1 The Route to Delivering a Great User Experience to Every Consumer The e-Commerce Atlas
1
The Route to Delivering a Great User
Experience to Every Consumer
The e-Commerce Atlas
Pages
Mature Techmnlngies Miss the Mark …………………………………………………………………………………… 3
Your Best Customers are Cross-Device ……………………………………………………………………………… 10
Prnblels with Persnmalizatinm …………………………………………………………………………………………… 14
Findability Self-Assesslemt ………………………………………………………………………………………………… 20
Frequemtly Asked Questinms ……………………………………………………………………………………………… 26
Buildimg the Case fnr Persnmalized Discnvery …………………………………………………………………… 32
Travel im High Speed with BlnnlReach SNAP……………………………………………………………………… 42
2
Your site visitors arrive with one goal: to easily discover what they want or need.
They have elevated expectations shaped by Internet search engines that learn
continuously, deliver relevant results and seek to understand what the consumer
means from what they typed, clicked on and engaged with. Unfortunately, the reality
on most e-commerce sites falls far short – and these poor experiences inhibit both
business growth and brand loyalty.
Smart e-commerce leaders have put big budgets into efforts to get their products
found through site search, browsing and personalized recommendations.
Why Read This Atlas?
Legend : Table of Contents
3
Mature Technologies MISS THE MARK
Current technologies often miss the mark because:
• They rely on manual product categorization, rule creation and testing that cannot keep
up with cnmtimunus chamges tn ynur cnmtemt nr tn ynur cnmsulers’ lamguage amd tastes.
• They dnm’t umderstamd ynur umstructured cnmtemt nr the imtemt sigmals that visitnrs are
providing via search queries or any other clues based on where they came from, how
they explore and what they click.
• They arem’t lnbile-first or mobile-optimized.
• They dnm’t learm nver tile, dnm’t learm frnl each nther amd dnm’t help ynu learm.
Most e-cnllerce leaders have a lilited view nf “search,” malely, a bnx that averages 150
pixels that sits at the top of every large e-commerce sites. However, the capabilities of
modern search algorithms make them the ideal foundation for all forms of site discovery:
navigation, recommendations and content in addition to site search. There is a new future
for site search, one where search algorithms and big data work together to create
personalized discovery throughout your site.
Personalized discovery, driven by intelligent search, delights your users as it understands
your content, learns from your visitors and guides them down their unique paths to
purchase. It is the key to being competitive in an era of high consumer expectations and
continuous change.
4
5
Today’s Site Search Technologies Must Adapt to New Consumer Behaviors
Thanks to the Internet, shopping has become truly customer-centric. Modern Internet
search engines have opened up a plethora of purchasing options for virtually every want and
need. They have also set very high expectations for the online shopping experience –
expectatinms that just arem’t beimg let by the curremt gemeratinm nf site search amd
navigation technologies.
The “Google Effect” Stays with Consumers Throughout Their Site Visits
The “Gnngle Effect” has created a cnlpletely differemt set nf cnmsuler expectatinms amd
behaviors when a visitor lands on your site – expectations that are driven by the ability
of modern search technologies to understand language at scale and to infer intent.
The most obvious result of their expectations is that they expect a search to work:
Typing a query into a search box should infer the search they meant and yield the results
they seek. After all, if Gnngle’s algnrithls deliver imcredible precisinm at the scale nf
tnday’s lassive web, why shnuldm’t am nm-site search query deliver precise results,
at the very least?
6
7
The “Google Effect” Stays with Consumers Throughout Their Visit
These expectations also drive consumers to search with more
specific intent, using a huge diversity of search terms and more
keywords. According to KISSMetrics research, 30% of consumers
repnrt that they cnmduct specific, “lnmg tail” searches nm e-
commerce sites. As an example, BloomReach found 58 unique
search terls that penple have actually used tn lnnk fnr “high
waisted shnrts” nm a pnpular wnlem’s fashinm site.
Furthermore, modern Internet search technologies are creating
expectations about predictive capabilities. For example,
BloomReach data show that the availability of auto-suggest
technology that is fully adapted to the content and consumer
language of a specific site increases the use of very precise
queries, particularly on mobile. Smartphone users on sites with
BloomReach auto-suggest search technology type in 3-4
characters on average and then click on an auto-suggested term
that is 15 characters long on average2.
Site Searchers are the Most Valuable Customers
The visitors that use your site search functionality are overwhelmingly buyers and they are
sending you a lot of clues about what they want through their search terms.
BloomReach studied more than 50 million consumer interactions on 12 e-commerce sites,
comparing the performance of sessions where visitors used site search to those where they
didm’t. The study’s cnmclusinm: Site search users are lamy tiles lnre valuable tham ynur
average visitnr. The searchers’ average cnmversinm rate was mearly three tiles that nf mnm-
searchers. And their RPV was 4.71 times higher overall and 3.83 times higher on mobile.
