15 Ways To Use Predictive Analytics To Become A Better Digital Marketer


This post is an excerpt from our Digital Growth Summit event in Los Angeles.Here are the digital marketing experts who contributed to this blog: Jesse Leimgruber Predictive Analytics Solutions Ben Williams Predictive Analytics Solutions Nick Metcalfe Swamy Narayanaswamy Predictive Analytics Solutions Dr. Alex Liu Predictive Analytics Solutions

15 Ways to Use Predictive Analytics to Become a Better Digital Marketer

Here are some key takeaways from the Predictive Analytics Solutions panel:

  • Know what you’re looking for
  • The data can be overwhelming. Rather than trying to use all of it, hone in on the questions you really need answered and look for the patterns.

  • Use the tools
  • Our experts have a lot of suggestions for predictive analytics solutions they use to break down the data, including AlchemyAPI and Google Trends.

  • When you find a pattern, test it out
  • Predictive analytics are great at revealing trends. But if you change strategy based on that data, run another test to see if how your theory plays out in the new data.

With so much data out there, where's a good place to start?

Know the rules and patterns that drive your business. —Swamy Narayanaswamy

If you know the rules and patterns that drive your business, a data expert can come into your business and help you a) capture that data and b) analyze it fast enough so that you can actually take advantage of it. —Swamy Narayanaswamy

Data analysis is finding previous things that have happened and trying to build a pattern out of it. —Ben Williams

When you think about data analysis and analytics in general, it is pattern matching. It is finding previous things that have happened and trying to build a pattern out of it. It can't tell you real predictions of the future, it can't predict unusual events. What it can do is give you a broad analysis and an idea of where you need to focus. It can help you find an area where something is happening that doesn't quite make sense, and then after you’ve taken the time to really dig into it, then you can use those tools to hone in on actual predictions and decision-making strategies. —Ben Williams

What you can monitor is changing constantly. One of our customers is a retail store where the data can literally tell you where a person is standing and how long are they spending in front of displays. If you created a kiosk as a major investment and you estimate it takes two minutes to engage, are most of the people spending two minutes to engage with this thing? The sensor technology has come so far that you can record that information and then analyze it, which you couldn't do five years ago. —Swamy Narayanaswamy

An IBM product that we use is called AlchemyAPI. It is a very cheap, inexpensive, API that allows you to put in text data and it gives you a one- or two-word summary of that data, and then you can structure it. So you can feed AlchemyAPI 1,000 tweets and it will say, 300 of these tweets are about technology, 100 are about fashion. It also does things like share what language those tweets are written in. —Jesse Leimgruber

IBM has quite a few predictive analytics solutions which can help you to identify patterns. —Dr. Alex Liu

IBM has quite a few predictive analytics solutions, which are all organized under Watson. You can input all of this data including social media data and you can set API, which can help you to identify patterns and help you to merge this data with your existing data. —Dr. Alex Liu

There are some really simple tools out there like Google Trends. Trend data is something that is a bit more accurate than actual predictions. With Google Trends, you type in any keyword and Google will say, "We have seen that every May this keyword goes up,” so you can make better sense of trends. —Jesse Leimgruber

Think about it like this. Let's say you have 1,000 customers come through the door and 100 actually buy something. What makes those 100 interesting? So you go back and look at their behavior and see what they did that is different from the 900 who didn't make a purchase. Then you can create a learning model, put in place a new rule and see if for the next 1,000 customers, does the prediction come true? —Swamy Narayanaswamy

You need to understand the data and make sure you don’t let the tools make the final decision. —Jesse Leimgruber

You need to understand the data and make sure that you don't let the tools make the final decision. This is huge. I worked with a giant e-commerce site a couple of years back, with millions of dollars in sales, and they looked at Google Analytics to see which marketing channel got them the most revenue. Google Analytics attributes a sale based on which marketing channel sent the last visitor. So if a visitor Googled the company and then clicked on an ad of the company and then clicked on a Facebook ad of the company and made a purchase, Facebook gets all of the credit for the sale even though there were multiple other channels involved in the process. So they said, "We have to double down on Adwords.” Meanwhile, their press was driving all of the awareness. So the tools can be very misleading. —Jesse Leimgruber

What challenges come with identifying data correctly?

It is all about targeting the consumer at the right time in the right place. —Nick Metcalfe

One thing that I have found that has worked is customer journey mapping. If you do a customer journey map, you bring all teams in including service, call center, social, broadcast and service-level. Everyone gets a better picture of "Okay, these are the data points that we would need to connect with you so it makes sense to have service information." This helps you see a bigger picture.

There is a formula I have heard over the years for campaign success, which is a 70-20-10 rule. 70% is the right targeting, 20% is the message, and 10% is to make it pretty. It is all about targeting the consumer at the right time in the right place. —Nick Metcalfe