December 1, 2016
Feature Story
Improving Marketing Data
with Human Computation
Sharon M. McIntyre
Innovation Researcher, Educator,
and Marketing Strategist

A VentureBeat report (Cifuentes 2015) on the state of marketing analytics indicated that while many companies are significantly increasing spending on related analytics services and technologies, it was also true that "marketers aren't that advanced in their analytical approaches" and there was a "gigantic skills gap in most organizations around data science."

Further, the report looked at more than 800 marketing technologies related to multiple business use cases including audience insights, e-commerce analytics, and unstructured data analytics—finding the chief marketing officer (not the CIO) was typically responsible for these technologies and there was a pressing need for marketing organizations to demonstrate business performance improvement related to this technology expense.

One way data professionals can help marketers get better insights and business results is to mentor them about designing better marketing research that includes unstructured data. Too many marketers simply rely on scanning corporate social media feeds, filtering online panel discussion threads, and examining open-text field content in surveys. Although there are many useful unstructured data analytics tools, such as Provalis or IBM Watson, that can make some sense of this flow of unstructured data, enterprises miss the opportunity to design more thoughtful, useful, and productive digital marketing research methods.

Data to Help Innovate

This is particularly pertinent when an organization is looking for predictive insights from marketing data that will help it innovate more successfully in new product (or service) design. Studies examining new product failure rates vary significantly in methods and results, but findings can range from 35 percent failure (healthcare products; Castellion and Markham 2013) to 95 percent failure (consumer products; Nobel 2011). Clearly, better insights into hidden needs and tacitly held beliefs are required.

Yet, as marketing author and speaker Seth Godin (2016) explains, "Too often, marketers do surveys, not polls, or bother everyone with a census, poorly done. Worse, they then use these results as an accurate prediction of the future instead of a reliable snapshot of now.... A poll doesn't predict the future.... If, on the day the iPhone was announced, you had done a well-designed poll and asked, ‘Do you intend to ever buy a smartphone?,’ the yeses would have certainly been less than 5 percent of the result.”

As a result, we find marketers—hoping to glean insights about consumers' future needs—turning to the aforementioned unstructured discussion groups, online chats, focus groups, user groups, and open-ended questions as expressive alternatives to reductive polls and surveys. What marketers can learn is that by augmenting these insight-seeking activities with creative and immersive research design, they can help participants better express their unspoken needs and beliefs by using interactive play, analogies, metaphors, images, symbols, and storytelling (Choo 2000).

Playful Human Computation

One useful way to address both unstructured marketing data analytics and the need for better innovation insights is by introducing human computation activities into the marketing researcher's toolkit.

The term human computation was first used in 1838, but researchers attribute its contemporary usage in computer science to CAPTCHA / reCAPTCHA inventor and Carnegie Mellon professor Luis von Ahn (Quinn and Bederson 2011). He is also well-known for creating the ESP game, which was used by Google to improve results for their image search function (von Ahn 2006). One of von Ahn's definitions of human computation is "the idea of using human effort to perform tasks that computers cannot yet perform, usually in an enjoyable manner” (Law and von Ahn 2009). This can take the form of participants providing image tags, meta coding for narrative texts, and ratings on the quality of contributions.

Even with powerful data analytics tools, preprocessing large volumes of unstructured data is required before data mining can occur effectively. As Barbier et al. (2012) outline, typical preprocessing techniques can include trust assessment, stop-word removal, vectorizing, and feature selection or extraction. By incorporating human computation design, research participants can take part in gamified trust assessment of unstructured marketing research data (text and images) by applying "likes," performing simulated spending tasks, and rating the quality of other participants' contributions.

Content Tagging

A popular human computation activity is content tagging. Researchers can show photos of products or use cases and have participants list the "three key words" they would use to describe how they feel about the scene in the photograph or video. When large numbers of participants are included in this activity, often through crowdsourcing platforms, the richness of the resulting data describing emotional responses to products and use cases can be interesting and useful.

“Fill in the Blanks” Sentences

When designing storytelling activities, researchers can incorporate a "fill in the blanks" question design that, in effect, presents the participant with precoded data intake activity that also embraces the enjoyable, immersive quality desired in generating insights for innovation. For example, instead of (or in addition to) an open online discussion, a human computation activity of this style might look like the following:

If I could only take one digital device with me on a vacation, it would definitely be my _____. I would use it mostly for _____ and _____. When I think I may have misplaced or lost this device it makes me feel _____ because _____. But the one thing I wish my device would do even better is _____.

Clearly, analyzing the content from these delimited text fields will be much more straightforward than open narrative texts. Further, depending on the research design, other participants can rate this completed sentence on whether it is similar or dissimilar to how they feel (e.g., they can vote with a "most like me" button and be eligible for gamified rewards).

Photo Elicitation

When researchers need to tap into participants' emotions and hidden beliefs, images are an excellent tool. The technique of photo elicitation can include having participants take their own photographs or they can select from a set of images provided to them to illustrate their ideas. However, Stedman et al. (2013) note that "photo elicitation is rarely used by social scientists" and acknowledge the irony because images are "more effectively tapping into informants' tacit, and often unconscious, consumption of representations, images, and metaphors," while producing "different and richer information than other techniques.”