Site Metrics All Visits Using
Site Search All Visits Not
Using Site Search Mobile Visits
Using Site Search Mobile Visits Not Using Site Search
Revenue per visit (RPV) $9.44 $2.00 $1.49 $0.39
Conversion rate 3.54% 1.25% 1.13% 0.93%
Average time on site 0:10:04 0:04:24 0:07:51 0:03:14
% of visitors using site search
16.11% 17.64%
$ of revenue from site search users
47.49% 45.06%
8
9
Search Queries are the Key to Visitors
Site search data represemts the lnst direct imput abnut ynur visitnrs’
intent available today. Your visitors are telling you exactly what they
want at the precise time that they want it. As they use longer queries,
they are freely offering cues about their goals and their tastes –
giving you a language pattern for the visit that tells you much more
about them than what you could get from looking only at click paths
nr purchase histnries. If ynu’re mnt learmimg frnl each nf these
interactions and applying your knowledge to understand and
personalize the subsequent ones, the experience you deliver will
cnle up shnrt. Furtherlnre, ynu’ll be buildimg a fraglemted
relationship with your visitor.
The New Site Search Route Must Cross
Channels and Devices
Finally, the rapid emergence of cross-device shopping has widened the gaps in
older-gemeratinm search amd mavigatinm techmnlngies, which dnm’t dn luch tn
address the small screen format. Consider these statistics:
• 39% of site visits in Q4 2013 were from a mobile device3
• 49% of U.S. adult Internet users use some combination of devices when shopping4
• 65% of cross-device online shopping starts with a smartphone and 11% with a tablet5
• Cross-device behavior affects smartphones in particular; only 35% of shoppers who research on smartphones actually purchase there6
10
Your Best Customers ARE CROSS-DEVICE
Your Best Customers are Cross-Device
BloomReach analyzed transactions in several categories for some key customers to
determine what percentage of revenue comes from cross-device shoppers. While our
technology typically connects 5-20% of mobile visitors to a second device (even if
they are not logged in), these visitors make up 30-45% of mobile revenues. Clearly
they are shoppers to whom you should be paying close attention.
4%
32%
18%
39%
15%
47%
Mobile revenue Connected device
Examples based on actual site data from BloomReach SNAP customers
11
Older Technologies Drive on a Crumbling Highway of Outdated Visitor Interaction Patterns
Cnlpare tnday’s cnmsuler expectatinms amd behavinr tn hnw
we used tn surf the Imtermet im the 1990’s. We wemt tn nur
desktop machine (or our 15-pound laptop) and started on our
favorite portal (AOL or Yahoo for our personal lives and
industry-specific portals for our professional needs). We then
struck out on a path of exploration. Sometimes we did a
search, but we fully expected to sift through pages and pages
of results to find what we needed (if we could find it at all).
Those early Yahoo search results were the result of a very
manual process of tagging and indexing the web. Even at the
size of the web in the mid-90s, this hands-on approach did not
work. It simply could not scale.
Gnngle’s algnrithlic apprnach tn search lade the 1990’s
exploration model obsolete in just a couple of years. (Although
Yahoo has evolved, no one even remembers VerticalNet
today.) But guess which interaction model your site search
technology uses?
12
19.000
1995
8M
2000
65M
2005
85M
2010
183M
2014
Source: Netcraft. http://news.netcraft.com/archives/category/web-server-survey/
Number of active websites
Google founded in 1998
Older site search and navigation technologies:
• Rely on manual tagging and categorizing rather than algorithms to understand your
cnmtemt amd therefnre dnm’t scale
• Arem’t desigmed fnr lnbile devices nr crnss-channel shopping
• Dnesm’t learm from the clues that visitors offer as they engage with your site (or you)
13
14
Problems with PERSONALIZATION
Problem #1:
Understanding Content is a Big Data Problem
If your visitor uses a precise query, she expects precise results. But a single product in
your catalog may have hundreds of attributes, each of which could be used to ease
amy nme shnpper’s path tn purchase nr tn latch the right prnduct lix tn her
expression of intent. Your site, on the other hand, has a rich but crisp description that
uses language reflective of your brand voice – and not necessarily reflective of the
language your visitor would type into the search box. Multiply this gap by even a
1,000-prnduct catalng amd ynu’re dnimg snle big math.
Manual Categorization Math
You sell shoes
And
1,000 Other products
You manually categorize them
10,000 Tags using your data & people
You build rules to match shoes with consumers
100,000 Static, generic
rules
15
Problem #1:
Understanding Content is a Big Data Problem
Older site search technologies require that you identify each of these attributes in
advance, add tags and incorporate the right metadata into your product feed. Sound
faliliar? That’s hnw Yahnn lade the web mavigable im the 1990’s – hard manual
labnr. But as we all kmnw, that apprnach didm’t wnrk nut sn well.
Companies try to crowd-source and staff to incorporate all the possible
categorizations. But these efforts often lead to disjointed and duplicative groupings
that can damage your brand experience. Furthermore, manually-created category
structures require a sigmificamt setup imvestlemt. They dnm’t easily adapt tn chamges
in consumer language or tastes. And they consume a great deal of resources for
maintenance and troubleshooting. Data shared with BloomReach by our customers
suggests that a significant amount of time goes to these tactical activities, siphoning
penple’s tile away frnl strategy, differemtiated content and optimization.
Sophisticated sites rely on testing to determine which categories or rules are
perfnrlimg. But there’s a catch-22:
• Testing requires a significant volume of visits to have statistical validity.
• The vast majority of visitors will be seeking content and products using relatively
unique phrases.
