APC Forum Poised Between A Wild West of Predictive Analytics

APC Forum: Poised Between ‘a Wild West of
Predictive Analytics’ and ‘an Analytics of

Things Westworld Frontier’
The convergence of advancements in big data, analytics, the Internet of Things, digital
transformation and artificial intelligence will transition organizations from their
current wild west of predictive analytics to an environment that begins to resemble
the American science fiction domain of Westworld. In this new domain, referred to
here as an era of the ‘Analytics of Things,’ (AoT) business to business relationships will
change dramatically and dynamically. New types of data and analytics contracts will
be needed, and academic research on preferred negotiation strategies to realize pre-
ferred contract scenarios is required. This collaborative effort with the SIM Advanced
Practices Council establishes early thought leadership for the emerging Analytics of
Things. Findings from this research provide recommendations on contract provisions
in different scenarios, and they inform AoT ecosystem partners on issues including, 1)
infrastructure capacity and communication channel costs can increase exponentially
when data and analytics are shared among partners, 2) early entrant data co-owner-
ship advantages exist when shared data remains shared after a contract ends, 3) well-
designed garbage collection processes for a partnering ecosystem are needed when
data does not remain shared after contracts end, 4) when partners share analytical
models and data ownership, there needs to be adequate communications capacity to
handle expected update and data transfer volatility and 5) democratization of data
and analytics ownership in an ecosystem with many partners comes at a higher cost
in infrastructure and communication loads than if data and analytics remain propri-
etary.
Michael Goul
Arizona State University (U.S.)
Executive Summary
Most CIOs are deeply engaged in leading their organization’s digital transformation strategy
while concomitantly minimizing business disruptions, promoting continuous innovation,
establishing big data and predictive analytics governance, and dealing with increasing security
risks. Even with all of those responsibilities, the relentless pace of technological change is
bringing new Internet of Things (IoT) and artificial intelligence applications and solutions into
vogue. Many now refer to a forthcoming Analytics of Things (AoT) as a convergence of IoT,
analytics, artificial intelligence, and digital transformation. AoT means more data, more and
December 2018 (17:4) | MIS Quarterly Executive 333

APC Forum: Poised Between Wild West of Predictive Analytics and Analytics of Things Westworld Frontier
faster analytics, machine learning at the edge, governance and the organizational shifts that
more complex infrastructures, and heightened must be addressed.2
analytics-based competition. AIM Suite’s digital signage observes people
Since AoT promises major disruption and watching streaming media created for a video
innovation, doesn’t it seem that proprietary, display. The suite uses predictive models to
secretive AoT strategies are best to foster discern watcher demographics from its camera
an organization’s next competitive edge? images. For example, it can detect a watcher’s
Surprisingly, a recent MIT study links data and gender, record the length of time the video is
analytics sharing to successful organizational watched (dwell time), and even assess whether
innovation strategies. Knowing what to share watchers are related. The AIM Suite likely
and with which partners will soon become a new conjures remembrances of the movie “Minority
purview of the CIO. And what is the glue that will Report,” in which Tom Cruise’s character is
hold partners together when it comes to data and digitally recognized as he walks through a
analytics sharing? Contracts. futuristic mall and receives personalized ads.
This report reviews AoT convergence with Suppose that kiosks equipped with Intel
an eye towards a new lens, referred to as an AIM Suite-like technology operate all around a
“assemblage,” that amalgamates AoT capabilities shopping mall.3 Figure 1 depicts the operations
into meaningful business bundles. CIOs will of such a kiosk. Because it captures watcher data,
be involved in developing products that evolve uses predictive analytics, and tailors content to
throughout their useful lifecycle due to embedded a recognized watcher’s demographics, such a
AoT capabilities. The assemblage, a collection kiosk basically represents a souped-up version
of interdependent devices, communication of what today’s sensors are capable of in the
protocols, predictive models, and machine IoT world. In other words, when Moore’s Law
learning capabilities, is viewed as akin to an AoT delivers more and more computing power to the
bundle. Based on an examination of contract edge, then today’s simple sensors will transform
provisions that capture assemblage notions, to be full-blown computers that can capture
guidelines are provided on negotiation strategies data, apply predictive analytics, use all means to
in particular scenarios.1 For organizations that engage a customer, leverage customer scoring
are considering hosting AoT platforms, findings analytics to offer coupons, transmit data to their
from this research suggest that particular owner, and even change their own behavior
contract provisions will increase assemblage when new intelligence is sent to them or they
infrastructure costs. For organizations that learn from their audience. Given this almost
anticipate deploying AoT assemblages in certain technological trajectory, organizations
infrastructures owned by others (e.g., Smart striving to get a handle on governing big data and
Cities, Smart Stadiums, etc.), findings from this analytics today will soon face even more difficult
research lead to recommendations on contract challenges with AoT.
provisions to maximize data and analytics In this AIM Suite, subsets of data are co-owned
sharing. by the kiosk owner and the media provider
whose video is displayed while demographics
Introduction and coupon scoring takes place. A predictive
The convergence of today’s Internet of model that issues a coupon might be co-owned
Things (IoT) business applications with ongoing by the kiosk owner and the media provider. Data
advancements in big data, artificial intelligence, streams collected at the kiosks could be shared
and analytics can be described as the era of between a mall owner who has contributed to the
“Analytics of Things” (AoT). Intel’s AIM Suite purchase or rent of the kiosk. Contracts between
foreshadows both the promise of AoT to innovate the participants might have provisions regarding
processes, products, services, and overall 2 Intel AIM Suite anonymous video analytics in digital signage,
business models as well as the challenges in published to youtube.com on January 9, 2013, https://www.youtube.
com/watch?v=Rp3Ltlcyo3g.
1 Goul, M., Mishra,V. and Dnyanmothe, D., “Organizational Data 3 Next generation retail & digital signage solutions, published
and Analytics Contracting in Smart City Fog Platforms,” Interna- to youtube.com on September 1, 2016, https://www.youtube.com/
tional Conference on System Sciences, 1/3/2018, 5035-5044. watch?v=8ehDRvSFjaQ.
334 MIS Quarterly Executive | December 2018 (17:4) misqe.org | © 2018 University of Minnesota
Figure 1: Channel by Channel Depiction of a Kiosk’s Behaviors
A Different View Time    
Viewer ID
Young Adult
Cell Phone
shared profit if coupons are exercised leading complex data and analytics ownership contracts,
to a sale by a co-owner. Moreover, kiosk owners complex profit and fee-for-data service contracts,
may possess a large ecosystem of networked and contracts with many different types of media
deployments across an entire geographical region providers. There may also be significant volatility
whereby there are many kiosk renters, many
AIM Suite AoT Mall Illustration
The Intel AIM Suite has a series of channels operating in real time. On channel 1, a subset of the kiosk’s predictive
analytics might be leveraged to discern kiosk watchers’ demographics. Those demographics could include gender,
age category, and dwell time for each viewer. Note that viewers may overlap, so as one moves in time from left to
right in Figure 1, demographics collected may capture multiple watchers at the same time slice. This is depicted in
Figure 1, the row associated with channel 1, by the overlaps in the colors representing the dwell time of individual
viewers. On channel 2, the kiosk’s ads are displayed through time, and by applying analytics that leverage data
from channel 1, predictive models may change the order of videos displayed in order to make more impressions
on watchers who represent certain demographic segments. Channels 3 and 4 may be used in the background by
the kiosk to transmit data and analytical model updates back and forth to its owners. Channel 5 may be reserved
for the kiosk’s internal coupon printer. A customer who chooses to share data from a cell phone on channel 6 may
subsequently be scored by a predictive model that directs the kiosk to print a coupon. Billing media providers for
videos displayed may be captured on channel 7.
The types of data transmitted through channels 3 and 4 might include channel 1 data streams, control information,
couponing counts, billing details, etc.

