Artificial Intelligence, deep learning, machine learning – whatever you’re doing if you don’t understand it – learn it. Because otherwise you’re going to be a dinosaur within 3 years.
– Mark Suster, “Mark Cuban on Why You Need to Study Artificial Intelligence or You’ll be a Dinosaur in 3 Years,” https://bothsidesofthetable.com/, February 7, 2017
Imagine a new world of Promotions:
Promotions and discounts are one of the most powerful tools to modulate demand. In conjunction with price, they create a compelling call to action.
Machine Learning and Artificial Intelligence can be fully leveraged to explore the power of promotions when adequate data, levers of creative control, and ability to customize and execute them are available. If all three are unavailable, promotions become weak and tired, and never fully result in capturing the promise.
What makes a promotion work might differ as time goes by, but there exist general rules and guidelines to make all of them work. The key learning from analysis of successful and failed promotions reveal that algorithms can play a vital role in ensuring success. The formula for Machine Learning success is usually a variant of the following items:
If we have data, let’s look at data. If all we have are opinions, let’s go with mine.
– Jim Barksdale, former Netscape CEO, https://www.goodreads.com/
The adage “time is everything” is particularly true of promotions. The timing of a promotion and what it is becomes particularly important, not just when a promotion is offered but rather where in the consumer’s journey it is offered. The consumer journey typically comprises awareness, information, inquiry, transaction, purchase consideration, advocacy, and enjoyment. In each stage of this journey the consumer seeks different kinds of things.
A promotion merely promotes the consumer efforts in that part of the journey, that is to say it is not reasonable to offer a price discount when a person is merely seeking information. At the time a consumer is seeking information a promotion enables the consumer to consider things to look for and perhaps features that the consumer has not thought about that the product presents. When a consumer is actively considering purchase it is important to offer not just details of the product or promotion but rather to offer the consumer competing information about other products and also the price discount that the seller wants to offer.
In every stage of the consumer journey, these are the seven elements the corresponding promotions must discuss:
20% of business content will be authored by machines by 2018.
– Heather Pemberton Levy, “Gartner Predicts Our Digital Future,” Gartner, October 6, 2015
Promotional templates grounded in neuroscience typically have the following five-part structure:
Machine Learning algorithms are first used to choose metaphors, and create a small library of metaphors to be served both at the opening and the closing of a promotion. Algorithms similarly create a small library of contexts – occasions, daily activities, life pressures, and so forth that are particularly tuned to the product. These contexts are then to be served with the promotional concept neatly embedded in them. Then a small library of calls to action are created. These calls to action may vary depending on which stage of the path to purchase the promotion is being served. Finally, algorithms serve a metaphor from the library of metaphors to close a promotion.
The use and power of Machine Learning now becomes clear – as data pertaining to the utilization or conversion of a promotion is collected. Now standard techniques (a combination of PCA, clustering, and classification) are applied to the data to determine which opening metaphors, which contexts, promotional concepts, calls to action, and closing metaphors work to maximize conversion. The promotion is optimized in real time or pseudo real time based on the algorithmic learnings in closed loop mode.
Heliograf, The Washington Post’s AI writer, created approximately 850 stories in 2016 during the Rio Olympics and the 2016 presidential election. Humanity’s saving grace? Editing and analysis polishing came from human editors.
– Capterra Business Intelligence Blog, November 28, 2017
Most services today, be they music streaming, entertainment media streaming, online dating, physical fitness, data storage, doctor home visits, language mastery, grocery delivery, massage, or just plain newspaper subscriptions use simple, brutish models to:
All of these models crumble eventually owing to either regulatory enforcement, or eventual customer dismay.
The critical question marketers face, to which there have not been many satisfactory answers, is how best to convert, upgrade, and upsell.
Here are observations from the field.
Context matters very much in the world of converting free to paying. That is to say, create a context that non-consciously communicates that you get what you pay for, and that there is pride in ownership, and there is joy in having all of it, and that the consumer deserves to have it all. In this messaging context, introduce the call to convert from free to paying. Machine learning algorithms parse contexts for conversion parameters – be they imagery, music, words, or content – to determine a “conversion index.” This conversion index is then used to identify the optimal moment to present a call to action to convert free to paying.
Upgrade and upsell happens in the presence of loss aversion, and “crowd advocacy.” Loss aversion is the instinctive cognitive urge that wants to avoid losing “what can be.” However, brutal frontal presentation of “what can be” is usually perceived as very sales-like, and is dismissed. It is therefore vital that loss aversion for the upgrade and crowd advocacy for the upgrade are communicated in a subtle and unobtrusive manner.
That begs the question as to how best to communicate the promotional message in a manner that appeals to the non-conscious. Algorithms identify a library of metaphors that have to do with loss aversion for the product and category in question. Libraries of contexts of daily activities, work pressures, or occasions that showcase loss aversion are subsequently identified. Promotional templates are then created that incorporate elements of each of these libraries. Data pertaining to success of upgrades is collected and then simple clustering of the data reveals the means to optimize the promotion for an upgrade.
The language and neurological codes of promotion are foundationally different based on age, gender, and culture.
