Digital Analytics – From Back Office to Front Page

Analytics have exploded into prominence in the past 15 months. What was once a mysterious statistical discipline understood by few has been elevated as the enabling technology that allows companies to unlock the potential of Big Data. Big Data was everywhere in 2012. There was a track devoted to it at the World Economic Forum in January, 2012. In October, the Harvard Business Review had a cover section on Big Data which characterized analytics as sexy and dubbed its leading practitioners Data Scientists. And at Ketchum, we made analytics training mandatory for ALL employees in 2012, a first for our industry.

The digitization of all forms of analog data is at the heart of the Big Data explosion.  From our click behavior online to purchases we make with a loyalty card to places we go in our vehicles, everything is captured as digital data potentially available for analysis. And with the accelerating use of cameras and sensors, the volume of data promises to keep rising for years to come.

For all the talk about Big Data, no one really wants Big Data. They want the insights hidden within the data that only digital analytics can unlock. Probably the hottest area within digital analytics is predictive analytics. Predictive analytics essentially predict how consumers will behave given certain conditions, assumptions and stimuli. This has powerful and tangible benefits to marketing. An Aberdeen Group study published in December 2011 found marketing organizations that applied data mining and statistical modeling to optimize marketing efforts saw as much as a 2X lift from marketing campaigns, a 76% higher click-through-rate and a 73% higher sales lift.

Marketing organizations are seeing the tremendous power and potential of predictive analytics across a broad spectrum of marketing activities. For example:

• Proactively optimizing marketing campaigns to improve engagement and conversions
• Identifying customers most likely to switch companies and targeting offers to them designed to keep them as customers
• Delivering customized offers designed to appeal to a prospect’s specific interests or life situation
• Predicting which ‘first visit’ customers are most likely to return or not return, and sending offers targeting the ‘not likely to return’ group while avoiding the costs of making offers to those likely to return anyway.

In 2013, we have seen analytics continue to be a mainstream news subject. Analytics is only in the early migration phase from early adopters to mainstream use. Expect this trend to continue and accelerate in the coming years as more marketers discover analytics are a clear path to improving marketing effectiveness, efficiency, and ultimately, the bottom line.

Three Keys to Insight Discovery in Social Listening

True social insights, as opposed to social findings or social observations, have the potential to inform, shape or drive marketing and even business strategy decisions, not just social strategy decisions. Discovering that tweeting with a link on Tuesday between 10:00 – 11:00 AM drives higher levels of engagement is a social finding, not an insight.

Social media is a microcosm of the larger Big Data problem/opportunity – too much data, not enough insights. Or if you prefer, too much noise, not enough signal. If you want to improve your ability to discover insights, here are three simple approaches you can take to improve your insight hunting.

First, start all analysis with a hypothesis or series of questions the analysis is designed to answer. It is much easier to prove or disprove a hypothesis, or answer specific questions, than it is to “find out what people are saying about us in social media”. The more specific the request, the better the answer is going to be. The hypothesis may be one you develop based on preliminary analysis or it may come from the ‘customer’ for the insight. Here are two examples of hypothesis:

• “Conversation about us in social media is quite negative. My boss believes ‘everyone’ is aligned against us. I disagree. My hypothesis is that there is a very vocal and active minority of consumers who are posting large volumes of negative content about us. And I believe this group is a small fraction of the total number of people who post. The majority of consumers are actually neutral toward us.”

• “When we look at Twitter, Facebook and Blogs we see pretty low levels of conversation about our product and the medical condition it treats. We believe there is actually a fair amount of conversation, but the conversations are occurring in Forums which may not be crawled by most social listening tools.”

While the hypothesis is a great way to begin to focus on what is important in the data, a further focusing mechanism is the second insight discovery key – the concept of targeted listening. With targeted listening we are not trying to capture all conversations that mention the brand or product. That is a ‘boil the ocean’ approach. Instead, we listen for very specific types of conversations or conversations by very specific groups of individuals within social conversations. The trick is to have the discipline to only listen to your focus areas and not be tempted to boil the ocean in hopes of finding a few pearls. Here are three examples of targeted listening strategies:

• An insurance company resists the temptation to try to capture ‘all’ mentions of the brand and decides to focus only on conversations where customers are thinking about non-renewal or switching companies.