to the renting agreements, with a mix of short, of smart machine deployment combined with AoT
medium, and long-term contracts. deployment remains elusive.
As a result, CIOs will be immersed in Ransbotham and Kiron also point out that
contracts about data and analytics ownership. effective data sharing practices among existing
Infrastructures will be needed to enforce customers, potential customers, suppliers,
contracts, and business negotiations regarding and even competitors is more common in
data and analytics ownership contracts will take organizations that exhibit a high ability to
center stage. innovate. Those practices are controlled by data
governance that specifies what can and cannot be
Towards an Era of AoT-driven shared.
Innovation Related Harvard Business Review articles
reinforce AoT’s potential impact on innovation.5
Although the AIM Suite experiment was not Although Gordon Hui doesn’t use the AoT
as successful as anticipated, there is no question acronym, he focuses on value creation and
that the potential for AoT to drive innovation capture aspects with an IoT mindset that
is significant. A recent MIT report asserts that incorporates the importance of data and
while companies were realizing only limited analytics. Hui asserts that IoT will require a shift
competitive advantage from data and analytics in corporate thinking from a traditional mindset,
in 2013-2015, that trend has reversed.4 Data and where products address lifestyle needs from a
analytics are being leveraged to innovate existing reactive perspective, to one where new products
operations as well as to create new business are proactive and capable of evolving through
processes, services, and business models. Authors their use of real-time predictive analytics. These
Ransbotham and Kiron define smart machines new products will be capable of being updated
as being capable of using algorithms to discern wirelessly, and some of their properties may
patterns from masses of data, or as above, able remain dormant until competitive forces require
to leverage machine learning and predictive that those dormant capabilities be turned on. In
analytics. They also conclude that smart machines addition, opportunities for recurring revenue
create opportunities for innovation, but evidence streams will be possible given that products
5 Hui, Gordon “How the Internet of Things changes business

4 Ransbotham, S. and Kiron, D., “Analytics as a Source of Busi- models,” Harvard Business Review, July 29, 2014, https://hbr.
ness Innovation,” MIT Sloan Management Review, February 2017. org/2014/07/how-the-internet-of-things-changes-business-models.
Table 1: AoT Business Shifts and Required Governance Extensions

