying.
- Activity Score is the measure of how much you do that is related to the topic communities you are part of. By being too active, your topic community members tend to get fatigued and may stop engaging with you; by taking a long hiatus on a particular topic, community members may not engage with a long absent member. Your Activity Score takes into account this behavior.
Kred is composed of two scores:
- Influence, measured by assessing how frequently you are Retweeted, Replied, Mentioned and Followed on Twitter. If you connect your Facebook account to your Kred profile, you get Influence points when people interact with your content on your wall and the walls of others who have registered their Facebook account with Kred. Facebook interactions counted towards your Kred include Posts, Mentions, Likes, Shares and Event Invitations.
- Outreach − We measure Outreach on Twitter by your Retweets, Replies and Mentions of others. When your Facebook account is connected to your Kred profile, you get Outreach points for interactions on your own wall and the walls of others who have registered their Facebook account with Kred. Interactions counted towards Kred include Posts, Mentions, Comments and Likes. your Outreach score is cumulative and always increases
Personally I wouldn’t read too much into these top-level scores as in most cases, we’re going to want have scores for a specific target market.
Scores Break Down Into Topics And Communities
Each service also associates people with thematic communities and calculates their score relative to the members of the community. Klout calls the themes topics, as does PeerIndex, while Kred prefers the term communities (Kred also hides them quite well, behind a blue arrow next the the Global Kred score).
This type of semantic analysis, while welcome, can be hit or miss. Klout believes yours truly is influential about Hurricanes and Poker, which is highly unlikely in both cases, although I’d like to think they’re right on the other topics they’ve identified including SEO, Google Analytics and Social Media.
While the general public focuses on the simple catch-all badge score, it is the ability to look at an individual’s rank within a topic which is going to be of interest to most marketers.
Influencer Identification, However You’d Like To Dice It
The raison d’être for the influence raters is to help marketers influence the influencers, help which also funds the services. Each service offers to help select influencers appropriate to a specific marketing campaign.
Kred mentions using keywords, hashtags, location, community or any combination of these, to identify influential individuals.
Beware: Those Who Participate May Be Scored Higher Than Those Who Don’t
Adding more networks helps us more accurately measure your influence and can only increase your Score.
(The emphasis is theirs). Translated: individuals who add Facebook and other social networks will likely have higher scores than those who ignore Klout, Kred and PeerIndex.
You should be aware of this when you ask Klout, Kred or PeerIndex to select influencers for you. Remember many, if not most, of the profiles represent just a user’s Twitter activity, ignoring the rest of their social activity on the web.
This article is part I of a two-part series. Here’s the second part, that looks at methodologies, strengths and weaknesses of these social scoring methods.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.