• A gaming company launches a new product and listens to understand what features are being discussed, what people like most/least about the new game and to gauge their specific reactions to the cover art.

• A consumer products company listens only for consumers who are actively in the purchasing process for the type of products they offer.

The third key to discovering insights is to provide context for decision-making. Remember with insights we are trying to inform, shape and guide decision-making. Context is incredibly important to making better decisions faster. Good social analysts understand how marketing and business work and how strategic alternatives might impact results. Understanding this helps you put your insights in the proper context for decision-making.

Here is an example of how context can lead to better decisions. Company X has a crisis. You are asked to do real-time listening of the crisis and help the PR team decide when and how to engage in the conversation. You come back the next day with a line chart showing a large spike in content mentioning the crisis – thousands of mentions. You know the sentiment in negative to neutral and on which channels the content appears. Unfortunately you have not given the people deciding

whether or not to engage enough information to make a decision. What information would provide the necessary context for decision-making? What questions do we need to try to answer? Here are a few:

• How much above ‘normal levels’ is the spike in content? (Normative data)

• How does this event compare to that event we had last year? Or, how does the event compare to competitor X who had their own crisis last year? (Comparative data) Comparisons help decision makers determine ‘how bad is bad’.

• How long do we anticipate seeing negative content at relatively high levels? (Comparative data) This might be the most important question to answer to provide context and guidance for the engagement decision. If we anticipate volume will drop back to normal in a reasonable period of time, then not engaging may be a viable and effective strategy depending on the brand involved and the nature of the crisis.

• Which stakeholder groups are active in the conversations?  With robust social analysis we always want to look at both the post – what is being said, and the source – who is saying it. In a crisis, who is talking is particularly important.

Normative data, comparative data and examining both post and source data are all effective techniques to provide context for decision-making. 

The tough part about discovering insights is there are no shortcuts and it is a human activity. No social media analysis platform that I have found has an insight button. The key barrier is lack of people who understand how to search for and discover insights. Hope these tips make you a more effective explorer.

Happy Holidays!

Where is Your Organization on the Social Media Listening Maturity Model?

Quite often I am asked to consult with a company on their social media listening strategy. Their initial question more times than not is about the listening platform they should use. But it is increasingly common for the questions to be more sophisticated and the ambition behind them to be much greater. Companies with experience in social listening know that it is all too easy to focus on rudimentary analysis of brand mentions and topics, Followers and Likes and never get to the truly actionable insights that lead to marketing or business actions. Experience in listening is an important element here but you also need a path to follow. I thought a maturity model approach to social media listening could provide a possible path to consider and would provide a construct that could be used in consulting with a company on their social listening strategy.

Maturity models are sort of hot – there seems to be a proliferation in the last two years or so. One that I find particularly insightful and helpful when thinking about social listening is Forrester’s Social Maturity Model.  Two really important points the folks at Forrester make is that listening is not the goal, social intelligence is, and that social intelligence informs actions taken by marketing or some other area of the business. Action being the operative word here. Social intelligence is a closely related topic to social business, and if social business is more your thing the Dachis Group has an interesting social business maturity model.  Big data more your bag? Check out IBM’s big data governance model. After looking at the models out there, I could not find one specific enough to social media listening so I took a stab at creating one.

Social Media Listening Maturity Model 

There are five stages in the Social Media Listening Maturity Model, beginning with reactive alerts and ending with social intelligence. Let’s take a brief look at each stage and some of the overarching differences or changes one sees with social listening maturity.

Reactive Alerts – Many companies or brands begin by establishing a reactive alert system that notifies them whenever their brand is mentioned or is mentioned with specific keywords. Think Google Alerts. Companies in this stage may only periodically check social media channels to see what may have changed or is new since the last check-in.

Monitoring Social Media – At the next stage, the company has begun active monitoring of all ‘owned’ social embassies. They also are monitoring social media conversations, often focused on trying to detect any ‘bad’ news, mentions or conversations.

Companies in these first two stages generally have a reactive stance toward social media, viewing it as another way to find out about news and circumstances that may harm or otherwise impact the organization. It is common for companies in these stages to use one or more of the various free tools available to gather web and social media data.