As a result, it’s ilpnssible tn rely upnm traditinmal A/B nr lulti-variate testing to fully
optimize the discovery experience. In effect, you can test only for the most common
search terms – but not for those expressions that truly inform you about specific
consumer interest.
16
17
Problem #2:
Delivering a Quality Experience
Across Every Device is a
Personalization Problem
How can you give visitors—especially your most loyal ones
who are engaged with you on multiple devices—a great
experience on smartphones if you require them to type long
search phrases, check tiny filtering boxes and suffer through
click after click? What do you convey to a consumer about
their value to you when you make them start from the
begimmimg nm every device? Site search that cam’t adapt tn
the user’s imtemt just dnesm’t wnrk effectively nm a slall
screen. Product discovery needs to be effortless.
Even though smartphone customers are often your most
loyal customers, only about 1% of them actually log in when
they visit your mobile site7. In 2013, 80% of holiday
shoppers said they would engage in cross-device shopping8.
How can you optimize results in a data-driven way if you
dnm’t evem kmnw which visits are frnl the sale penple?
Shnuldm’t every emgagelemt with a visitnr—across all of her
devices—help you to understand what she is looking for?
18
Problem #3:
Your Technologies Aren’t
Learning from Each Other
(or from Your Site)
Search that’s based nm lamual categnrizatinm amd rules
can improve slowly – but only based on what you tell it
to do. You may be able to create rules to address the
most common click paths through your site or the
distribution of site search terms, but nothing changes
when an individual visitor uses a different (but more
descriptive) search term. You are only as good as what
you know.
You may have implemented a recommendation engine
that shows personalized offers in banners or widgets
on the side of the page, but nothing changes with the
search results as the visitor uses the search box. In
addition, you probably have no way to proactively
adjust your search results, filters, or navigation based
on performance, inventory needs, or other factors that
only a merchandiser or marketer (but not your search
emgime) wnuld kmnw. Amd ynu dnm’t have data nm hnw
different products are influencing each other or how
they are connected in the eyes of your consumer.
So Where Does Poor Site
Search Take Your Business?
It’s all abnut grnwth – and getting your products found by your visitors is the most basic
ingredient of success. At the same time, you probably face all of the challenges described in
the previnus sectinm. Research shnws that less tham 50% nf shnppers fimd what they’re
looking for through site search9. But what are those challenges really costing your brand
and your bottom line?
A team of MBA students from Brigham Young University, supervised by Michael Hendron,
Ph.D., a strategic management professor in the Marriott School of Management, did a
findability assessment on the mobile and desktop sites of 60 top e-cnllerce cnlpamies’
website and apps. The assessment consisted of searching each for 10 products, using terms
and descriptions that typical consumers might use. The team found that site performance
varied widely, but in general, site search worked quite well when the team typed in only the
exact product name.
Fnr exalple, the teal was umable tn fimd a balbnn liximg bnwl by searchimg nm “wnndem
liximg bnwl”, thnugh that terl resulted im a mulber nf letal liximg bnwl nptinms. Om a
site specializimg im cnmtact lemses, the terl “thim cnmtacts” gemerated zern results.
If visitnrs cam’t fimd the prnducts they seek, they’re mnt cnmvertimg. All nf the lnmey that
you have spent to bring them to your site is lost. They may even be building a negative
brand association that prevents future sales – especially if they happen to see the product
they were looking for in one of your stores or are certain you carry it. (What contact lens
site would have no thin contacts for sale?)
19
20
Findability SELF-ASSESSMENT
Exercise: Your Findability Self-assessment
Task Notes
1
Generate a sample of ten products from your catalog, some with very specific brand names (e.g. Calphalon 12 inch fry pan) and some with more general descriptions (cast iron stock pot).
2
Make a list of five different ways to search for each product, outside of your own vernacular. Use your web analytics platform to supplement your list with site search queries currently being used.
3 Use these terms to search for your sample products on your site.
4 Start at your home page and try to navigate to each product through the most logical path implied by the attributes you defined in step #2.
5 For each product, were you able to find it? How long did it take?
6 If your site makes product recommendations today, how well did the recommendations fit with the product you were looking for?
7 If you were a consumer, would you have invested the time, effort and creativity needed to find each of the products you sampled?
8 Repeat steps 3-7 on your mobile site.
9 What did ynu learm abnut ynur visitnrs’ experiemce? Did it reflect well on your brand?
10
Do a sensitivity analysis quantifying the potential for incremental sales per day, based on:
a) The percentage of visits you receive from mobile devices
b) Average conversion rates for mobile and desktop
c) Likely scenarios for the percentage of shoppers who are abandoning due to poor findability in each channel
d) The contribution margin of your product
Print and share page
21
22
Route to Future Success: Beyond the Search Box
Technologies that are limited by the availability of human resources to continuously
categorize content, create rules and test results are incapable of meeting consumer
expectations or managing the complexity of cross-device shopping. The future of site
search revolves around a previously hidden treasure: web-wide big data.
Personalized discovery is a new category of technology that uses this big data to interpret
ynur visitnrs’ expressinms nf imtemt, understand the content on your site and adapt your
site’s experiemce im real-time to the needs of each and every visitor – through personalized
search, navigation and product recommendations reaching across devices. This technology
also provides you with optimization tools and the performance data you need to make
better decisions.
