Affinity Solutions - White Paper - AI in Retail Marketing
1. RETAIL AND AI:
WHAT’S IN STORE FOR
ARTIFICIAL INTELLIGENCE
Predicting future consumption with machine
learning and purchase behavior data
By
Amit Seth
President, Affinity Solutions
2. AI evolves to conquer predictive marketing
Artificial intelligence (AI) is today’s hottest business and tech topic, truly
powering the future; everything from self-driving cars to pizza-making
robots has AI in its DNA. But apart from AI’s highest-profile innovations,
artificial intelligence technology is rapidly enhancing a wide range
of more familiar domains, including marketing. Here, AI is helping to
improve analytic models to predict two types of consumer demand:
• Latent demand, which comprises unmet needs unarticulated by current
consumers (pre-Starbucks, no one was asking to pay $4 for a cup
of coffee), or demand from an unexpected population of consumers
(like Xbox drawing in women and kids via Kinect).1
Harvard Business
Review noted the perplexing nature of latent demand in 2013:
It is possible to quantify latent demand, but the real challenge is
estimating when the dam will break. Investing in latent demand
requires as much faith as it does fact, as there is a risk of being
perceived as over-investing if the upside doesn’t materialize soon.1
• Active demand, which is existing demand for currently available
products. As described below, many methods are used to forecast
active demand, but all are plagued by two shortcomings: they have
high levels of latency, and they extrapolate large-scale demand from
a small data subset.
Predictive analytics have been used
for decades to predict all manner of
consumer financial behavior, including
payment card fraud. Here, the fraud
detection technology answers the
question, “Based on this individual’s past
purchasing behavior, and the purchasing
behavior of similar people, is it likely
that the person we think is making this
transaction is actually doing it?”
1 “Is Latent Demand Netflix’s Secret Weapon?” Eddie Yoon, Harvard Business Review, January 28, 2013.
https://hbr.org/2013/01/is-latent-demand-netflixs-secr
Using past purchase
behavior to predict
future consumption
affinitysolutions
1
3. In both instances, purchase behavior data is central to identifying
demand. By applying AI to the massive amounts of retail and purchase
behavior data available today, marketers can break free of outdated
research constructs that are no longer sufficient in meeting consumer
expectations for personalized offers and experiences.
AI is replacing decades-old forecasting constructs
Artificial intelligence techniques, including machine learning, can predict
consumer behavior at extremely granular, individual levels, obviating
decades of reliance on:
• Market segmentation vs. marketing to individuals:
Markets are traditionally broken into demographic segments, such as
women ages 18-34. In today’s exceptionally crowded marketplace,
in which every brand competes for both mindshare and wallet-share,
this generic approach in no longer sufficient. Sophisticated consumers
expect personalized experiences and products targeted at their
individual needs––not broad offers directed to an enormous slice of
the general population.
• Panel vs. census: Marketing panels are small segments chosen
to represent a larger population. For decades, for example, the
television viewing habits of a tiny subset of American families
drove the entire TV entertainment industry. It’s impossible to capture
individual behavioral nuances through panel marketing, which
contrasts with census marketing, a method that captures information
from all participants in the population.
The availability of AI and large amounts
of fresh data has the potential to solve
many voids in the retail arena, allowing
marketers to predict future purchases
with pinpoint accuracy.
Today, marketers
can have direct or
indirect access to
limitless amounts of
data,including past
purchase data, and
use AI to answer an
even more refined
question: “Based on
this individual’s past
purchasing behavior,
what can we learn,
dynamically, about
what he or she may
buy in the future?”
2
4. AI can help to
determine future
buying propensities,
allowing marketers to
target their offers with
fine precision.
In these ways, understanding past
purchase behavior, and applying
artificial intelligence, can predict both
latent and active demand far better than
traditional methods.
One-off purchases such as buying a new
computer or television are often followed
by new accessories such as a printer or
surround-sound speakers, respectively.
Marketing to a “segment of one”
Armed with this insight, marketers can create a “segment of one” and
deliver offers that:
• Are highly accurate: Consumers’ online behavior can indicate they’re
in the market for a purchase, such as a new TV or sunglasses.
Ad retargeting can be a useful reminder of purchases that were
considered, but what about ads that continue to be shown after a
purchase is made? Such promotion is not only inaccurate, but it’s
also annoying. The use of fresh purchase data in analytic models
can indicate when the item has been bought and promotions are no
longer appropriate.
• Reflect customers’ individual preferences: As an example, consumers
increasingly watch entertainment content online, on sites like Netflix
and Hulu. AI can help marketers to get to know consumers and their
preferences at an individual level, such as, “It’s OK to market to me at
the beginning of a movie or show, but not during.”