Social Listening – The third stage is most likely where the largest percentage of companies reside today. Companies in stage three are listening to social conversations about their company, brands and products. They are tracking mentions of competitors and calculating share of conversation. Many also track issues and topics that are important to their brands/products/company. At this stage many begin to put additional emphasis on ‘who’ is talking (source) not just what is being said (post). Most companies in the social listening phase have transitioned from free tools to paid platforms.

Companies in the first three stages often suffer from having too much data and not enough insights. They are up to their necks in ‘big data’ but lack the experience and expertise to analyze the data and reduce it down to crisp, actionable insights supported by the data. They look for the Insight button on the tools they use but increasingly realize insights are the product of human analysts, not tools or data.

Strategic Listening – The transition to strategic listening brings with it a bias toward ‘listening with a purpose’. I first heard this turn of phrase from my friends at Radian6 and use it often. Listening with a purpose is just that – listening to specific sets of conversations with a specific goal or objective in mind. Often in insight work, the goal or objective may take the form of a hypothesis we are trying to test. Here are a few examples of listening with a purpose:

• Listening for conversations of consumers in a particular phase of the buying decision process
• Listening to customers whose subscriptions or policies are about to expire that are expressing thoughts of changing vendors
• Identifying, tracking and building relationships with key influencers
• Listening for consumer reactions to new packaging or product features
• Mining the emotional content of specific stakeholder groups to determine potential risk around a sensitive issue.

During this phase, an Enterprise listening strategy is often developed and implemented. Some also begin to integrate data from sources beyond social media – search, web analytics and customer data for example.

Social Intelligence – Forrester defines social intelligence as the process of turning social media data into actionable marketing and business strategy. Social intelligence therefore is not about the best times to tweet or whether or not a twitter party would be an effective tactic, it is about informing strategic decisions that impact the company’s success. For me, three concepts are crucial:

1. Action – social intelligence is designed to drive true actions.
2. Integration – although the definition focuses on social media data and insights, the fact is that true insights often require more than just social data. Integrating data from multiple data sources – consumer survey, behavioral tracking, social posts, search analytics, advertising data, customer records, scan/sales data – allows for greater understanding and richer insights. Integration of multiple data types often requires multiple tools and platforms to aggregate and analyze the data.
3. Sharing – For social intelligence to truly take root within an organization, the data and insights should involve cross-disciplinary groups that can look at the data from different perspectives and collectively arrive at better insights than any one group could in a vacuum. The insights then need to be systematically shared broadly across the organization so they may be acted upon in a manner that will create the most impact. Social intelligence can be a catalyst to the silos within an organization tumbling down.

Since the social listening and social intelligence ‘markets’ are relatively immature, this model will continue to evolve and be refined.

Where is your company today on the social media listening maturity model?

Time to Get Real About Social Media Audience Reporting

Though almost everyone would agree that social media is about engagement and not eyeballs, too much of digital and social media measurement is focused on audience size. How many Followers do we have? How can we get a million Likes? How many unique visitors did we have to our site this month? And unfortunately, audience size estimates in social media grossly overstate the actual relevant audience. We seem fixated and oriented toward ‘how many’, while our focus should be on ‘who’ and specifically, ‘who within our target audience’. Generally speaking, the advertising industry has led the way with audience measures and is ahead of where the public relations and social media camps are with respect to level of sophistication.

In television advertising, the concept of Target Rating Points is a refinement of Gross Rating Points where you only measure and get ‘credit’ for the percentage of the gross audience that meets your target audience criteria. In an effort to keep refining the audience data available, Nielsen has evolved from diary-based data to electronic data to software at the set-top box level that allows operators to monitor channels choices and changes. In audio-based media, Arbitron’s Portable People Meter recognizes today’s mobile world and begins to address cross-platform measurement. It is also interesting to reflect on the U.S. Congressional involvement in television audience ratings accuracy (or lack thereof as it were) that led to the formation of what is now known as the Media Rating Council in the early 1960’s. The time has come for social media audience research to greatly increase in sophistication, accuracy and relevance.