AoT Governance-related
AoT Business-related Shifts
Extensions
From products to services; from Mitigate risks of model decay
dormant to emergent; from reactive resulting in outdated models and
Mindset Shift
to proactive; from single sale to data at the edge - related model
recurring revenue update latency issues
There are risks of bad data and
From standardized to personalized;
inappropriate algorithms being
systems leads who understand
used for personalization; there
Design design and are part of product
are security and privacy concerns;
teams; teams follow a build-test
moving computation to the edge is
and learn process
new and perhaps less reliable
Real options methods are Need for mechanisms guaranteeing
challenged by difficult data center adherence to analytics and data
ROI Evaluation scale cost estimation, data mining sharing contracts; cross-company
discovery uncertainties and the digital ecosystem creation requires
possibility of chaos new negotiation strategies
Table 2: European Union AoT Governance Concerns

A. Governance
High-level Issues B. Security
C. Privacy
1. Legitimacy & Transparency
2. Size & Heterogeneity
Upper Level 2 Governance Issues 3. Scalability
4. Data Storage Protection
5. Data Communication Protection
a. Audit & Monitoring
b. Identity Management
c. Device Authentication
Lower Level 2 Governance Issues
d. Context Awareness
e. User Data Anonymization
f. Metadata Anonymization
can be personalized and even reconfigured AoT-based product design given that many of the
after purchase. From an AoT perspective, the mindset shifts required will involve a company’s
ongoing role of data, analytics, and the automated IT infrastructure.7
embedding of product capability will be key to Iansiti and Lakhani report on GE’s shift in
managing the evolving product lifecycle. While mindset from a product focus to one in which
this may not be very surprising given that modern data and analytics provide decision support
washers, refrigerators, and even children’s toys services.8 Nelson and Metaxatos encourage new
now generate data and can be updated over the AoT-based product engineers to pay attention
Internet, taking those design ideals to other to design.9 They recommend starting with a
product areas will be rapid and pervasive. Hui clear problem statement and asking questions
and others point out that new product designers about why features matter and what customers
should consider that network effects between will pay for those features that matter to them.
products can and will shape markets. In this Therefore, they encourage appointing a system
context, he advises that organizations will need lead who understands design and designers who
to take into account how different stakeholders understand the technology. And they recommend
within a product’s ecosystem make money. Being a build-test-learn process, which borrows from
savvy about positioning AoT-based products into Apple’s success with AoT-influenced iPhones,
existing ecosystems and carefully contemplating iPads, watches and, iPods.
the types of new ecosystems that can be created
with new genres of products are both part of the AoT Governance
new AoT mindset. The era of AoT-driven innovation through data
D. Chivers concludes that companies will and analytics sharing implies significant changes
find that AoT can: “… change the category you to traditional IT governance. Mindsets must
compete in, the products and services you sell
and how you market them, and even the talent 7 Gandhi, S. and Gervet, E. , “Now That Your Products Can Talk,
you acquire.”6 Many agree that the CIO and IT What Will They Tell You?” Sloan Management Review, Spring 2016.
8 Iansiti, M. and Lakhani K.R., “Digital Ubiquity: How Connec-
team should be an important partner in new tions, Sensors and Data are Revolutionizing Business, Harvard Busi-
ness Review, November 2014.
6 Chivers, D. “The Boundaries Around Your Industry are About to 9 Nelson, S. A. and Metaxatos P., “The Internet of Things Needs
Change,” Harvard Business Review, November 2014. Design, Not Just Technology,” Harvard Business Review, April 2016.