• Target future buying propensities: Purchase data shows seasonal
or habitual buying behavior that can be used as the entry point
for complementary offers. For example, annual trips to specific
destinations, such as visiting family abroad, usually entail purchases
such as airfare, ground transportation, tourist activities and gifts.
The Holy Grail: Tying together online and offline purchase behaviors
Most retailers have a firm grasp on what’s sold within their own four walls. But what about purchases that
customers make at other online and offline retailers, and their ability to predict future buying behaviors?
A large home improvement retailer recently leveraged external purchase behavior data, in combination with
its own in-store and online data, to identify customers considering home remodels, and increase its share of
wallet with them.
3
5. A home improvement retailer
captures more revenues
The home improvement retailer had in-store and online purchasing
data captured from tens of millions of active customers. It wanted
to target customers planning a significant renovation in the
immediate future by understanding what those customers were
spending at the retailer relative to its competitors.
It also sought to gain insights to identify which incentives would
be relevant, depending on whether a customer was already pre-
disposed to shop at the home improvement or more likely to shop
its competitors, that latter of which might suggest a higher-value
incentive.
Finally, the home improvement retailer wanted to develop
a reliable, verifiable method for measuring sales lift from its
advertising efforts, and develop insights to more effectively re-
target and up-sell additional products and services.
Working with an external partner providing purchase data
analytics, the home retailer built a custom analytic model designed
to identify customers in the market for a major home renovation.
The purchase data analytics provider’s ability to track and analyze
tens of billions of debit and credit card transactions annually, from
tens of millions of consumers, provided a unique and exclusive
view of consumer spend behavior. 4
6. The solution allowed the home improvement retailer to leverage the
purchase data analytics provider’s core assets:
• Third-party credit and debit card transaction data
• Purchase data down to the SKU level
• Machine learning models that derive actionable insights from that data
• A technology platform that operates on those toolsets and models
at scale, to more accurately predict the near-term purchase intent of
customers planning a home renovation
A full 88% of the renovators identified by the purchase data analytics
provider were found to be dual shoppers, dividing their share of
wallet between the home improvement retailer and competitors,
demonstrating a significant opportunity to drive share shift. Among
those dual shoppers, the home improvement retailer had a 58% share
of category spend.
Impressively, the home improvement
retailer experienced a 12x lift
from the purchase data analytics provider’s
predictive model, relative to the control
group, and saw an increase in sales (post
vs. control) of up to 87% relative to the
control group.
The upshot: By effectively targeting “dual
shoppers,” shifting just 1% of these customers
back to the home improvement retailer would
drive $5 million in incremental revenue.
LIFT12 X
Control Group Target Group
5
7. Technologies like
AI-enhanced purchase
data analytics are
today empowering
marketers to do
exactly that, to create
a “segment of one”
unique to every
customer.
How the science of AI
advances the art of marketing
Artificial intelligence is taking data-driven marketing to a higher level,
providing predictive insight into individual behaviors with:
• Self-learning and continuously updating models: AI incorporates
self-learning analytic models that continuously update and
improve over time, as more data is ingested. In this way,
marketers can expect to improve their understanding of
customer behaviors, toward the goal of addressing each
individual as a “segment of one.” Also, self-learning predictive
analytic models can also be deployed when there is no existing
data set, such as for the launch of a new product or retailer.
• The ability to predict future demand: In a November 2016 article
in Harvard Business Review, Amit Sharma wrote: Retail giants
have been using machine-learning algorithms to forecast demand
and set prices for years. Amazon patented predictive stocking in
2014, and saying that AI, machine learning, and personalization
technologies have improved since then is an understatement.
Retailers need to think more like tech companies,
6
8. using AI and machine learning not just to predict how to stock
stores and staff shifts but also to dynamically recommend products
and set prices that appeal to individual consumers.2
Technologies like AI-enhanced purchase data analytics are today
empowering marketers to do exactly that, to create a “segment of
one” unique to every customer.
• An entry point to the data and analytics boom: It’s a fact that the
predictive value of AI and analytics models correlates directly with
the quantity and freshness of the data used within. The availability
of limitless cloud-based compute, and storage resource makes
it easy and cost-effective to combine vast amounts of data from
internal and external resources—everything from social media likes
and dislikes to weather data—into marketing predictive analytics,
driving further value.
Affinity Solutions delivers
AI solutions to retailers today
On the retail analytics front, numerous initiatives are now looking
at how artificial intelligence techniques can replicate the human
behavior to predict future purchases.
Affinity Solutions Buyer Graph provides marketers with far greater
visibility into which individuals are likely to make a relevant
purchase, as well as when the purchase may be made.