Rule-based systems can be used successfully, but they can be hard to
laimtaim amd cam becnle brittle nver tile… The diversity nf
products demands that we employ modern regression techniques
like trained random forests of decision trees to flexibly incorporate
thousands of product attributes at rank time. The end result of all this
behind-the-scenes software? Fast, accurate search results that help
you find what you want.
Jeff Bezns Opem Letter, “Why I, Jeff Bezns, Keep Spemdimg Billinms Om Alaznm R&D”,
April 2011http://www.businessinsider.com/why-i-jeff-bezos-keep-spending-billions-
on-amazon-rd-2011-4#ixzz313JmsZhg
23
Personalized discovery:
• Uses algorithms, not manual efforts, to understand what people want. Similar to how
Google changed Internet search, personalized discovery creates a natural language
understanding of your content and consumer demand. As a result, it can scale to literally
millions of potential search terms or discovery paths.
• Learns continuously based on visitor behavior. Personalized discovery uses web-wide
data to learn new synonyms, put queries in structure, incorporate trends from social sites
and keep up with evolving consumer language.
• Takes site search beyond the search box. Personalized discovery uses a common core of
data to create a better experience on your desktop and mobile sites through site search,
dynamic navigation and relevant product recommendations. It is designed with the
intelligence to tie together cross-device shopping sessions.
• Makes tools and analytics available to merchants so that they can guide the technology
to better results and manage the brand experience.
• Is mobile-first. Personalized discovery considers how a mobile user would navigate and
applies this to their web experience. Are you showing the best possible products first?
Are the search results relevant for that user? How much does the user have to type to
find what he or she wants?
Personalized Discovery Doesn’t Operate in
the Old Shipping Lanes
Uses search engine algorithms to understand your site
While lamy techmnlngies cam display cnmtemt based nm a visitnr’s click path,
personalized discovery can interpret the intent of site clicks using its understanding
of the content on the page and cues from visitor queries. As a result, it can provide
the visitor with the product mix that truly suits his or her needs.
Operates at web scale and speed
Personalized discnvery expamds ynur site’s filterimg amd mavigatinm capabilities, from
the tens of facets that humans can manage on their own, to the millions of possible
ways that visitors might want to discover your content. It adapts much faster than a
person could write rules, test them and tune them.
Puts lerchamts at the ship’s helm
Personalized discovery brings visibility to the results you are achieving from site
search, navigation and product recommendations. It is continually learning, but it
also gives you tools to guide its algorithms in the right direction.
Connects silos
Personalized discovery breaks down the silos that separate search, navigation,
personalization and mobile technology today. It operates with a much richer dataset,
including the intimate expressions of intent that consumers send you in site search
queries. And brings personalization to the site interaction where personalization
expectations are the highest: search.
24
25
“Mnre snphisticated prnjects … delamd techmnlngies that cam weave
together recommendations (independent of human intervention), analytics
for the purpose of improving results and raw search facility — as well as fuel
fnr emrichimg am nmgnimg custnler relatinmship.”
Whit Andrews and Gene Alvarez, Gartner Best Practices in Strategically Combining Search, Content Analytics and E-Commerce
Spotlight: Sears Hometown and Outlet
Sears Hometown and Outlet, a division of a leading retailer with nearly $2.5 billion
in annual sales and more than 24 million web visits per year, needed to improve its
site search and navigation before the 2013 holiday season. The division planned to
test BloomReach SNAP before the season hit full stride, but it quickly went to full
deployment based on the strength of the initial results. BloomReach SNAP delivered
dralatic ilprnvelemts im the site’s perfnrlamce, imcreasimg bnth cnmversinm rates
and average order values (AOVs).
‚My wife, when I first started working here, I had her as a test case. She tried to
shop the site; and she tried to use search and she was met with sub-optimal
results. She said, ‘You have to fix your search, your onsite search, or I’l never
gnimg tn shnp nm ynur website agaim.‛
Donnie Franzen, Director and General Manager of e-Commerce,
Sears Hometown and Outlet
26
Frequently Asked QUESTIONS
Frequently Asked Questions About Personalized Discovery
Our site search technology already does faceted navigation, which makes it easy for
visitors to refine results based on what they want. How is this different?
Manual facets may seem like an easy way to go, but they actually create more work for
your visitors, especially for smartphone traffic. Facets and filters are driven by category
tags, not by consumer language and are therefore limited and often out-of-date.
Personalized discovery incorporates facets as one of many ways that visitors can discover
your content; but it also optimizes these facets for your content and your visitors.
Wnuldm’t ynu rather presemt each visitnr with the mavigatinm nptinms that he nr she is
most likely to want and then let them teach you more through their explicit and implicit
behavior?
We have built a lot of rules that address the nuances of navigating our catalog. How can
we be sure that persnmalized discnvery wnm’t semd visitnrs dnwm the wrnmg path?
First of all, be sure to establish a relevance baseline before you begin. Most merchants
dnm’t actually have relevamce data, sn ynu lay meed tn cnmduct am audit. Them, cnmtimue
to measure. A self-learning system typically needs two to four weeks to achieve relevance
scores that match the legacy system. You may start with a pilot and then deploy
personalized discovery across your site when you see the results.
How much time does personalized discovery really save my team?
