Despite the recent hype, predictive analytics has been around for over 80 years. At least the mathematical models underpinning this branch of data science have. It has only been in the last few years that thanks to increasing computing power and nifty software interfaces, analytics has become accessible for the average person -- and the marketer.
So what is Predictive Analytics?
Let's start with the word predictive. As marketers, we can "predict" a lot of things. We know intuitively that certain products sell better in certain seasons (think of ice cream or leg warmers), at certain times of day, and at certain physical locations. Supermarkets and convenience markets "predict" the demand for anything from alfalfa to yogurt.
B2B marketers like myself know that we will be busier during certain times of year, when corporations plan ahead for the coming year for example. We know pretty much from experience what is going to happen at any time or day. Social media marketers know that posts at 4 pm will receive more likes than those at 4 am, and so on.
The value of prediction is specificity
Intuition is all very good, but the real value of prediction lies in specificity. That is to say, if you email that person on these days and those times, you have a 50% higher chance of receiving an order. This is where machines come in. The amazing advances in computing power mean that marketers using predictive analytics can save a lot of money and time while achieving measurably better results.
So what do you need to get started with predictive analytics
Predictive analytics has become a lot easier. You don't need a math degree anymore; software does most of the work. But a marketer still needs three skills (or a marketing team three people) to get the best value from predictive analytics. Firstly, you need someone who knows how to get the data and understands it. Whether it's Google Analytics, Facebook, Instagram, or any other platform, or CRM system, you need to be able to extract the data from the platform you are using.
This is more complicated if you are including internal legacy systems; most social media platforms offer good API that make this relatively easy.
Secondly, you need someone to clean the data up, a "data scientist". Clean it up, because data can be messy and misleading. Fields can be misaligned, language codes badly set, or combinations of data sets lead to entirely misleading outcomes. You can try this by looking at your own preferences on Facebook. Why on earth does Facebook think I like sailing and expensive watches? Turns out I once watched a video of the Rolex Cup. Ah well.
Finally, you need marketers familiar with using data to achieve marketing outcomes. What is the point of measuring anything if you won't act on it.
Small marketing teams benefit most from analytics
Most marketers work with limited resources; some are even all on their own. The biggest benefit you will get from predictive analytics is better time management. At Geber Brand Consulting, we use our CRM data for the past 12 months to forecast which times we will be busiest interacting with clients, and which weeks or months we get the fewest incoming requests. We can then schedule our blogging, podcasting, or website revamp at the most suitable times.
Search data from Google is very useful in this regard. We see spikes of 18-25-year-olds in Taiwan googling "social media marketing" in October - those will be students deciding which course to take. But we see the same spike in older people, combined with "marketing jobs" in August - this is when we start our hiring push on LinkedIn. If we correlate this with post engagement on our blog, we see that in November and December, companies are looking for help with social media marketing - which is when we run our Google and Facebook ads to lock in business for the next year. Learn predictive analytics to keep your jobFrom the example above you can see how important it is to understand the data. Marketing is no longer a hit-and-miss strategy, it can be based on real-time, valuable insights.
In the near future, Artificial Intelligence will make this task even easier. (Read: 4 Ways How AI Can Augment the Digital Marketer) You will be able to feed more and more data into smart systems, and the AI will tell you what to do and when to do it.
We are currently experimenting with a tool that lets us combine ZOHO CRM with Google and Facebook data to predict actual earnings potential of the company for the next 12 months. Not easy, but the boss loves it.
Organizations using this kind of data in the right way gain a competitive advantage. (Read: How digital companies are leaving the rest behind) Marketers using it will become better marketers, with better reporting skills, better outcomes, and higher earning - which in a competitive market place is a great way to make sure you don't get fired.
Is predictive analytics right for my business?
There are certain things you cannot predict or do not need to predict. If your business is entirely dependent on the purchasing behavior of a few big companies (in the B2B space, say, semiconductors for example), you really don't need to waste time predicting the behavior of other market participants. If you are hiring students out of school, you know when school ends and young engineers are looking for jobs; you don't need software to predict that.
However, there might be applications for predictive analytics you are overlooking. We are handling the social media marketing for the HR department of a big B2B company and used analytics to determine that 8-10 pm was the best time for Facebook Live to engage with potential hires. Turns out that at those time, they were mostly watching entertaining videos and sports or news livecasts, and didn't want to think about finding a job. The best time to reach them was when the data showed the lowest engagement with Facebook Live, i.e. when there were no distracting other livecasts available. In this case, basic A/B testing may be your savior.
Your actions skew the data
On a final note, there is one aspect of analytics that is a bit problematic. Your own actions may lead to misleading interpretation of the data. Every time we publish an insightful article that gets picked up by the marketing community, the subsequent search data and incoming job requests spike. You may have thought that this month was a good time to post about this topic, but it is only so because you did that last year, and the spike was actually caused by your activity. Sometimes one-off events such as conferences, or media hype over a vaguely related issue can skew the data too. This is why clean data, and understanding data, is so important.
On the other hand, for many marketers working primarily with social media and websites, the skewed data doesn't really matter so much. Don't worry about getting things a little wrong at the beginning. It is far more important to get started and do it than to be afraid of failure! "If you can move the needle, enjoy the success!"
