Predictive Analytics and the Future of PR

Public relations, marketing, and communications aren’t known for their futurist perspectives. More often than not, we’re reacting, from crisis communications (when the news is bad) to rapid response/newsjacking (when the news is good).

Until now, our ability to keep a finger on the pulse of the news has been good enough. We’ve reacted fast enough to keep crises mostly contained; for every major PR crisis in the news, there are dozens, if not hundreds that are intercepted by talented PR professionals before the spark turns into a conflagration. We’ve jumped on the bandwagon enough to insert our perspectives, our stories, and our brands into trending news stories. However, relying on just what’s happening now isn’t enough any longer. In the new era of intelligent machines, we marketers and communicators must learn to predict what will be.

Public relations, a communication-driven field, can and will continue to benefit from predictive analytics.

What is Predictive Analytics?

So, what exactly is predictive analytics? By definition, it’s the “statistical discipline of predicting what’s likely to happen based on existing data.” Essentially, in its most basic form, this is the practice of connecting numerous points on a plot, giving us the ability to “connect the dots.”  The expansive variety of software and technology at our fingertips allows everyone the opportunity to turn data into valuable insight.

It’s nothing new; we’ve attempted to predict the future since we first became a rational species, from Mayans using it in astronomy to children learning basic predictive skills in math class. What’s changed is our ability to predict with extremely large, complex datasets - and how accessible such prediction technology is. With advances in software and computing power, we now make predictions with many more data series, increasing the accuracy and richness of our powers of prediction.

For example, in years past, if we wanted to predict the health of our business, we might assemble revenue, sales numbers, and perhaps a few marketing statistics. Today, we ingest thousands of news articles, millions of social media data points, web analytics, customer service data, CRM data - the list of data sources is nearly endless. Nearly every possible way for us to communicate with our audiences has a digital twin - a digital way to measure the activity, from location-based check-ins to photos posted online to voice conversations we have with company representatives.

With new open source technologies, turning data into valuable insights is within everyone’s reach.

How Does Predictive Analytics Apply to PR?

A great example of how predictive analytics applies to PR is the process of creating a crisis communication or community management plan. By modeling public forum, product reviews and social media data, PR pros can use machine learning to extract the most common complaints and talking points. Using this information allows you to build out an effective data-driven action plan.

Or suppose we want to understand how an audience is likely to react to a negative corporate announcement. Since very little is completely new in the world, we should be able to find a dataset of a similar circumstance - perhaps even in our competitive set - and build a model of how the audience reacted to that announcement. For example, if we’re a telecom company raising rates, finding reactions to a competitor’s rate increase is straightforward. Once we have the data, we can model it over time and use machine learning to extract the most common complaints/talking points from the past and develop an action plan for the future.

In yet another example, suppose a company wants to understand when demand for its product or service will be highest. Using publications, search data, and social media conversations, we construct a model of demand for the product or service looking forward and give the company a proactive calendar to plan against.

What Data Is Needed to Leverage Predictive Analytics as a PR Team

If you’re a PR Team looking to begin using predictive analytics for future endeavors, look no further. Discover first-hand how to ensure you are using clean, compatible data that was chosen well. The purpose of PR is to generate awareness and trust, so start exploring how to apply data collected about your customers to drive initiatives.

If our data lacks any of these attributes, creating reliable predictions will be impossible.

• If our data is corrupted, filled with junk, or flat-out wrong, our predictions will simply magnify these errors. To prepare for predictive analytics, we must clean our data as best as possible.
• Most predictive analytics software and systems are based on open-source libraries and technologies. As such, we must ensure our data is in compatible formats so that as little conversion/manipulation is required prior to import. Some of the most popular data formats for compatible numerical data transfer are: CSV files, SQL files and TSV files
• To prepare for predictive analytics, we must choose our data carefully, with the right amount od detail. Choosing data well comes from great data governance, including documentation of what’s in our data, how we acquired it, and how we prepared it.

Sources of data for predictive PR can include: 

• Web analytics from systems like Google Analytics™
• Social media data
• Publication data from the many media monitoring systems out there
• Massive numbers of public data sets
• Trend data from reputable providers like Google
• SEO data
• CRM data

Matching customer behavior to public relations

The purpose of public relations is to generate awareness and trust. We help connect our companies with audiences who matter most to the company, from investors to activists to customers. However, public relations operates on a calendar largely dictated by either internal operational mandates - “it’s end of quarter, we need more website traffic!” - or editorial calendars for publications.

