Big data has grabbed headlines primarily because of its quantity and complexity. But what often gets lost in the discussion is the nature of speed. Not only do instant-gratification consumers today want responses in real time; the sheer mass of data also requires speedy processing so companies can do something useful with it. It’s no use getting a great piece of insight after the customer has walked out the door.
Algorithmic marketing is already starting to solve that speed-to-market conundrum. Employing advanced analytical methods, algorithmic marketing provides real-time offers targeted to individual customers through a “self-learning” process to optimize those interactions over time. That can include predictive statistics, machine learning, and natural language text mining. It harnesses big data such as customer location and behavioral information along with powerful computing systems to match customers with context-sensitive products and services.
To go algorithmic, companies need to move from batch systems (where work is done at regular intervals) to algorithmic systems (real-time updates). The way a batch system works, for example, is a retailer tracks keywords on a spreadsheet and uploads them once a week or once a day. Algorithmic marketing, however, tracks keywords automatically and makes updates every 15 seconds based on changing search terms, ad costs, customer behavior, etc. It can make price changes on the fly across thousands of products based on customer behavior, price comparisons, inventory, and predictive analysis.
Algorithmic profits
Algorithmic marketing is allowing companies to do things they couldn’t do before, and some early signs show it can deliver big value, especially in financial or information services.
In North America, Amazon.com grew 30 to 40 percent, quarter after quarter, throughout the United States’ 2008-2012 recession, while other major retailers shrank or went out of business. From 2006 to 2010, Amazon spent 5.6 percent of its sales revenue on IT, while rivals Target and Best Buy spent 1.3 percent and 0.5 percent, respectively. That investment and focus has yielded increasingly sophisticated recommendation engines that deliver over 35 percent of all sales, an automated e-mail/customer service systems (90 percent are automated, versus 44 percent for the average retailer) that are a key component of its best-in-class customer satisfaction, and dynamic pricing systems that crawl the Web and react to competitor pricing and stock levels by altering prices on Amazon.com, in some cases every 15 seconds.
Another example is a large Latin American bank which transformed itself from a little known player to an institution that, by 2010, ranked 11th worldwide in market capitalization. All offers are delivered to customers in a personalized way, based on an understanding of their preferences. In addition, information received via one channel is used to inform and update intelligence across the system in real time. For instance, if a customer rejects an offer on an ATM, the “next product to buy” (NPTB) engine is updated to ensure that the customer’s next interaction with the call-center results in a different, more suitable offer. It also used its capabilities to stay far ahead on straight-through processing (STP) across channels. ATMs are capable of 190 different transactions, which cover key sales types, such as fixed deposit creation, personal loans, credit cards transactions, loans against pensions, and simple life and accident insurance offers.
Telecommunications companies that have traditionally used data mining techniques are also now pursuing the benefits of algorithmic marketing. For example, a top-three Asian telecommunications company with over 100 million subscribers established automated, real-time churn triggers that create tailored and progressively more aggressive offers in order to retain customers in an environment of rapidly diminishing loyalty.
Getting into that algorithm rhythm
Invest in tech: Making the shift from batch to algorithmic is like going from the age of propeller flight to jet engines. And the implications are just as momentous. To be real time, companies need very different system architectures, investment programs, programing, and security protocols. In-memory processing, an emerging technology that gives users immediate access to the right information for more informed decisions, can’t be found in off-the-shelf packages. Throw in all the security issues in protecting all those data and money transfers and you’re talking about a sophisticated and complex system. Getting these IT systems working is the greatest challenge CMOs face, according to a poll we recently ran on the topic.
Algorithmic marketing requires custom programs or heavily modified packages because the nature of a company’s business, organization, and processes tend to be unique. Firms must, therefore, invest in creating integrated information systems that not only transcend organizational silos but also tie into systems operated by suppliers and partners.
New organizational capabilities: Algorithmic marketing is a wasted opportunity unless companies resolve how it fits in with current organizations and processes. For example, companies need to understand how a merchandizing manager negotiates deals with a dynamic pricing system running in parallel. Or when a mobile customer walks into a store after receiving an offer via SMS, the sales person needs that offer and customer information to understand not only how to fulfill the order but also how to cross- and upsell.
Other opportunities require new teams. For example, big data and intelligent, self-learning algorithms offer the potential for companies to conduct controlled real-world testing. But how to enable, manage, and make sense of potentially thousands of these experiments, while also using time, resources, and money efficiently. Organizations need to set up and manage “test factories” with streamlined IT systems, robust underlying processes, and automated results interpretation tools, which would allow managers to test several thousand ideas every month.
Winning the war for digital talent: Google, Facebook, Amazon have made recruiting talent a top priority. Amazon employs 25 PhDs to manage and conduct tests to optimize its site. Google, which pays its engineers more than its MBAs, provides engineers with leadership roles and well-defined career paths. One retailer we know spent three years trying to attract the right talent but ended up only hiring two people. That’s because they weren’t willing to build its HR strategy—compensation packages, career paths, and leadership opportunities—geared around the data engineers that they need for algorithmic marketing.
Algorithmic marketing isn’t really a choice. Leading companies are already doing it and seeing tremendous growth. Companies need to develop their own algorithmic capabilities or risk being locked out of the next wave of growth.