When we think about social media audience size measures today, the emphasis is on Opportunities To See (OTS), although almost never by this name. We might call them Impressions or Reach, but what we really mean is how many people had the potential to see this content item. There are two overarching issues here:

• Opportunities to see are not the same as actually seeing
• The metrics count all possible members of the audience, regardless of whether or not they are part of the targeted audience or can even buy the product or service.

OTS is also a prevalent metric in the public relations industry which has always focused on stating the highest possible audience measures. In traditional media we know the probability of any one person in the audience actually seeing the article in question is a fraction of the total audience – a reasonable estimate is 10% or less. So OTS greatly overstates the actually number of people who saw a given article. To compound the audience overstatement, we have the practice of using audience multipliers to ‘credit’ earned media for either a perceived credibility advantage over advertising or to account for pass-along circulation (see this IPR white paper for more on multipliers). Thankfully the practice of applying multipliers (and its evil cousin AVEs) is out of favor and rapidly on a path toward extinction.

In social media one can make the case the audience metrics situation is actually exacerbated in that the probability of any one follower seeing any one tweet, for example, is most likely an order of magnitude less than in earned media – my guesstimate is 1% or less. Before you call BS on this guesstimate, play around with a few Twitter factoids – the recent Pew Research study suggesting only 8% of Twitter users use it daily, the perishable nature of most individual’s twitter streams, and the fact that a reasonably high percentage of Followers of a brand are bots, and the reality is that only a small fraction of twitter followers actually see tweets, let alone find it interesting enough to share or comment on. And, of course, not all Facebook Likes see every post you make either. Riffing on the old, ‘if a tree falls in the forest…’, if you tweet into the twitterverse and no one sees it does it make an impact?

Evolving from ‘opportunities to see’ to ‘relevant audience’ measures.

Most social media campaigns have a specific target audience in mind, often described with demographics (Female, age 18 – 34), psychographics (who worry about feeding their family healthy food on a budget) and behavioral (access deal and coupon sites regularly) dimensions. Yet when it comes to reporting and measurement we take credit for the entire audience (total OTS) rather than the percentage of the audience that meets our targeting criteria. Trying to promote lingerie to 22 – 29 year old ladies? No worries, count all your Twitter Followers and all the visitors to your website – the men, the young and the old – everybody counts. Trying to sell camo clothing to male hunters? No worries, everybody counts – male, female, hunters, non-hunters and PETA members, too. Of course this all seems a little silly and strange and I suppose it would be if it wasn’t the way most social audience reporting is done today. It is unusual to see someone in social media, or PR for that matter, report only the relevant audience opportunities to see. Why is this? I believe there are three primary reasons:

1. Legacy – the PR industry has historically reported gross potential audience size rather than the relevant audience size. When social media came around, this same orientation toward gross audience measurement was used.
2. Data – there is a lack of consistent social media demographic and psychographic audience data available and it often resides in channel silos rather than cross-platforms. And often the audience data from one platform (e.g. ComScore) does not match the data available from another platform (e.g. Compete).
3. Standards – there are no standards for social media audience metrics and no codified best practices for audience measurement.

Where do we go from here?

First, we need a change in mindset of how we think about audiences. From ‘how many people theoretically had the potential to see our content’ to ‘how many of the people we were targeting actually saw our content’. Big audience numbers are irrelevant. Relevant audience numbers are big.

Next, as the demand for audience data that contains demographic, psychographic and behavioral data grows, it is reasonable to assume one or more of the large media data companies might start to aggregate and make the data available. Privacy concerns, cookies and other issues are also in play here.

And last but not least, industry standards for social media audience and engagement metrics and definitions are necessary for transparency and replicability that will increase credibility of social media measurement and reporting. 2012 will go down as the year that serious cross-industry progress on social media metrics standards began and gained momentum. There has already been a lot of progress (See this post from Katie Paine), and this week in Dublin at the 4^th AMEC European Summit on Measurement the theme is around attempting to define standards for social media metrics and measurement. To tune into the debate as it occurs in Dublin, monitor #SMMStandards and #AMEC2012.

What  are your thoughts on the need for social media metrics standards and the use of target rather than gross audience size estimates?