Affinity Solutions Buyer Graph taps the power of historical purchase
activity and advanced analytics, producing predictions that are far more
accurate, and which consistently correlate with actual purchase outcomes.
2 “How Predictive AI Will Change Shopping,” Amit Sharma, Harvard Business Review,
November 18, 2016. Emphasis added. https://hbr.org/2016/11/how-predictive-ai-will-change-shopping
Amazon patented
predictive stocking in
2014, and saying that
AI, machine learning,
and personalization
technologies have
improved since then is
an understatement.
7
9. Affinity Solutions has developed 63 predefined product categories
and more than 400 named brands that represent the most widely
used segmentation for audience targeting and can generate custom
segments to address specific marketing objectives. It presents a
highly intelligent way for marketers to engage and influence likely
buyers at the right time in the buying cycle, and convert them into
loyal customers.
Affinity Solutions Buyer Graph is powered by the
Purchase-Driven Marketing Cloud, which leverages leading-
edge machine learning algorithms to support real-time pattern
recognition, new customer “buy signals,” marketing intelligence and
CRM applications.
This intelligence and analysis, together with purchase data,
empowers marketers to make more informed and better marketing
decisions that deliver high-impact business outcomes.
Affinity Solutions Buyer Graph
is unlike other data analytics
solutions, which are focused
on past customer behavior;
it can focus on what is most
likely to happen in the near-term
by continuously discovering
individuals purchase patterns and
behaviors, in real-time.
RETAIL
Accessories
Jewelry
Footwear
Men's Apparel Stores
Women's Apparel Stores
Discount Stores
Luxury Department Stores
Children's Apparel & Toys
Apparel Stores
Department Stores
Sporting Goods
Optical
MASS RETAILERS
Mass Merchant
Digital Retail & Services
TRAVEL
Travel Services
Luxury Hotels
Mid-tier Hotels
Economy Hotels
Cruise
Airlines
Auto Rentals
FOOD & PHARMACY
Grocery
Convenience Stores
Vitamin Shop
Drugstore
Club Warehouses
RESTAURANT
Fine Dining
Casual Restaurants
Quick Serve
HOME
Home Furnishing
Home Improvement
AUTOMOTIVE
Auto Parts
Gas Stations
CONSUMER SERVICES
Tax Tools & Services
Insurance
Logistics
Cable, Satellite, Other
Investment Institutions
Wireless Carriers
Health Clubs
SPECIALTY RETAILERS
Florists
Pets
Arts & Craft
Beauty/Cosmetics
OFFICE,
ELECTRONICS, GAMES
Video/PC Game Stores
Consumer Electronics
Office Supplies
ENTERTAINMENT
Movies
Amusement Parks
Tickets
Streaming Music
& Video
8
10. Affinity Solutions Buyer Graph, along with Spend Insights and
Closed Loop Measurement, together form a comprehensive suite
of purchase-driven marketing solutions that no longer just inform,
but deliver accurate predictions, allowing marketers to create the
messaging and offers required to drive revenues, without guesswork.
PRECISION
MARKETING
CAMPAIGNS
CLOSED LOOP
MEASUREMENT
BUYER
GRAPH
SPEND
INSIGHTS
Unparalleled
marketing
intelligence turns
likely buyers into
customers
Precision
marketing
campaigns leads
to higher
proven ROI
Optimize
marketing
spend to generate
the highest
proven ROI
Better visibility
leads to high
impact decisions
Affinity Solutions provides a
powerful suite of purchase-
driven marketing solutions
that you can leverage
to identify new market
opportunities, shift share in
a competitive market, and
optimize marketing spend to
generate profitable revenue
outcomes.
About Affinity Solutions
Affinity Solutions makes all marketing more productive by driving
significantly greater business outcomes for marketers using the power
of purchase data and analytics. Through our partnerships with over
4,000 financial institutions, Affinity Solutions has real-time and
secure access to where and when consumers are spending across
brands, categories, geographies, and channels.
Key intelligence is gleaned from the continuous, real-time discovery
of an individual’s purchase patterns to identify, reach, and influence
likely buyers at the right moments of choice; as well as to measure and
optimize the sales impact of marketing campaigns. This gives marketers
the ability to deliver unique and engaging experiences throughout the
customer journey––from discovery to purchase to loyalty––to maximize
and increase customer lifetime engagement and value.
Affinity Solutions
New York City HQ Office:
1180 Avenue Of The Americas, 3rd Floor
New York, New York 10036
Phone: 212.822.9600
United States
info@affinitysolutions.com
More at www.affinitysolutions.com
9