“Almost everything we do is a recommendation.” That’s the essential design philosophy articulated by then-Netflix engineering director Xavier Amatriain five years ago, where personalizing and customizing choice is the coin of the realm. “I was at eBay last week,“and they told me that 90% of what people buy there comes from search. We’re the opposite. Recommendation is huge, and our search feature is what people do when we’re not able to show them what to watch.”
Unlike search, recommendation systems seek to predict the “rating” or “preference” a user would give to an item, action, or opportunity. Thoughtfully managed, recommendations can prove far more valuable to marketers than for the customers they ostensibly serve. Recommendation engines not only generate useful data for analyzing customer desires; they can be harnessed to make tactical and strategic recommendations for marketers. Think of an enterprise Netflix, Amazon, or Spotify for marketers; the same technology enhancing customer choice now empowers managerial decision. The “Netflixization of analytics” will increasingly shape how serious marketers make up their minds.
That means digital dashboards won’t merely measure and monitor marketing KPIs, but will offer data-driven suggestions, options, and advice, as well. Optimization will defer to recommendation; “the right answer” will matter less than “really good choices.” Gmail, for example, can already draft professional correspondence based on past e-mail exchanges. LinkedIn proactively prompts value-added introductions. Salesforce software computationally qualifies and ranks leads. Calendar managers visually and acoustically suggest scheduling options and priorities. Practically everything a digitally-dependent knowledge worker sees, hears, or swipes can become a recommendation. That makes simple and engaging user experience design for management key.
For Chief Marketing Officers to brand managers alike, the experience of identifying actionable analytic insight will come to resemble binge-watching on Netflix, shopping on Amazon, or swiping left (or right) on Tinder — an exercise in selecting customized and contextualized options determined by data-enriched algorithms.
Analytics-oriented marketing leaderships must recognize that their best people, just like their best customers, want intelligent exposure to meaningful choice. That’s the design goal. Marketer recommendation systems will fuel customer recommendation engines; customer recommendation analytics will inform and inspire marketing’s recommendations. We see here the power and opportunity for network effects.
Empowered by supervised and unsupervised machine learning, that virtuous recommendation cycle will drive customer insight and growth. Marketing’s machine-mediated role will be striking a profitable balance between optimizing recommendations and recommending optimizations. Whether managing sales funnel conversions or customer journey touch-points, marketers will spend less time searching for answers than weighing recommended alternatives. In an “almost everything we do is a recommendation” analytics environment, suggested alternatives could range from segmentation advice to possible hypotheses for A/B and multivariate experimentation. Marketers planning mobile promotional campaigns, for example, would spend less time data mining than exploring Amazon/Netflix-type suggestions, such as: Prospects who clicked on this advertisement also responded to that advertisement; 2-for-the-price-of-1 seasonal promotions attract more repeat purchases from gift buyers than 50% discounts, and so forth. At larger firms with broader product and service offers, enterprise marketing recommendations might be framed as, Brand managers like you seeking to appeal to X customers used Y campaigns and Marketers who launched these kinds of promotions also used those kinds of advertisements.
Many digital marketing agencies develop search engine optimization recommendation engines designed to correlate with user profile themes and characteristics. Their goal is not to sell more, but to learn more about prospects. Recommendation then becomes a catalytic precursor to insight.
Even more provocative, perhaps, are nascent recommenders that suggest marketing hypotheses to test and experiment. These “automated hypothesis recommenders” identify correlations that might merit real-world exploration. For example, at one ecommerce company, data suggested that customers who constantly checked their order arrivals gave higher Net Promoter Score ratings when they saw photos of their purchase, not just estimated delivery dates. The recommender proposed testing whether including purchase photos with delivery notifications would lead to better ratings. (It did, but just for women.)
Marketing recommenders also have roles to play in everyday “explore vs exploit” (EvE) decisions. To wit, what’s more valuable to marketing efforts: running additional experiments or profitably exploiting what one’s just learned? EvE recommendations explicitly highlight trade-offs between optimizing for the moment versus investing in future knowledge.
As marketing KPIs become more sophisticated, understanding those trade-offs becomes more important. Context matters. That’s what makes recommenders so cognitively appealing: they invite agency; they enable choice. In data-rich, artificially intelligent markets, one of the most challenging design decisions organizations make is how best to inform and empower their people. Should analytics emphasize prescriptive, directive, and comprehensive compliance? Or will organizations enjoy better morale and results when technology delivers smarter, better, and bespoke recommendations? These are as much questions of culture and values as business process.
It is no secret or surprise that technological advances will effectively automate many marketing tasks previously handled by humans. That appears inevitable. Consequently, much of the marketing challenge going forward will be determining — recommending — when and where human agency adds more value than machine autonomy. This will come down to managerial discretion.
But make no mistake: at least some of that human agency will be illusory. After all, Netflix, Amazon, Alibaba, and Facebook ruthlessly and relentlessly track which of their recommendations “win” and which ones are ignored. Again, the benefits reaped from consumer analytics map to workplace analytics, as well. Identifying decision biases, for example, becomes easier and clearer. Assessing which recommendations most influence marketing decisions will prove at least as revelatory as determining which ones most impact consumer choice. These insights will be immensely useful for top management.
As technologies that generate data, enhance analytics, and empower choice, recommendation systems enjoy both an excellent reputation and great success. The irony is that marketers have done a better job deploying recommendations for their customers than for themselves. Going forward, the most successful marketing organizations will be ones that find innovative ways of aligning recommendations inside the enterprise and out.
No comments:
Post a Comment