Three Fundamentals of Great Social Media Measurement

If you want to evaluate the robustness and effectiveness of your approach to social media measurement, ask yourself these three fundamental questions:

• Does the approach measure the ‘right’ things in order to show the business impact of the programs and initiatives? 
• Will stakeholders of the report receive the data and actionable insights required to make strategic decisions?
• Are the data and insights presented in a clear and concise manner that tells a story and makes it easy to understand and act upon?

Measuring the ‘Right’ Things

Social media metrics are derived from three primary sources:

Ideally, a robust social media measurement program will have a rich metrics set that contains metrics from all three areas. Metrics tied to program objectives allow for direct measurement of program success. Fundamentally, measurement is about assessing performance against objectives. It is surprising how often social program objectives are slanted toward channel-specific metrics (e.g. Likes or Followers) and not the specific outcomes desired for the program – what you hope to accomplish by implementing the program. Also, relying too heavily on channel metrics limits you to what you can measure rather than what you should measure. Business outcome metrics are used to connect the dots between social media programs and the business results they are designed to drive. Social programs that cannot answer, or at least address, the management question, “How is this impacting my business”, are more susceptible to resource allocation scrutiny (#pleasecutmybudget). Stated another way, if management asks how we’re doing in social media and we reply, “great, post virality is up 6.1% this month”, we make it difficult for that individual to understand how social media/business initiatives are helping move the business forward.

Getting to Data and Insights that Inform Strategic Decisions

Expectations for social media measurement and analysis have risen. In addition to sound analysis and reporting of performance against key metrics and KPIs, understanding audience dynamics and developing actionable insights are rapidly becoming de rigueur. Insights may be defined as synthesizing and interpreting data to provide actionable information and knowledge that informs strategic decisions. Too many social media measurement programs take a social-centric rather than a business-centric approach to insights. They often feature insights and recommendations that are tactical in nature – the best time of day or how many times to tweet, or what type of content seems to be most successful. Ideally, insights and recommendations in social measurement reports would be operating one level above this, informing strategic decisions about how social programs and conversations are impacting, or could impact, the business. To do this requires an understanding of the business function (e.g. marketing, customer service) impacted by the social program and an ability to ask the right questions prior to starting a social media analysis. 

For example, let’s say Company X plans to introduce a new video game. A social listening program has been implemented to analyze the early consumer reaction to the game. Based on the listening analysis, changes to the packaging, marketing or even the product itself are possible. If you are in charge of the marketing campaign for the game, what are the types of social media insights you need to make decisions about the game and the marketing campaign?

• What is the level of buzz about the game?  What is the overall sentiment? How does this compare to previous game launches?
• What are people talking about in social media – availability, cost, specific features of the game, packaging, marketing campaign?
• What features of the game do consumers seem to like most?  Least? Specifically, what do they like or dislike?
• What are the most influential gaming enthusiasts saying about the product?
• Who are the promoters and detractors? What is the ratio of promoters to detractors? How does this compare to promoters and detractors from previous game launches?
• How much social media conversation contains recommendations or expresses purchase intent?  How does this compare to previous launches?

Answering these types of questions provides actionable insights that provide context and can inform strategic marketing decisions.          

Presenting Results

Dashboards have gotten a bit of a bad rap – not because dashboards are not useful, but because some have used them as THE measurement report rather than just one aspect of a good report. I’m a dashboard proponent for a few reasons:

• Deciding which metrics to feature on a dashboard is a good strategic exercise requiring you to focus on the very most important and relevant metrics for the intended audience
• Online, dynamic dashboards are an effective user interface that can be used as a launching- off point for drilling into data to understand the underlying story
• Good dashboards present a snapshot of overall performance that is easily absorbed and understood.   

A dashboard-driven social media measurement report is versatile and effective in many situations. A typical report might consist of one of more dashboards and then a deeper dive on each of the key metrics featured on the dashboards, along with audience insights, strategic insights and recommendations. This format provides a quick snapshot (dashboard) of results, ideal for those stakeholders interested only in topline data, and provides sufficient depth to satisfy those more interested in the underlying drivers of the metric  

Social media measurement programs that are built around metrics tied to business outcomes and show how programs are performing against objectives are important. Reports that deliver clear insights that inform strategic decisions are important. And delivering those reports in a compelling format that enhances usability and effectiveness is important. How do your programs stack up?