shift from reactive products to proactive ones as will be a key challenge. At a more detailed level
predictive analytics anticipate customer needs of analysis, there will need to be ongoing auditing
and personalize responses. Recent research for and monitoring of devices in an AoT ecosystem.
the Advanced Practice Council addresses best Governance overhead costs will include polling
practice governance guidelines for predictive devices to audit their performance and monitor
analytics and business intelligence.10 Table 1 their proper behavior. Managing identity will
summarizes the shifts that are shaping AoT- require verifying identity before updating,
related strategic business thoughts and describes maintaining a global assessment of ecosystem
extensions to traditional IT and predictive device status, and maintaining protocols for
analytics governance those shifts create. new device authentication. Context awareness
For AoT, analytics (data gathering and issues may vary by implementation domain and
computation) will move to the organization’s ecosystem performance expectations. Similarly,
edge as demonstrated by the Intel AIM Suite. anonymization requirements may be more
Updating analytics throughout a complex stringent in, for example, health care than in some
network of devices creates new governance closed-loop industrial systems.
challenges. For example, there would likely be Given the strong positive relationship between
product liability issues should an individual’s data sharing across organizations and AoT-
AoT-based pacemaker be updated too late due to related innovation, effective governance related
decayed or outdated data and models. This issue to legitimacy and transparency will be essential.
extends to many types of AoT-based embedded Inter-organizational contracts can be the glue
devices in the medical domain and in many that ties together AoT relationships. And there
industrial applications. New risks associated with is an important role for CIOs in negotiating such
algorithms at the edge that could be accused of contracts. Consider the AIM Suite, for example.
discriminatory personalization are also a concern. Ecosystem partnerships could have been at
Ethical concerns abound as well. In addition, the core of the business models for capturing
the reliability of sensors and analytics at the and sharing dwell time data and for parlaying
edge can be an issue if redundancy has not been predictive analytics to motivate potential
adequately built in. Privacy and security are also customers. Sharing dwell time data among
major governance concerns. video providers, mall owners, mall department
A European Union (EU) report lists security stores, and others could, according to the MIT
and privacy concerns related to AoT governance.11 report, catalyze innovation. However, each
Table 2 displays the upper and lower level sets entity will need different types of data and data
of concerns. In the upper level are issues that aggregations. Exploring AoT contracts is the
are more general; lower level issues are more purpose of the remainder of this report.
aligned with specific technical solutions that
may already exist or that need to be addressed New Unit of Analysis for AoT
through new technical research. Authorized Contracts
devices added to an organization’s AoT ecosystem
will need to follow protocols to indicate they In order to make savvy decisions about AoT
legitimately belong to that ecosystem. The variety contracts, an organization must define how AoT
and volume of devices and data types will be a will drive new revenues, cut costs, and optimize
concern. Since AoT involves data and analytics, assets. This step often includes considering the
ensuring their proper storage at the edge will be broader ecosystem within which the new AoT
an important consideration. Similarly, protecting solutions will be deployed. This larger context
data communications to and from AoT devices may require a new lens or unit of analysis
for fusing AoT deployment elements such as
10 Goul, M. “Governance Guidelines for Predictive Analytics analytics, data, inter-organizational systems,
and Business Intelligence Across the Firm,” Society for Information customer interfaces, dormant capabilities that
Management Advanced Practices Council Report, 2016.
11 IoT Governance, Privacy and Security Issues, Report of the
can be “turned on”, and the AoT strategic mindset.
IERC – European Research Cluster on the Internet of Things, 2015,
retrieved October 7, 2017 http://www.internet-of-things-research.eu/
documents.htm.
Figure 2: The Star Wars Assemblage
Star Wars Assemblage

UI Assemblage
Object Assemblage
C-3PO: [to R2-D2] No! Shut them *all* down,

C-3PO: Where could they be? hurry!
[R2 beeps at him] [R2 shuts down the compactors]
C-3PO: Use the comlink? Oh my! I forgot, I turned it Luke Skywalker: What? HAHA! Hey, you did it 3PO!
off. [Luke, Leia, and Han start laughing hysterically; it
[over the comlink] sounds like screaming]
C-3PO: Are you there, sir? C-3PO: Listen to them, they're dying R2! Curse my
Luke Skywalker: 3PO? metal body, I wasn't fast enough, it's all my fault!
C-3PO: We've had some problems... My poor Master.
Luke Skywalker: [interrupting] Will you shut up and Luke Skywalker: 3PO, we're all right! We're all
listen to me! Shut down all the garbage smashers right! HaHa! Hey, open the pressure maintenance
on the detention level, will ya? Do you copy? Shut hatch on unit number... where are we 3263827.
down all the garbage smashers on the detention C-3PO: Don't call me a mindless philosopher, you
level! Shut down all the garbage mashers on the overweight glob of grease.
detention level!
This report proposes assemblage theory as such a later movies in the series). One might think of
lens.12 the Star Wars Assemblage as having the user
Figure 2 includes a “Star Wars Assemblage.” As interface sub-assemblage, C-3PO, and an object
the term assemblage denotes, the lens includes a sub-assemblage, R2-D2. While C-3PO could
bundle of different types of AoT capabilities. For communicate well with humans, R2-D2 could,
the first Star Wars film, millions of moviegoers through a connecting device, interface well with
were introduced to robots C-3PO and R2-D2. other machines and tap into their assemblages
C-3PO was a chatty robot, very much designed to intercept information and to exert controls
for dialog and human interaction. In contrast, over other assemblage’s components. Thus, an
R2-D2 spoke in a whistle language, and while assemblage can be thought of as the synthesis of
capable of communicating with C-3PO using this sub-assemblages, and each sub-assemblage may
language, the robot did not directly converse have certain capabilities, including direct cross-
with humans using natural language (until assemblage communication. Assemblages may be
part of a broader ecosystem in the same way that
12 Hoffman, D. L. and Novak T. P., “Consumer an Object Experi- C-3PO and R2-D2 were both robots in that early
ence in the Internet of Things: An Assemblage Theory Approach,”
August 21, 2016, retrieved October 9, 2017 from https://papers.ssrn. Star Wars movie.
com/sol3/papers.cfm?abstract_id=2840975. The Star Wars Assemblage highlights many
new issues associated with AoT assemblages