There are many ways for marketers to combine photo elicitation and human computation. One simple method is to present participants with an image of a product in use as the main visual for a print advertisement, targeted at an audience very different from them (e.g., their grandparents). By providing fields for the participants to create an advertisement headline, compose a delimited text product description, and use two words to describe the main product benefits to this dissimilar audience, the researchers are tapping into powerful psychological projections about the product and producing data that is already categorized.

Summary

These are just a few examples of how data professionals can advise marketers to creatively delimit unstructured data collection with the impact of (1) more engaging user experiences, (2) tacit need elicitation, and (3) improved data analytics and innovation insights.

Sharon McIntyre is an innovation researcher, adjunct professor, and marketing strategist. She currently consults on open innovation methodology for crowdsourcing pioneer Chaordix, whose clients include LEGO, IBM, and Ford. Sharon is also advising the Cameroon, West Africa, energy utility company Eneo with a scope that encompasses frugal innovation and technology marketing mentorship. Her doctoral research focus is on entrepreneurial innovation process and policy. Visit shazzmack.com or follow Sharon on Twitter: @shazzmack.

References

Barbier, G., R. Zafarani, H. Gao, G. Fung, and H. Liu [2012]. “Maximizing Benefits from Crowdsourced Data,” Computational and Mathematical Organization Theory 18(3): 257–79. doi:10.1007/s10588-012-9121-2

Castellion, G., and S. K. Markham [2013]. “Perspective: New Product Failure Rates: Influence of Argumentum Ad Populum and Self-Interest,” Journal of Product Innovation Management 30: 976–79. doi:10.1111/j.1540-5885.2012.01009

Choo, C. [2000]. “Working with Knowledge: How Information Professionals Help Organizations Manage What They Know,” Library Management 21(8): 395–403, http://choo.fis.utoronto.ca/lm.

Cifuentes, J. [2015]. “The State of Marketing Analytics: Insights in the Age of the Customer,” VentureBeat, December 16, http://venturebeat.com/2015/08/21/new-research-companies-plan-to-massively-increase-spend-on-marketing-analytics.

Godin, S. [2016]. “What Does the Poll Say?,” blog entry by Seth Godin, October 31, http://sethgodin.typepad.com/seths_blog/2016/10/what-does-the-poll-say.html.

Law, E., and L. von Ahn [2009]. “Input-Agreement: A New Mechanism for Collecting Data Using Human Computation Games,” CHI 2009, April 4–9, Boston.

Nobel, C. [2011]. “Clay Christensen's Milkshake Marketing,” Harvard Business School Working Knowledge, February 14, http://hbswk.hbs.edu/item/clay-christensens-milkshake-marketing.

Quinn, A., and B. Bederson [2011]. “Human Computation: A Survey and Taxonomy of a Growing Field,” CHI 2007, May 7–12, Vancouver, BC.

Stedman, R. C., B. Amsden, T. M. Beckley, and K. G. Tidball [2013]. “Photo-Based Methods for Understanding Place Meanings as Foundations of Attachment,” in Place Attachment: Advances in Theory, Methods, and Research, ed. L. Manzo and P. Devine-Wright (London: Routledge), 112–24.

von Ahn, L. [2006]. “Human Computation,” GoogleTalks, July 6, https://youtu.be/tx082gDwGcM.

Additional Resources

Alvesson, M., and K. Skoldberg [2009]. Reflexive Methodology: New Vistas for Qualitative Research (London: Sage).

“IBM Watson Analytics for Social Media Tools,” available at: https://www.ibm.com/marketplace/cloud/social-media-data-analysis/purchase/in/en-in.

IDEO [n.d.]. “Design Thinking for Educators Toolkit,” available at https://www.ideo.com/post/design-thinking-for-educators.

Mankoff, R. [2008]. The New Yorker Cartoon Caption Book (Kansas City: Andrews McMeel).

Michalko, M. [1991]. Thinkertoys: A Handbook of Business Creativity for the ‘90s (Berkeley, CA: Ten Speed Press).

“Provalis Research Text Analytics Tools,” available at: https://provalisresearch.com.

TDWI Onsite Education: Let TDWI Onsite Education partner with you on your analytics journey. TDWI Onsite helps you develop the skills to build the right foundation with the essentials that are fundamental to BI success. We bring the training directly to you—our instructors travel to your location and train your team. Explore the listing of TDWI Onsite courses and start building your foundation today.

 
Announcements
NEW Best Practices Report
BI, Analytics, and the Cloud: Strategies for Business Agility
NEW Ten Mistakes to Avoid
When Defining an Enterprise BI and Analytics Strategy
NEW Business Intelligence Journal
Business Intelligence Journal, Vol. 21, No. 3
NEW TDWI E-Book
Big Data, Analytics, and the Cloud: Strategies for Success
NEW TDWI Checklist Report
Evolving Toward the Modern Data Warehouse
contents
Feature
Improving Marketing Data with Human Computation

The Data Warehouse Feedback Interface
Feature
Interest in Improving Data Preparation

Mistake: Overreliance on Open Source NoSQL
Education & Events
TDWI Conference in
Las Vegas (Pre-registration Now Open!)