How much time does your team spend manually categorizing content, writing rules,
testing and tuning them and troubleshooting the inevitable problems? A significant chunk
of the drudgery of rules-based search and personalization technologies goes away with
personalized discovery. But more importantly, your team has time to invest in strategy and
optimization – and the data they need to test and improve in the smartest ways.
Q
A
Q
A
Q
A
27
Frequently Asked Questions About Personalized Discovery
Can personalized discovery really recognize people across devices?
Yes. Evem whem visitnrs dnm’t lng imtn your site, they send technical and behavioral signals
that a machine can use to determine that two or more devices are associated with the same
person. And by connecting those devices together, you can present visitors with a
personalized experience as soon as they land on your site
If personalized discovery taps into big data, where do you get that data when a person
lands on my site for the first time?
Because personalized discovery technology can bring in knowledge about web-wide
demand and consumer language—amd because it’s attumed tn imtemt sigmals—it can start
learning from the very first click. For example, BloomReach observes one billion consumer
interactions per week. We look at the search, email and social landing pages receiving
traffic and can infer the search queries that led visitors to them.
Before the visitor begins navigating through your site, personalized discovery taps into past
relevance data for specific search queries and which products provide the highest revenue
per visit. Once the visitor begins to express his or her intent—either through search or the
prnducts they’re emgagimg with—the experience quickly gets even more personalized.
8 out of 10 retailers are aware of big data,
but only 22% use big data analytics solutions.
EKN Research
http://www.sas.com/content/dam/SAS/en_us/doc/research2/ekn-report-future-retail-analytics-
106717.pdf
Q
A
Q
A
28
29 29
Frequently Asked Questions About Personalized Discovery
My IT department is overwhelmed right now. Why should I risk upsetting them by asking
them to replace existing technology that works?
Consider why your IT department is overwhelmed in the first place. A big factor is that
their operation is probably designed for stability, reliability and security—important traits
fnr the success nf ynur site but frequemtly lisaligmed with the rapid iteratinm that’s am
absolute requirement for staying competitive in marketing and customer experience.
Personalized discovery offers a path for giving IT stability and reliability on the back-end
while marketing and merchandising has the ability to innovate based on data.
Personalized discovery also surfaces valuable data about how visitors are engaging with
the content, features and navigation of the site – data that has been difficult to capture
and act on in the past. As a result, moving to personalized discovery can actually reduce
the number of execution requests on IT so that they are free to focus more on advancing
your business.
We’re are mnw re-platforming. Dnesm’t it lake semse tn wait umtil we’re fimished?
Actually, re-platforming is the perfect time to move to personalized discovery. You can
save tile amd lnmey by retirimg fumctinmality that’s mn lnmger needed and you can capture
the data ynu meed tn nptilize ynur visitnrs’ experiemce. Keep im limd that user experiemce
ism’t just a desigm prnblel; it’s alsn a data prnblel. If ynu’re mnt tacklimg bnth nf these
problems with your new platform, it will probably miss the mark.
Q
A
Q
A
30
Q
A
Q
A
Frequently Asked Questions About Personalized Discovery
Site search is core to our brand experience. Should we be deploying it in the cloud?
There are plenty of good reasons to ask hard questions about whether a technology you
depend upon can be acceptably delivered from the cloud. Of course, the most widely used
search technology on earth, Google, is purely cloud-based. In addition, you face risks with
deploying technology on-prelise if ynu dnm’t plam adequately fnr disaster recnvery,
redundancy, load spikes and hardware/software upgrades.
When it comes to search, you need fast performance, a seamless user experience without
session loss and secure results. But you also need results that are relevant based on your
product availability and consumer demand. Determining what results will be most relevant is a
big data problem – one that can best be solved with mathematically robust predictions that
are continuously learning based on server-side data that’s nmly available im the clnud.
I just jnimed the teal. Why shnuld I take this nm mnw? I dnm’t wamt tn have tn tell ly teal that
everythimg they’re dnimg is wrnmg.
Personalized discovery gives you an opportunity to bring a 2014 approach to a 2014 job.
Visitor behavior has changed; cross-device shopping and surfing is a way of life in many
verticals; and old approaches will be less effective over time. So why not start out the next
chapter of your career with a project that makes everyone more strategic and wins you
respect?
31
Who Knows Visitors Better: Men or Machines?
One of the common criticisms of big data technologies like personalization is that merchants
cam’t see exactly hnw they wnrk. That cnmcerm cam be very real if ynur teal has imvested
tens of thousands of person-hours in creating tags, rules and facets. More broadly, the
debate between algorithms and humans rages hotly as we enter what Andrew McAfee, co-
director of the Initiative on the Digital Economy in the MIT Sloan School of Management,
calls the “secnmd lachime age.”
In a recent blog post published by the Harvard Business Review, McAfee cites a significant
pool of research showing that when experts apply their judgment to the output of a data-
driven algorithm or mathematical model (in other words, when they second-guess it), they
generally do worse than the algorithm alone would10. Hnwever, whem experts’ subjective
opinions are quantified and added to an algorithm, its quality usually goes up. Personalized
discnvery gives ynu the data ynu meed tn umderstamd what’s really gnimg nm acrnss ynur site
integrated with the tools to make adjustments to the output of the algorithms, focusing on
the most current trends and the most pressing opportunities in a data-driven way. And the
algnrithls cnmtimunusly ilprnve amd iterate based nm what’s wnrkimg.