Detailed Star Wars Assemblage
In the classic scene depicted in Figure 2, Luke Skywalker, Leia, and Han Solo were caught in a trash compactor
on a starship. As the trash compactor’s walls began closing in on the three, there was a dialogue that captures
assemblage-to-assemblage and human-to-assemblage interactions. That dialogue began with C-3PO asking R2-
D2, “Where could they be?” in reference to the whereabouts of Luke, Leia, and Han. When R2-D2 whistles back a
suggestion to 3PO to use the “comlink” this demonstrates a machine-to-machine communication, where there is
effort on the part of both sub-assemblages to address the broader problem at hand. R2-D2 understands that the
whistle language cannot be used over the comlink to communicate with humans, and the robot knows that C-3PO
would not think of using the starship’s internal systems to send a message to solve the problem unless the robot
was so advised. R2-D2 understands some of C-3PO’s inherent limits likely because it learned that lesson through
partnering on problem solving in prior experiences.
C-3PO talks over the comlink as per R2-D2’s advice. When C-3PO recognizes Luke’s voice in response, C-3PO begins
a dialog about problems that the two robots had. Luke immediately cuts C-3PO off, he knows that C-3PO can tend
to ramble on, and he therefore utters “shut up” and repeats the command to “shut down the garbage smashers”
several times. Next, C-3PO infers that R2-D2’s typical operating mode would be to talk to each smasher one-by-
one as they are shut down. He emphatically asserts that R2-D2 needs to issue the command to shut them all down
at once. R2-D2 can extend an arm that connects into a special socket on the starship as depicted in the picture in
the upper right of Figure 2. Here, C-3PO understands R2-D2’s reasoning limits and takes appropriate action. Once
again, 3PO has likely learned that this is R2-D2’s predicted behavior through prior experiences and interactions.
After R2-D2 shuts down the compactors, C-3PO hears Luke talking wildly, and the robot interprets hysterical laughing
as indicating that the humans have died. This interpretation is likely due to C-3PO’s lack of prior experiences with
such a (possible) death scenario, so C-3PO does not possess the necessary understanding. However, with this new
set of data points and the outcome, C-3PO’s predictive capabilities can be updated and improved. With a final
interjection of some humor into the dialogue, R2-D2 apparently teases C-3PO about this lack of knowledge about
the difference between hysterical laughing and dying (C-3PO is a “useless philosopher”), and C-3PO responds with
a jab at R2-D2’s structural limitations (no arms and limbs like C-3PO has, but R2-D2 has a significant amount of
lubricants). Here again, this subtle understanding between the two sub-assemblages is a result of learned findings.
in general. When consumers and assemblages In Figure 3, C-3PO is replaced with Apple
interact with AoT, new consumer and assemblage Watch, R2-D2 with an iPhone and other
identities can emerge. Through data capture, assemblages that can connect to this assemblage,
assemblages can leverage experiences through including the Nest assemblage and the Apple
machine learning and predictive analytics to HomeKit accessory assemblage (e.g., light
behave differently in subsequent interactions. fixtures, wall outlets, fan controls, alarms,
This implies that data enables the transfer of smoke detectors, audio equipment, etc.). The
identity through interactive experiences–identity watch and the iPhone have machine-to-machine
can transfer from consumer to assemblage and communication capabilities akin to R2-D2’s
from assemblage to assemblage. As in the Stars special connector that links into a starship’s
Wars assemblage example, C-3PO and R2-D2 have control systems. R2-D2 also has its whistle
experiences enabling each to have an implicit language that it uses to communicate with C-3PO.
understanding of the behavior of the other. Apple’s IoT HomeKit devices can communicate
Luke had an understanding of C-3PO’s tendency with both the phone and watch since they run
towards being long-winded; humans give identity the IOS operating system. The watch-iPhone
to assemblages based on experience. While it connection is through Bluetooth and Wi-Fi–
may seem that such new scenarios are far in our after the devices are paired. The Apple Watch
future, consider Figure 3: An Apple Ecosystem can connect to a Nest assemblage through an
Assemblage. app, therefore creating a user interface to Nest’s
thermostat that uses machine learning. The Apple
Figure 3: An Apple Ecosystem Assemblage
Assemblage
UI Object
Assemblage Assemblage
OX
Assemblage
m2m
communication
User experience Object experience

dynamics dynamics
• User impacts • Object impacts
on the on the
assemblage assemblage
• Assemblage • Assemblage
impacts on impacts on the
the user object
assemblage can impact Nest through experiences, user experience dynamics with an assemblage.
thereby illustrating assemblage experience Through my interactions, my actual commutes,
dynamics. the assemblage learned a pattern and then had an
If you own an Apple Watch, you may have impact on me in a way that altered what I thought
discovered a pop-up message informing you was an extreme aversion to giving up my privacy.
of the time it will take you to drive home. While Figure 4 shows the range of predictive analytics
this was startling to me, I found the information and machine learning that might occur within an
quite valuable. Where did this come from, and assemblage.
how did it know this? Apparently, this was a An assemblage bundles data, user interfaces,
dormant feature in the assemblage. I assume that machine-to-machine communication, predictive
my iPhone was tracking GPS on my commute and, models, machine learning, and dormant
over time, it discerned the pattern, and leveraged properties. Companies that don’t take the
current traffic information to predict my drive AoT assemblage perspective to what they
time. Interestingly, the privacy implications might create or fail to consider the broader
may be enormous, but for me, personally, I was assemblage ecosystem in which their new AoT
delighted enough by my consumer experience might participate can miss the mark in a highly
that I was willing to give up some of my notions competitive marketplace. With the assemblage
of privacy. This interaction is an example of as the unit of analysis, developing strategies

