Email Marketing is a Predictive Analytics Problem
Email Marketing is a Predictive Analytics Problem
Join the DZone community and get the full member experience.Join For Free
Access NoSQL and Big Data through SQL using standard drivers (ODBC, JDBC, ADO.NET). Free Download
In his book Permission Marketing, Seth Godin referred to email marketing as “the most personal advertising medium in history”. That was 1999.
Where does email marketing stand in 2012 in the age of social media, omni-channel marketing and big data analytics? Here are some interesting data points.
A recent Forrester Research report posed a multiple-choice survey question – “What are the biggest email marketing challenges you expect to face in the next two years?” The top five responses were:
- Integrating email with other channels - 57%
- Leveraging dynamic content - 50%
- Leveraging segmentation - 43%
- Managing email frequency and cadence 38%
- Increasing open and click-through rates 38%
A common theme across these questions is that each of these responses relates to making decisions such as the right marketing channel with the right content for the right audience at the right time. Making the right choices is a daunting task for today’s digital marketer because of the large number of permutations and combinations – similar to finding needles in haystacks.
Email marketing effectiveness is a big data analytics problem and it has not been solved. Yet.
Occam’s Razor, a blog by Avinash Kaushik states the email problem in simple terms as follows:
“One of the core challenges with email is that you have to deal with multiple data sources. There are three primary sources:
1. Your campaign data. How many emails went, to whom, what happened to them, yada, yada, yada.
2. Your website data. What happened after someone clicked on your email links?
3. Your company cross-channel outcomes data. Multi-channel customer purchase behavior, customer lifetime value. You know, Analysis Ninja territory!”
Predicting Intent to Buy – Monthly, Daily, Hourly
Predictive analytics provides a measure of propensity based on statistically significant correlations. In the past statisticians and marketing experts have successfully used data analytics as a basis for testing marketing conjectures and developing campaign rules. A typical campaign can consist of hundreds of rules to choose the right combination of content, channels and timing to maximize the effectiveness of a marketing campaign.
What if we could have a few hundred thousand rules applied to target “nano-segments” of consumers with marketing campaigns optimized by the month, week and hour? Real-time precision targeting of consumer is considered a utopian goal for the marketer and often limited by the paucity of data and the ability to get meaningful insights into the data. The availability of technologies such as Hadoop has made advanced analytics more accessible. Machine learning can capture thousands of attributes of consumers including temporal and geo-spatial which have a better chance of predicting behavior. This approach can produce marketing rules on a much bigger scale and precision than what is possible by human beings.
There is a fundamental difference in the output from machine learning and traditional modeling. Machine learning models use self-learning systems that grow smarter over time with increasing prediction power. Traditional predictive models have a half-life and lose their predictive ability if they are not periodically refreshed. The ideal predictive model takes advantage of traditional statistical techniques as well as big data.
This combination of machine learning, micro-segmentation and analytics creates a new set of possibilities for marketing managers who are looking at integrating email with other marketing channels.
InsightsOne is one of the pioneers in the application of big data analytics to email marketing. Their multi-channel marketing solution generates a real-time propensity matrix for consumers for every product in an email campaign based on data gathered from multiple channels such as transaction data, web logs and call center records. A recommendation engine produces scores that can be used to target offers through traditional outbound mechanisms such as email, web and call centers.
According to Waqar Hasan, Founder and CEO of InsightsOne, “Companies using advanced analytics on big data have consistently seen both revenue and margin increase over 10% while decreasing customer fatigue.”
Email, Social Media and Online Consumers
An interesting report entitled The Collaborative Future by ExactTarget and CoTweet classifies online consumers into three categories:
- Subscribers of email. 96% of these consumers subscribe to at least one permission-based email daily from a brand.
- Fans of Facebook. 69% of daily Facebook users are a fan of at least one brand.
- Followers on Twitter. 68% of daily Twitter users follow at least one brand.
The report indicates that email is the best channel for reaching and retention of customers but ranks low for customer acquisition. Facebook and Twitter on other hand are far superior for customer acquisition and rank lower on reach and retention. It further recommends that the key to success on email is targeted, exclusive content. The report concludes with, “Despite common trends and helpful tips, there’s no magic formula to predict how and when consumers interact the way they do.”
Marketers will benefit from looking at ways to optimize investment in all channels of interaction with consumers that are most effective in each medium and personalized to each consumer. Personalization of relevant content has always ranked high in consumers minds. However, three things have changed since personalized technologies were introduced in the 90’s.
- Focused Content. The sheer volume of information has grown exponentially while the human capacity to absorb it has not. This means any personalization solution has to be much more focused and relevant from all touch points of direct marketing.
- Time and Place. Consumer behavior is strongly influenced by time and place and circumstances especially for online mobile consumers. On-line offers that are personalized in real time and situations will trump those that are not.
- Technology Leverage. Advanced analytics and machine learning technologies available in the last five years can provide an advantage to early adopters.
Published at DZone with permission of Ravi Kalakota , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.