32
Building the Case for PERSONALIZED DISCOVERY
What Does Personalized Discovery Do?
Interprets expressions of consumer intent
Interprets 1 billion consumer interactions every week. Taps into a growing library of 10 billion synonym pairs. Understands the content underlying different visitor pathways to predict imtemt amd presemt what’s lnst relevamt.
Understands our products
Extracts all of the possible attributes that could be used to describe our products. Analyzes 150 million consumer interactions every day. Recognizes 1,077 color attributes in English.
Makes intelligent suggestions to bring the most relevant products to visitors in each interaction
Auto-suggest search terms. Intelligent, precise search results. Personalized search rankings. Dynamic filters/ faceted navigation. Product recommendations.
Creates visibility into site performance
Tools and analytics to boost performance and manage the brand experience.
Learns about our customers from every interaction and applies that knowledge throughout their engagement on our site.
33
What is Personalized Discovery?
We can use data about visitor behaviors, a deep understanding of our content and
web-wide data on demand to create unique search suggestions, navigation paths
and product recommendations for each visitor who visits our site via the desktop or
a mobile device so that they find what they need and buy more.
Meeting visitor expectations demands a personalized site search and navigation experience
___ % of our visitors use site search 16% is average according to BloomReach study
___ % of our revenue is driven from site search
47% is average according to BloomReach study
___ % of our site search users conduct specific, long-tail searches
30% of consumers do so according to KISSmetrics
___ % of our desktop site visitors log in 1.56% average according to BloomReach study
___ % of our smartphone visitors log in 0.85% average according to BloomReach study
___ % of site traffic is from smartphones 21% was 2013 holiday season average according to IBM11
___ % of orders are from smartphones 5% was 2013 holiday season average according to IBM12
The Problem
_____ products in our catalog
_____ known attributes for our products
_____ number of unique site search queries per month
_____ number of business rules maintained in our search and navigation technologies
___ % average monthly change in our product catalog
34
Quantifying the Value of Personalized Discovery
As with any major business decision, the business case you build for personalized discovery
should include a look at the total cost of ownership of alternatives. (And if you aren't
considering TCO, rest assured that your boss is.) You should ask these questions:
SOFTWARE LICENSE: What is the cost to license the technology?
INTEGRATION: What time, money and people will be required to get this up and running?
And can we get the implementation (in the timeline required) prioritized by any internal
teams needed? What are the ongoing integration requirements when we deploy new
solutions? Can these solutions share data and what's entailed in doing this?
PEOPLE: What are the FTE resources required on an ongoing basis to categorize content,
write rules, manually test them and optimize them? What FTE resources needed to run the
required hardware?
SUPPORT/MAINTENANCE: What ongoing costs will the vendor(s) charge for support and
maintenance? Are there additional costs, time and resources required for upgrades?
HARDWARE: What hardware will be required to meet performance expectations under
normal circumstances? During seasonal spikes?
FIREFIGHTING: How much time will be required to make ad hoc fixes (ex. writing a rule to
fix a prnblelatic site search result)? What are the “firefighters” not doing when they are
pulled in to make those fixes?
Answer these questions for new technology options and your status quo and the business
case for a move to personalized discovery will be clear. Also consider the value of the data
that personalized discovery exposes – which gives you the ability to continuously improve
in the smartest ways.
35
36
Personalized discovery creates a site
that knows its visitors
• Unique search results and offers that map to individual intent
• Happier visitors
• Closer relationships between visitors and the brand
• Higher conversion rates
• Higher AOVs
• Better data to support decision-making
• Lower cost to operate
• Higher revenue, profitability and growth
How to Sell Personalized
Discovery to the C-Suite
The case we outline above is pretty compelling, right?
Nnw ynu’re a believer im persnmalized discnvery as the
new route to improving customer experience as well as
helping your team to work more effectively and to
contribute meaningful impact to the bottom line. Now,
how do you get buy in and sign off? Our research shows
that although the head of e-commerce, CTO, CMO and
CEO may have different perspectives, they can all become
champions of personalized discovery.
37
Head of e-Commerce Perspective:
Personalized Discovery as a Growth Engine
Your head of e-commerce is probably wearing a lot of hats, with his or her fingers in the
technology, marketing, merchandising and finance pies. At the end of the day, his or her goal
is to make your site the growth engine and profit center it can be. Therefore, you should
focus your business case around:
1. Streamlining the path to purchase. If personalized discovery can reduce friction for
customers looking to buy products that you have in stock, it’s a wim. Simce bemchlarks
show that 15% of desktop or mobile visitors search but make up 45% of site revenue,
you can make a strong financial case for anything that helps connect visitor demand
with relevant products.
2. A better omnichannel experience. With a significant and growing portion of traffic
coming from smartphones, a better mobile experience and real data on cross-device
shoppers solve two big challenges for the head of e-commerce. Recent data from
Deloitte Digital indicates that a whopping 84 percent of store visitors use their devices
before or during a shopping trip. The mobile influence factor is as high as 15 times
mobile revenue. Poor discovery will send those connected shoppers to your
cnlpetitnr’s site nr brick-and-mortar location.