Figure 4: Assemblage Predictive Analytic/Machine Learning (PMs: Predictive Models/

Machine Learning Algorithms)
Assemblage
PMs Guide the UI-Object

communication protocol
PMs Guide Assemblage UI OX

Object
interaction with a user Assemblage Assemblage
Assemblage
m2m
PMs turns on dormant communication
properties
User experience Object experience

dynamics dynamics
User Data
• User impacts • Object impacts Object
on the on the Data
assemblage assemblage
Assemblage Data • Assemblage • Assemblage Assemblage
impacts on impacts on the Data
the user object
New PMs constructed

from data mining of user,
object, and assemblage
data
for how to engage with other companies in an mining data, is key. It is the consolidation of
AoT ecosystem is more straightforward than it these models and the collected dwell time data
would be if one just considered each individual that together shape the character and emergent
predictive model or a specific subset of data behavior of the assemblage. Consider the role of
to own or share. Data and analytics sharing predictive models as part of the user interface,
contracts, for example, can be discussed from the the manner by which sub-assemblages interact,
broader vantage point of the whole rather than and how predictive models enable interaction
the parts. Table 3 elaborates the Intel AIM Suite in with video providers. These models capture
the assemblage context. the behavior of the entire assemblage, and
In this view of the Intel AIM Suite assemblage, this perspective can help to shape data and
the set of predictive models, constructed by analytics ownership contract negotiation. At
Table 3: Intel AIM Suite as an Assemblage

Nature of PMs Interactions Intel AIM Examples
Consumer with User Interface Sub- Predictive models capture the
User Interface Predictive Models
assemblage demographics of video watchers
User interface Sub-assemblage with Predictive models issue coupons to
User Interface Predictive Models
Consumer video watchers
Predictive models change the video
Sub-assemblage with another Sub-
Assemblage Predictive models display order based on watcher
assemblage
demographics
Sub-assemblage with Ecosystem Predictive models up-sell time/
Object Predictive Models
renters video space to retail suppliers
Figure 5: AoT Assemblage Contracting Issues for an Agriculture Example
stake is deciding what data needs to be solely created or derived through data that is co-
owned in order to maintain control over the owned is thereby co-owned by the same
assemblage’s revenue production capabilities. entities who co-own the data; if there is
Similarly, one must decide which predictive no co-mingling, the entity that creates the
models can be shared in order to help all parties model exclusively owns the model.
of the assemblage ecosystem innovate and ●● Data post-use vs. No data post-use: If data
thereby increase the value of the assemblage post-use is a contract provision, then if a
to all stakeholders. For example, from Intel’s data owner leaves a partnership governed
perspective, it may be advantageous to co-own by a contract, the data that entity owns
the predictive models that issue coupons with becomes the property of the remaining
merchant and video provider stakeholders in owner(s); if there is no data post-use
order to innovate around the value that the contract clause, the data solely owned by
assemblage brings to the bottom line of all an exiting owner is deleted/out-of-bounds
stakeholders. (Note: Ownership often implies data can
By focusing on ecosystems and the contracts be sold).
that glue stakeholders together, AoT assemblages
provide a business perspective for negotiating ●● Analytical model post-use vs. No analytical
contracts. Consider the following concepts model post-use: If an analytical model post-
that can be used in actual contracts between use provision is included in a contract,
assemblage stakeholders: then an analytical model owned/co-
owned by a exiting partner remains the
●● Data Exclusivity vs. Non-exclusivity: When property of the remaining owner(s); if
data ownership is exclusive, there is a there is no analytical model post-use, then
specific entity that owns the data; data an analytical model solely owned by an
is co-owned or open when contract exiting owner is deleted.
provisions stipulate that data is non-
exclusive. ●● Simple: Contracts are for a specific
duration with agreed to fixed monetary
●● Co-mingling vs. No Co-mingling: In co- exchange based on specific units of
mingling, an analytical (predictive) model