We don’t investigate often enough the way our audiences - the public in public relations - behaves and time our efforts to what the audience wants most. The PR industry operates this way mainly because we’ve not had access to predictive software or good data. That’s changed in recent years as the data floodgates have opened and predictive software has been democratized. Services like IBM Watson Analytics allow us to do powerful statistics and analytics without being statisticians.

Example: matching search intent to PR

How might we understand customer behavior to inform our PR efforts? Suppose we looked at a very common data source: search engine data. Using services like Google AdWords™, Google Search Console™, Google Analytics™, or even Google Trends™, we’ll find clean, compatible, well-chosen audience intent data to predict customer behavior. Let’s take the organic traffic from my personal blog as an example. Suppose I extract five years’ data from my blog of traffic from search:

This traffic data tells me when my audience is searching for topics of interest that my blog answers. If I knew when my audience was searching for topics the most intensely, I’d incorporate this information into my public relations program so that I’d be in the news right at the time when interest was highest. Using machine learning software such as timekit, scikit-learn, etc. we take this data series and project it forward, using the existing 5 years’ data as the training data. This is what my year looks like:

What we see above is the future: a prediction of what my blog’s likely traffic over the next 12 months will be, based on machine learning analysis of the past 5 years. Let’s look at some key points:

• A: The fall is a good time - and unlike other B2B marketing sites, I don’t see a Thanksgiving slowdown, just a Christmas one. This I know from other research is because a significant portion of my traffic is from outside the United States.
• B: The first quarter look strong until right around the middle of March. I receive a ton of search traffic in January. This is due in part because I have lots of content about marketing plans and strategies, which are hot search topics in January.
• C: I don’t see much of a summer slowdown. The winter holidays are when I see the biggest search traffic volume drop; otherwise, the summer months are good to keep publishing.

What would PR do?

A PR team should advise me to craft stories, topics, articles, bylines, etc. to match key industry themes, news and editorial calendars in the months when my search volume is highest, so that I’m in print when people are already finding my website. Using the above as an example, my PR team would advise me to write content they can pitch for precision marketing. Second, they'd help me create content and creative assets (video, audio, and infographics so that they’re in publication - with links to my website - in January. Third, my PR team would hit up events planning calendars for prominent weeks, like the week of November 26, based on my search traffic. What current events that week could I participate in, even remotely?

Finally, my PR team would inventory the greatest hits I’ve received over the past few months that have faded into legacy and prepare paid syndication campaigns to bring them back to life at peak times. Paid syndications make old news fresh again and having great coverage (as long as it’s still relevant) resurfaced helps ensure I’ve got mindshare when I need it most. Finally, at the inevitable lulls and dips throughout the year, my PR team would help me craft bylines and other content pieces for a regular drumbeat of coverage at times when my audience’s interest in me is lowest, like in section C in the chart above.

Watching out for predictive analytics pitfalls

While incredibly helpful, it's easy to take missteps when working with predictive analytics. The first and most common scenario in predictive analytics is flat-out bad data. If we have data which is poorly formed, broken, incompatible, etc. - and we don’t know it - then our predictions are likely to be very wrong.

Another circumstance in which predictive analytics often fail is with confounding variables. These situations occur from our failure to understand our data and the context it occurs in. To use a classical data science and statistics example, suppose we’re modeling and predicting ice cream sales. We’ve got great sales data from the last 50 years, and we’re building our model based on it. Yet, the next year we look back and we see our predictive forecasts were terribly wrong because the summer was unseasonably cold. We need to engineer anomalies and "black swan" events.

The final way predictive analytics goes wrong isn’t with the prediction, but what we do with it. All descriptive and diagnostic analytics, being based in mathematics and statistics, can only tell us what happened. Predictive analytics models, built with the same math and statistics, will only tell us what is likely to happen. None of these analytics ever explain why something did happen or why it will happen. None of these mathematical models understand the humans often at the root of the data we’re studying. Never mistake what for why. Our models help us plan and predict what is to come, but we still require human insight and judgement to determine whether the circumstances of the model remain appropriate.

The power to predict

The power to predict is neat in and of itself, but combined with great PR strategy and tactics, the power to predict makes our PR efforts a force to be reckoned with. Instead of guessing when to conduct our most aggressive PR campaigns, or scrambling to react to internal needs, predictive analytics enables PR professionals to bring real strategy to the table.Predictive analytics is the epitome of data-driven PR. We know, based on data and careful analysis, combined with insight and the most advanced technology, what is likely to happen. Once we know what’s likely to happen, we’re able to build thoughtful strategies, relevant tactics, and timely execution to maximize the impact PR has on our overall business goals.

‍