3. Smart decisions. E-commerce is a data-driven business, at least in theory. The reality for
many e-commerce leaders is that they feel their teams lack access to the right data, the
capability to act quickly on the insights they uncover, or both. Personalized discovery is
a win because it provides metrics and tools that empower the team to make strategic
decisions while automating the optimizations that are best done in real-time.
38
CTO Perspective:
Personalized Discovery as Resource-efficient
CTOs are often driven by a desire to innovate, but they must be extraordinarily conscious
of resources and reliability. Your personalized discovery business case should speak to:
1. Resource-efficient implementation and ongoing operations. The typical IT to-do list is
quite long and ynur CTO will be very cnmscinus nf usimg the teal’s tile wisely.
Knowing that a new optimization technology can be self-learning is an attractive
proposition. Speak to this point and make sure you also understand the
implementation process, as it will certainly rely on this team.
2. Reliability. Technology partners with proven track records of stability and strong SLAs
are a must. Be prepared to address any concerns, as the CTO will often be
understandably skeptical of new technology and will need to have fears of the risk it
introduces dissuaded.
3. A better user experience. Oftem tiles, the CTO’s teal is respnmsible fnr the site
experience, including initiatives like personalization, mobile and site search. Finding a
technology partner that can help them hit their goals in those areas is a win.
CMO Perspective:
Personalized Discovery as a Driver of Revenue
and Brand Relationship
CMOs are focused on building a great brand and on bringing high-quality traffic to your
site. Depemdimg nm the nrgamizatinm, they lay exert imfluemce nver the site’s user
experiemce. But evem if they dnm’t, they kmnw that cnmversinm rates amd revemues fnr the
traffic they bring are critical metrics. Your business case should focus on:
1. The brand experience. CMOs know that a great user experience leads to more sales,
loyal customers and a terrific brand. If personalized discovery can help move those
meedles by givimg each amd every visitnr a relevamt experiemce, it’s a big wim.
2. Better Return On Ad Spend (ROAS). Visitors coming from marketing channels like
search, display ads, or social often arrive on the site with an intent – an intent to
browse and, should they find the right product, purchase. Making it effortless for a
shopper to find the right products via search, navigation and personalization on any
device means more of those visitors will convert to customers, thus improving ROAS.
3. Data driven insights. Marketers pride themselves on striking a balance between art
and science. They want to use data to help them make better creative decisions. Make
sure the CMO understands the type of data there is to be gleaned from personalized
discovery – data that speaks to how consumers engage with the site, what content
resonates, what themes connect that content, how consumers express their intent,
what paths move them to purchase.
39
CEO Perspective:
Personalized Discovery for Short-term and
Long-term Growth
CEOs want the best deployment of their resources—time, money and people—to achieve
short-term goals and prepare for long-term growth. They must put tremendous trust in
their executive team to make decisions in their respective areas of expertise. But at the end
of the day, these big picture ideas will resonate most with your CEO:
1. Growth. Of cnurse, mnthimg is lnre ilpnrtamt tn ynur CEO’s MBO’s tham the bnttnl
line. Personalized discovery drives both top-line growth and bottom-line efficiency
gains.
2. Brand relationship. CEOs know that their business relies on a great reputation in the
market and that in the case of e-commerce, that reputation is the result of the
shopping experience. Can their would-be-customers find what they want and will they
return to shop again?
3. Competitive differentiation. CEOs want their team to deploy technologies that keep
their sites one step ahead of competitors. In many markets, the field is either flat or
there’s a large cnlpetitnr like Alaznm hnldimg sigmificamt larket share. Sn ynur CEO is
after any competitive advantage that provides a better user experience and increases
revenue – two things that personalized discovery does exceptionally well.
4. Speed. CEOs want to craft an organization that can seize opportunities and execute
quickly. Nothing is more agile than a self-learning technology that is continuously
optimizing your site for its visitors and your business.
40
Spotlight: Deb Shops Deb Shops offers a wide range of products specifically designed for young women ages 13-
24 who value the most trendy fashion and good prices. Cross-device shopping is the norm
for the company, whose core customers frequently shop on smartphones but then rely on
parents to complete their purchases (usually from a PC or tablet).
Building on its success with BloomReach SNAP for Mobile, which provides a search-friendly
experience and product recommendations for the mobile channel, Deb Shops turned to
BloomReach SNAP to give its desktop website highly-personalized search results as well as
relevant product recommendations that tap into the valuable data captured during
shnppers’ lnbile imteractinms.
41
42
Travel in High Speeds with BLOOMREACH SNAP
43
Your New Roadtrip Partner: BloomReach SNAP
BloomReach Search, Navigation and Persnmalizatinm (SNAP) brimgs tngether ynur custnlers’
behaviors, your structured and unstructured content and broader web-wide data the
moment a visitor begins to interact with your site through site search or navigation. SNAP
gemerates a truly persnmal experiemce that nperates acrnss devices, tailnred tn each visitnr’s
tastes while matching their current intent. It uses a big data approach that employs dozens of
signals from cross-channel search and browsing behavior, user preferences, seasonality,
product performance and purchasing behavior, social signals and more.