Table 4: Summary of Research Findings
analysis (e.g., data volume, customers, ecosystem, and partnering refers to the nature of
etc.). the relationships among stakeholders. Finally, the
●● Value-based: Contracts specify how last column summarizes findings relevant to the
value derived from the relationship given scenario.
is split between parties (incentives to Row I refers to the context where the contract
both stakeholders); requires sharing of is simple, data and analytics are openly shared,
measurable outcomes. and the perspective is from the vantage point
These contract provisions can be used of the predominant assemblage and sub-
to formulate a large set of possible contract assemblage owner. Timing doesn’t matter in this
scenarios. For example, a value-based contract case, and the partnering relationship is one of
that shares revenue among Intel AIM Suite open sharing. The important finding here is that
stakeholders on actual profit made from sales the owner needs an infrastructure with a large
generated could be negotiated where dwell capacity, all other things being equal, in contrast
time data is non-exclusive, there is co-mingling, to an architecture where contract provisions are
and analytical models, and data are available proprietary. Democratization (complete sharing)
post-use. This is a fully democratized (in terms of analytics and data comes at a cost to the
of sharing) case, but Intel may wish to retain assemblage owner.
exclusive ownership of the analytics that Row II is taken from a participating
predict demographics from images, and video stakeholder’s perspective. If the partner is an
providers may wish to have exclusive ownership early entrant into an assemblage ecosystem, and
of predictive models that issue coupons. The that participant sustains for a relatively long
assemblage perspective provides a lens by which period of time, then (under the data post-use
the business of AoT solutions comes into focus for paradigm) that participant can anticipate co-
contract negotiation–a bundle that is more than owning a large percentage of the assemblage’s
the sum of its individual data stores or predictive ingested data during that time period. This could
models. apply to a farmer in the agricultural example of
As another example, consider an agricultural section IV. Suppose an AoT assemblage provider
context where an AoT assemblage includes video tries to entice farmers in the area to leverage its
data from drone flyovers of farm land, data from services (e.g., watering, drone flyovers, etc). If
crops with sensor-equipped pest tests, seed there is data sharing post-use, then a farmer who
source data, sensor measured - predictive model- is an early entrant can gain ownership of a large
driven watering, etc. amount of the ecosystem captured data.
Figure 5 summarizes the contractual Row III is similar to the case of Row II,
complexities in any AoT assemblage that except that an early entrant in a data post-use
are relevant to contract negotiation.13 New contracting assemblage partners with another
research is needed to guide stakeholders in their early entrant that operates in a complementary
negotiations in these situations. area. Together they can co-own even more data
than in the scenario of Row II. Row IV focuses on
Recommendations: Contract the scenario where there are no data post-use
contract provisions. In this scenario, a participant
Guidance should not exchange any value for a post-use
Table 4 summarizes nine contracts that contract with another participant, because the
can serve as an early guide to AoT assemblage data ingested into the whole ecosystem will likely
contract negotiation. The type of each contract be fairly quickly flushed. The scenario of Row V
is noted. Next, the contract’s provisions and is similar: it indicates that trying to negotiate for
perspective are stated. Timing refers to when data post-use in an assemblage where no other
a participant may opt to enter into an AoT parties are sharing data post-use is not a good
option.
13 Adapted from Archer “Agree to Agree: Data Ownership, Protec- In scenario VI, the focus is on a late entrant
tion and Precision Ag,” September 2015, Retrived from http://www. who may target a long-term, early entrant for
tmtindustryinsider.com/2015/09/agree-to-agree-data-ownership-
a data sharing agreement. The research shows
protection-and-precision-ag-part-2/ on October 15, 2017.

that the late entrant can indeed benefit from contract negotiation. After contracts have been
that relationship. Scenario VII addresses the decided, CIOs should be actively involved to
situation where participants have exited an ensure that contract provisions are adhered
AoT assemblage ecosystem characterized by to. As discussed above, AoT contracts and
data non-exclusivity and analytics post-use. The assemblage ecosystem investments will be
predominant assemblage owner will require linked as certain contract provisions imply the
some sort of garbage collection mechanism that need for certain infrastructure capacities and
executes on a regular basis in order to avoid networking capabilities. The very fact that AoT-
unnecessary infrastructure costs. based products have dormant properties that
In Row VIII, where there are predictive can be “turned on,” and that behaviors and
analytics co-mingling and data non-exclusivity, identities of assemblages emerge after usage (as
predictive model updates can become an predictive models and machine learning adapt
issue. The research resulted in highly bi- sub-assemblage behaviors) implies that CIOs
modal distributions for communication and IT areas will need to engage with products
and coordination costs for predictive model throughout their useful lifecycles.
updates. In instances where update latency is More research is needed to provide guidance
significant, it will be very important to have large to CIOs regarding contract negotiation issues
communication pipes between sub-assemblages surrounding data and analytics ownership.
in this contract scenario. This report provides first steps. Since industry
Finally, in Row IX, the research indicates is still in the initial stages of analytics and IoT
that early performance results from an convergence, there is some lead time. That said,
AoT assemblage were not very reliable; the companies who are early movers are already
simulations used in the research always took stipulating contract provisions for prospective
significant time to stabilize. This means that partnership models for AoT. Companies such
predominant assemblage owners should not rely as Axon (formerly Taser) have introduced body
too heavily on early ecosystem performance in cameras for police officers, thereby including
developing strategies to improve infrastructure cloud providers and judicial systems.14 All Traffic
performance. Solutions has advanced sensors and wireless
communications to usher in a new era of traffic
Conclusion management, thereby including cloud providers
The emerging AoT convergence combining and hardware vendors.15 These early movers
complex analytics, artificial intelligence, and IoT will be watched closely; some may fail and
highlights the need for research on contractual others could succeed beyond expectations. With
instruments that will shape data and analytics business and competitive shifts underlying these
sharing in order to foster inter-organizational ventures, AoT is likely to be a major disrupter for
innovation. The notion of an assemblage can a long time to come.
serve as a lens that focuses on the business
aspects of complex AoT. Contracts that glue About the Authors
together stakeholders in AoT assemblages should Michael Goul
be viewed in the context of ecosystems that Michael Goul is Associate Dean for Faculty and
involve multiple stakeholders. Research and a Professor at the W. P. Carey School
This report’s early guidance on contract of Business at Arizona State University. Michael
negotiations with AoT considers different oversees faculty and department affairs and
ecosystem perspectives, e.g., the predominant
owner of the ecosystem and participants. 14 Pasternack, A. “How the Lucrative Fight to Put Cameras on
Considerations should address data and analytics Cops is Changing the Way Police Work,” The Future of Policing,
June 2017, retrieved from https://www.fastcompany.com/40425969/
sharing, the nature of ecosystem investments, and how-the-lucrative-fight-to-put-cameras-on-cops-is-changing-the-way-
different contract strategies based on timing. police-work on October 15, 2017.
The CIO’s role in AoT deployment is crucial 15 “All Traffic Solutions Closes $8 Million in Series A Funding,”
June 2016, retrieved from http://www.alltrafficsolutions.com/press-
beginning with the early stages of design and release/all-traffic-solutions-closes-8m-in-series-a-funding/ on October
continuing through data and analytics sharing 15, 2017..
leads initiatives to advance faculty excellence. Research Director