Caesars Palace
February 12–17, 2017
TDWI Boston Accelerate
(Pre-registration Now Open!)

Marriott Copley
April 3–5, 2017
TDWI Conference in Chicago (Pre-registration Now Open!) Hilton Hotel
May 7–12
Webinars
Emerging Best Practices for Data Lakes
Wednesday, December 14
Maximizing the Value of Your IoT Data
Thursday, December 15
Marketplace
TDWI Solutions Gateway
Informatica – Data Management for Next-Generation Analytics
TDWI White Paper Library
7 Data Visualizations You're Not Using Enough
TDWI White Paper Library
Big Data Analytics Buyer's Guide

Member Discounts

Ready to take the CBIP Exams or attend our next conference? Take advantage of these exclusive member discounts.

10%

Discount

on TDWI Las Vegas

$26

Discount

on CBIP Exam Guide

Flashpoint Insight
The Data Warehouse Feedback Interface

A data warehouse receives transaction data from multiple operational systems and stores that data in a form specifically structured for query and analysis.

End users are able to view that data through a combination of standard reports, dashboards, and ad hoc queries to gain greater insight into their business.

In this manner, the data warehouse acts as a standalone system. Other than through BI tools, the data warehouse does not provide data to downstream systems. Therefore, the relationship between the operational systems and data warehouse is unidirectional.

This article considers a growing industry trend of the data warehouse returning data across feedback interfaces to the operational systems. This changes the relationship into a bidirectional one and allows the data warehouse to directly enhance the business.

We discuss real-world examples of data warehouse feedback interfaces that illustrate how they can provide a useful mechanism from a business perspective. We describe several ijmplementation architectures and discuss the impact of the data warehouse following an Inmon, Kimball, or hybrid design.

Learn more: Read the entire article by downloading the Business Intelligence Journal (Vol. 21, No. 3).

 
TDWI Research Snapshot
Interest in Improving Data Preparation

Before delving into data preparation processes specifically, we wanted to get an overall sense of the satisfaction of users in research participants’ organizations with how easily they can find relevant data and understand how to use it appropriately for BI and analytics.

These two related objectives are core to the mission of data preparation. The biggest percentage said users in their organizations are somewhat satisfied (36%), with 7% very satisfied (see Figure 1). More than a third (37%) indicated dissatisfaction. The results suggest significant room for improvement.

Organizations confront barriers when trying to make upgrades to data preparation, many of which are not about technology implementation. Our study finds that among research participants, an insufficient budget is the most common barrier to improving how data is prepared for users’ BI and analytics projects. The second most common barrier is not having a strong enough business case. The results highlight the difficulty many data professionals have in convincing executive management to invest in improving the quality of shared data assets (quality being a cornerstone of well-prepared data).

(Click for larger image)
Click to view larger

In conducting interviews for this report, we found that organizations tend to act only when a harmful incident has occurred, such as a new regulatory policy that must be observed or public embarrassment from customer complaints about bad data. Some will budget for solutions if there is a specific project, such as a migration or consolidation of systems after a business merger or acquisition. It can be difficult for data stewards to demonstrate and quantify the relationship between poor data quality and preparation and lost revenue, missed business opportunities, and costly process inefficiencies.

Complicating efforts to make sustained improvement to data preparation across the enterprise is the third most common barrier cited by research participants: lack of skilled personnel or training. Organizations frequently do not have enough internal expertise to work effectively and efficiently with data—or their expertise is spread unevenly and is dependent on a few “power users” rather than built up as part of a broader user training program. In many cases, most of the expertise resides in BI teams that are part of the IT function. However, BI teams themselves may lack skills in data preparation analytics projects if they are trained for more traditional BI reporting requirements.

Read the full report: DownloadTDWI Best Practices Report: Improving Data Preparation for Business Analytics (Q3 2016).

 
Flashpoint Rx
Mistake: Overreliance on Open Source NoSQL

The main benefit of commercial distributions for NoSQL is that they provide additional software or enhancements to the open source software such as support, consulting, and training.

Tools, however, that help administrators configure, monitor, and manage NoSQL are lacking for enterprises in open-source-only software.

Another need is for enterprise integration. Commercial distributions provide additional connectors of availability, scalability, and reliability as is common with other enterprise systems. These are well covered by the major commercial distributions. Some of the vendors push their wares back into the open source en masse while others do not.

When selecting how you will deploy NoSQL, keep in mind the process for getting it into production. You can spend a great amount of time developing and productionalizing if you do not use a commercial distribution. You already have a well-fit database for the application that should be much less expensive than a relational database, so some expenditure for a commercial distribution can be worthwhile.

Read the full issue: Download Ten Mistakes to Avoid in NoSQL (Q3 2016).