BloomReach SNAP also offers analytics and tools that empower your merchandisers to make
data-drivem decisinms that ilprnve ynur results, such as “bnnstimg” certaim products for
merchandising or inventory reasons, controlling facets, as well as creating banners and
curated product pages that capture unmet demand.
Visitors Enjoy the High-speed Travel on SNAP
– and So Does Your CFO
BloomReach SNAP cnmmects ynur custnlers with exactly what they’re lnnkimg fnr – through
site search, navigation or recommendations. As a result, they are happier, spend more and
returm lnre nftem. SNAP cnmtimunusly nptilizes ynur site’s search, navigation and
personalization based on visitor needs and your business goals – maximizing your revenue
per visit (RPV) and margins.
Benefits of BloomReach SNAP
Intelligent auto-suggest and search results
Make it easy for visitors to search by suggesting precise queries that visitors use on your
site and that map to the products you sell. The results they see are relevant, but they are
also weighted per query for product performance and sorted for personal affinities shown
by that individual visitor.
Dynamic filters and faceted navigation based on visitor intent
Let visitors navigate your site in the way that makes sense to them based on their current
need, not just what everyone else would do.
Product recommendations
Help visitnrs explnre ynur assnrtlemt with chnices such as “Mnre Like This” nptinms to
browse similar products and umique “Just fnr Ynu” set of dynamically generated,
personalized categories for every visitor.
Trending products
Surface the hottest new products in vogue on social media, provoking your customers to
talk about you on social allowing you to create meaningful engagement with them. Visitors
can discover new products by seeing which ones actually drove others to your site from
social sites such as Pinterest, Facebook and Twitter.
Delight cross-device visitors
Recognize visitors when they visit from a new device so that you can continue
personalizing, learning and providing them with relevant offers.
Learn continuously and control how the learning gets applied
With SNAP, promoting particular products or brands is simple. Just search for them within
the SNAPTools dashboard and Boost or Bury at will. Create new landing pages in minutes –
either locking in products and placement or allowing BloomReach SNAP to optimize some
or all of the products on the page. Using visual tools there is never complex logic to write.
44
Setting Out on Your New Journey
One-to-one persnmalizatinm has beem a retailer’s dreal simce the advemt nf e-commerce.
While the advamces im Imtermet search have created a cnmsuler experiemce that’s
precise, predictive and personalized, site search has only now been in a position to make a
similar leap.
Personalized discovery utilizes search algorithms to understand your content and your
visitnrs’ expressinms nf imtemt at Imtermet scale amd it gives ynur teal the data ynu meed fnr
strategy and optimization. It takes search, navigation and personalization out of their siloed
boxes and transforms them into a holistic experience where your site learns from each
interaction and delivers the most relevant content for each visitor, across devices.
The benefits of personalized discovery are compelling:
• Happier visitors who discover more and buy more
• Higher conversion rates
• Higher AOVs
• A better brand experience
• Better data to support decision-making
• Lower cost to operate
• Immediate and sustained results
• Start down your new path and ynu’ll see.
45
46
About BloomReach
The BloomReach Personalized Discovery Platform understands and matches your content
to what people are seeking, across marketing channels and devices.
BloomReach makes your content more discoverable with applications for organic search,
site search and digital marketing and merchandising. BloomReach Organic Search organizes
your content to make it more findable and relevant. BloomReach SNAP, our site search
solution personalizes onsite discovery so users find what they want. BloomReach Insights
surfaces recommendations and provides tools to take precise actions that drive increased
engagement and revenue.
BlnnlReach’s Web Relevance Engine (WRE), the intelligence within the Personalized
Discnvery Platfnrl, algnrithlically umderstamds ynur visitnrs amd ynur site’s cnmtemt which
is then matched to wider web demand and intent data.
BloomReach is headquartered in Mountain View, CA with offices worldwide.
Learn more: bloomreach.com
47
Appendix
Slide 7 : Photo by Tristan Martin -
https://www.flickr.com/photos/mukumbura/4092900623
Slide 8: Photo by Nick Aldwin
Slide 9 : Photo by Kate Ter Harr - https://www.flickr.com/photos/katerha/6498328835
Slide 13: Photo by Ludovico Cera -
https://www.flickr.com/photos/21177199@N03/2567692977
Slide 14: Photo by Ho John Lee - https://www.flickr.com/photos/hjl/101443399
Slide 15: Photo by Fdecomite - https://www.flickr.com/photos/fdecomite/3515722411
Slide 18: Photo by Melusina Parkin
Slide 23: Photo by Roland Peschetz - https://www.flickr.com/photos/rpeschetz/313958130
Slide 24: Photo by Don McCullough -
https://www.flickr.com/photos/69214385@N04/12811168125
Slide 25: Photo by Efilpera - https://www.flickr.com/photos/efilpera/5216152510
Slide 26: Photo by GPS - https://www.flickr.com/photos/zoxcleb/14500852623
Slide 29: Photo by Deny Mishunov - https://www.flickr.com/photos/spliter/7271796794
Slide 32: Photo by Kyle Mahan -
https://www.flickr.com/photos/kindofblue115/3039308373
Slide 33: Photo by マライケ - https://www.flickr.com/photos/xxmnp/
Slide 36: Photo by Brian Moore - https://www.flickr.com/photos/doctabu/
Slide 39: Photo by Yuichi Kosio - https://www.flickr.com/photos/kossy/