In addition, he oversees the School’s portfolio Richard Watson, Ph.D.
of research centers, coordinates the School’s 706.542.3706
Ph.D. Program, and he represents the School rwatson@terry.uga.edu
on University research initiatives such as those
associated with advanced analytics. While SIM Headquarters Staff
Associate Dean, standings in the National Science Vincenzo Nelli
Foundation’s Higher Education Research and 856.380.6827
Development rankings improved from #43 to vnelli@simnet.org
#14 in Economics and #31 to #14 in Business &
Management. From 2013 - present, the School’s
published research ranks #21 in North America
and #22 Worldwide in a prominent top 100
business school research ranking. In 2017-
18, Goul steered a ‘Professionals in Practice’
task force designed to engage non-tenure track
faculty in establishing a new matrix governance
structure for advancing Carey School teaching
and learning. For the six years prior to his
current role, Goul served as chair of the school’s
department of information systems. As chair,
Goul spearheaded the development of onsite and
online platform versions of the MS in Business
Analytics program and the undergraduate BS
in Business Data Analytics degree. Michael’s
most recent research efforts are in the areas
of big data/data science, services computing
and analytics in the Internet of Things. He
recently guest edited dual special issues of
IEEE Transactions on Services Computing and
Informs Service Science on aligning business and
technical services research. Michael is a member
of the Society for Information Management’s
Advanced Practices Council Research Advisory
Board. In summer 2016, Goul was recognized
with the Outstanding Leadership Award by the
IEEE Computer Society Technical Committee
on Services Computing. He has published over
one hundred articles, authored cases and he
conducted analytics research at companies
including adidas, American Express, Blue Cross
and Blue Shield, eBay, Intel, the U.S. Army and
Teradata.
APC Contact Information

Program Director
Madeline Weiss, Ph.D.
301.229.8062
Madeline.Weiss@simnet.org

Copyright of MIS Quarterly Executive is the property of MIS Quarterly Executive and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

APC Forum Poised Between A Wild West of Predictive Analytics

Uploaded by

Copyright:

Available Formats

APC Forum Poised Between A Wild West of Predictive Analytics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

APC Forum Poised Between A Wild West of Predictive Analytics

Uploaded by

Copyright:

Available Formats

APC Forum: Poised Between ‘a Wild West of

Predictive Analytics’ and ‘an Analytics of

December 2018 (17:4) | MIS Quarterly Executive 333

Figure 1: Channel by Channel Depiction of a Kiosk’s Behaviors

A Different View Time    

AIM Suite AoT Mall Illustration

December 2018 (17:4) | MIS Quarterly Executive 335

5 Hui, Gordon “How the Internet of Things changes business

Table 1: AoT Business Shifts and Required Governance Extensions

Table 2: European Union AoT Governance Concerns

December 2018 (17:4) | MIS Quarterly Executive 337

Figure 2: The Star Wars Assemblage

Star Wars Assemblage

C-3PO: [to R2-D2] No! Shut them *all* down,

December 2018 (17:4) | MIS Quarterly Executive 339

Detailed Star Wars Assemblage

Figure 3: An Apple Ecosystem Assemblage

User experience Object experience

December 2018 (17:4) | MIS Quarterly Executive 341

Figure 4: Assemblage Predictive Analytic/Machine Learning (PMs: Predictive Models/

PMs Guide the UI-Object

PMs Guide Assemblage UI OX

User experience Object experience

New PMs constructed

Table 3: Intel AIM Suite as an Assemblage

Figure 5: AoT Assemblage Contracting Issues for an Agriculture Example

December 2018 (17:4) | MIS Quarterly Executive 343

Table 4: Summary of Research Findings

December 2018 (17:4) | MIS Quarterly Executive 345

leads initiatives to advance faculty excellence. Research Director

APC Contact Information

December 2018 (17:4) | MIS Quarterly Executive 347

You might also like

C-3PO: [to R2-D2] No! Shut them all down,