Metaphors seek to create extended conversations while limiting themselves to just a few words or a phrase. They become powerful to activate in the structure of a promotion. Machine Learning algorithms extract metaphors and look for their presence and activation in the conscious communication of humans. Classifying metaphors into emergent and dominant ones enables further smart use of metaphors. Use dominant metaphors for a promotion of an established product, and consider using emergent metaphors to either revitalize an existing product, or to promote a new product.
The language of promotion contains the semiotics of the underlying metaphor and the imagery conjured by the metaphor. The metaphor clearly dictates the kind of imagery to be used. Neuroscience then dictates the particulars and nuances of the image, and the juxtaposition of words, numbers, and additional concepts with the image.
Artificial Intelligence algorithms embed the neurological codes into appropriate linguistic context of promotions.
44% of executives believe artificial intelligence’s most important benefit is “automated communications that provide data that can be used to make decisions.”
– “Outlook on Artificial Intelligence in the Enterprise,” Narrative Science, 2018
Today, data and customer loyalty go hand in hand to help drive a company’s brand strategy. Loyalty-data-driven promotions are based on the assumption that it is less expensive to keep existing customers than it is to find new ones.
The information contained in loyalty data offers a huge competitive advantage, as the most loyal customers can be pinpointed, then targeted appropriately with the right offers and promotions to try to make sure they don’t shop somewhere else.
Simplistic use of card data enables targeting an audience with promotional ads for quality products (or product bundles) that specifically suit their individual needs and interests. More sophisticated uses are possible depending on frequency of purchase, basket purchased, and the construction of promotions based on loyalty to the cliched yet useful hierarchy of needs models.
Algorithms can systematically map the purchase of individual items to inferences regarding the core personality of consumers; algorithmically map consumer purchase behavior to the common five-factor model; algorithmically tag each consumer to belong primarily to one of the five factors and secondarily to another; and alternatively, cluster groups of consumers to each of the five factors. That is to say, they can tag features of consumer pools to each of the five factors of the model and design promotions based on loyalty card data in one of the following ways, by:
Algorithmically inferring personality traits from purchase data, and being able to tag consumers to one of the factors in the five-factor model naturally creates the structure of promotions for them.
Promotions typically target value-seeking customers offering one of the following offers:
Note that most offers involve a reduction in price. This reduction in price usually confers self-esteem and a bit of self-actualization for the consumer, according to the well-known hierarchy-of-needs theory. Self-actualization occurs as the promotion encourages spontaneity and a bit of creativity; it continues to occur as it confers confidence and accomplishment by “snagging” a good deal, thereby earning the respect of others and giving bragging rights.
It is the general approach of promotions theories that nothing in a promotion promotes the more than the following basic human needs: physiological comforts that have to do with the clothing, food, water, hygiene, sex, and sleep; safety needs that have to do with security of person, work security, security of property, security of health, and resources; love and belonging needs that have to do with friendship, family, relationships, and intimacy. The hierarchy of needs suggests a linear progression of physiological comforts, safety, love, self-esteem, and self-actualization as the ladder of human needs fulfillment. While it is a rational ladder, human behavior suggests the irrational emotional ladder that places needs further up in the hierarchy as being more important.
Charity is a simple connection that connects the top of the hierarchy to the bottom. That is to say, catering to the physiological comforts and safety needs of others satisfies one’s self-actualization and self-realization needs, and thereby facilitates the fulfillment of one’s need for love. This is why, time and again, neurological studies of promotions that involve charitable acts indicate they always work.
Loyalty programs that are built not on providing discounts to customers, but enabling the satisfaction of the physiological comforts and safety needs of others are very powerful. Loyalty card data reveals which of the physiological and safety needs might be useful in structuring charity-based promotions.
41% of consumers believe AI will make their lives better.
– Christopher Dodge, Strategy Analytics, August 31, 2017
Loyalty card data identifies a few unique aspects of consumers that are underleveraged today. Loyalty data identifies:
Knowing that a lower-income demography seeks a store may be part of the data that supports creating an entire mini store-within-a-store comprised entirely of store brands – devoid of fancy packaging and fancy graphics – thus creating a tremendous value perception. By creating the mini-store within a bigger store, the perceived social stigma of shopping at a “value store” is removed, yet the consumer can gain the value from shopping at a “thrift store.”
Knowing that an affluent demography frequents a store may support similarly creating a mini-store to cater to high-end and luxury items. By just changing layout, textures, lighting, and materials, the perception of luxury is created thereby inviting customers to seek and spend time perusing higher value items. If a category seems to spark consumer interest, it becomes a fertile area for a store to explore with its generic store brands.
How can promotions enable switching a consumer from Brand A to Brand B? Promotions based on price simply become a race to the bottom and rarely win.
Switching happens when the dominant metaphor that a competitor has built their brand on, effectively dovetails into an emergent metaphor that the product is based on. Thereby, the switch is performed in the non-conscious, where the old dominant metaphor is evolved into a new emergent metaphor. This “sleight of hand” happens at the level of a metaphor, and not by touting conscious attributes of the functional excellence of one product over the other.
Context switching happens in a similar fashion. The life pressures, cultural tensions, occasions, locations, and daily activities that have formed the backdrop of the competitor brand are effectively swapped for the product contexts. Not all of the contexts should be switched, however, as maintaining the tie with the old context might be necessary.