Successful Visualization Projects Require Managing Cultural Changes

by   |   November 12, 2013 5:17 am   |   0 Comments

Johns Hopkins Hospital maintains a data warehouse of patient details, including diagnoses, demographics, admission and discharge information, insurance and payment data. The information is used to measure the cost of care and the hospital’s overall performance. Hospital analysts use Tableau to access data from this and other sources.

Johns Hopkins Hospital maintains a data warehouse of patient details, including diagnoses, demographics, admission and discharge information, insurance and payment data, to measure the cost of care and the hospital’s overall performance. Hospital analysts use Tableau to access data from this and other sources.

Business intelligence (BI) dashboards have been staples of enterprise IT for decades and visual query and data discovery tools aren’t new. But the past several years have seen a proliferation of easy-to-use graphical applications that allow non-experts to conduct sophisticated analysis of high-volume and varied data sources. This has allowed organizations to provide a wide range of decision makers with analytical capabilities, reporting and collaboration that draw on unified, up-to-date views of data.

Specialist vendors like QlikTech, Tibco Spotfire, and Tableau Software have made inroads in organizations with low-cost applications for individual users that are easily scaled up to department level and beyond, prompting business intelligence and enterprise software companies like, SAS, MicroStrategy, Microsoft, SAP, IBM and Oracle to add or bolster visual data discovery and analytical options as well. Gartner pegs this fast-growing segment of the business intelligence software market at roughly $1 billion in 2013.

Related Stories

Advances in data visualization software empower business users.
Read the story »

7 tips for evaluating data visualization software.
Read the story »

Tactics for building and sustaining a data analytics team.
Read the story »

More articles in Data Informed’s Visualizations section.
Read the story »

Whether organizations settle on a stand-alone data discovery application or one integrated within a data warehouse or enterprise framework, the impetus is typically to get more data, and more appropriate and useful data, in front of more users in a timely manner.  This often means giving business analysts or other subject experts untrammeled access to vast stores of structured and unstructured data and allowing them to create dashboards and reports for their colleagues.

Such projects usually come with a heavy dose of organizational change.  In some cases, the new data flows—and accompanying transparency and accountability—cut across departmental lines or require a change in entrenched procedures.

That change represents a cultural challenge that managers and business analysts need to overcome. But the benefits of unlocking data that might previously have been slow in coming, not widely accessible, or not available at all generally outweigh the pushback and inertia, according to business analysts who’ve worked with the software.

That lesson cuts across vertical industries and use cases, as the experiences shared in this article by data visualization practitioners from the fields of health care, HR recruiting and manufacturing show.

Overcoming Resistance to Change a Hospital Admission Process
The Johns Hopkins Hospital in Baltimore gathers and disseminates huge volumes of data, on clinical and research activities, finance and accounting. A key driver of this is the state agency overseeing hospital costs and the delivery of service, Maryland’s Health Services Cost Review Commission. The commission, which sets the rates Maryland hospitals can charge for a given service, requires extensive tracking and disclosure of financial information and patient data (abstracts of medical records, demographics and billing).

Hetal Rupani of Johns Hopkins

Hetal Rupani of Johns Hopkins

The hospital funnels data from thousands of in-house and third-party sources  into its reporting system for the commission, according to Hetal Rupani, senior project administrator in the department of medicine.

Given the welter of information coming from so many disparate sources, hospital analysts were spending much of their time querying and generating reports. “We used to make our queries in Access and transfer that data to Excel to create some pretty graphs, and from Excel transfer graphs to PowerPoint for presentation,” Rupani said.

The department of medicine, which has over 3,000 employees and 570 fulltime faculty, uses Tableau’s desktop, mobile and server software to, among other things, speed the process of transferring patients from the emergency department to available beds when they’re admitted. The Maryland commission tracks the flow of patents through the hospital as part of its ongoing performance review.

The key metric is “boarding time”—the time it takes to find beds for emergency room patients once an attending physician decides to admit them. The department of medicine used Tableau to create a dashboard application, which, by providing doctors, nurses, shift supervisors, and others involved a centralized, up-to-date view of data on available beds and other information needed to make the complex hand-off between the emergency department and an inpatient unit, helped cut the average boarding time from 12 hours to 4, according to Rupani.

The effort to change this process ran into resistance, particularly from the emergency department, which was reluctant to open access to its stove-piped internal systems. The culture shift to better communication, transparency and mutual responsibility for processes involving multiple departments took about 6 months, according to Rupani.  “We had to show them the value of sharing data from bed management and emergency and other departments,” she said.

The project benefitted from buy-in and active pressure from senior managers.  The hospital also assigned analysts to each group to help in the transition to using the dashboard, according to Rupani. Emphasizing the tangible early results and mutual benefits of the effort helped bring the departments onboard, she noted.

IT was easier to bring around, since Rupani’s group was able to present a very specific set of requirements for the project. “Rather than depending entirely on my IT team to figure out what was going on and how they were going to help me, we as an analytical team, said to IT, ‘This is how we are going to do it.’  They just helped us in data modeling.”

New Data Views Shift Work for Analysts and Clients at HR Services Firm
International job posting service eQuest places ads for about 350,000 jobs per month on almost 1,000 online job boards. The company also advises its clients, which include more than half of the Fortune 500, on buying recruitment ads and complying with federal employment regulations. Its analytics group helps customers plan and measure the effectiveness of their recruitment marketing campaigns, according to David Bernstein, eQuest’s vice president of big data.

The analytics group earlier this year began using QlikTech’s QlikView data discovery and visualization software. Clients typically subscribe to eQuest’s service on a yearly basis and the company assigns an analyst to work with them. “Our data analysts jockey the dashboards that are used to extract the insights that we then feed to our customers,” Bernstein said.

HR clients aren’t typically data-savvy, and they are pressed for time and resources, so rather than provide them with business intelligence dashboards and data visualization tools, eQuest uses QlikView internally to advise them and produce customized reports, he said.

Whereas previously analysts were performing historical analysis on each customer’s own data, “now, in a forecasting kind of model we’re able to look globally across job titles or labor markets. We’re able to quickly navigate through huge amounts of information in minutes,” he said. That information includes data from customers’ systems, internal eQuest systems, and Internet sources.

Analysts at the San Ramon, Calif.-based HR services company had been working with basic batch-oriented systems for generating Excel spreadsheets. A quarterly report would take about five days to compile. The same report now takes about half a day to produce, according to Bernstein. “More importantly, we’re able to immediately interrogate the data, start pivoting and shifting it around and start seeing patterns that I then want to take to the visual layer. I can forecast and look at information in a more real-time way,” he said.

The swifter analytics and visualization also means each of eQuest’s half dozen analysts can handle a larger client load—30 or more now versus 10 in the past, according to Bernstein.

The insights are also more advanced. Rather than simply filling in clients on their recent recruiting and hiring—what they spent and what they got for it, eQuest analysts can now help them understand how long it might take to attract a candidate through a given media outlet, how many candidates they’re likely to get through it, which days are best to advertise, and what text to include in their ads. They can also check in with clients during their recruiting campaigns, rather than afterwards, and provide relevant data, including statistics aggregated from eQuest’s extensive customer base.

The new analytical tools have required a shift in the approach analysts take to interpreting data and framing their advice to clients. Already versed in historical analysis and reporting, they have had to become proficient with predictive modeling and real-time data flows, according to Bernstein.

EQuest clients have also had to get comfortable with the predictive nature of the data. “Most of what they’re used to is looking at a pile of old data and saying if we ever do this again, what might we do differently? It’s a complete mind shift. Many of my big data customers are still leveraging me as a historical look-back engine,” Bernstein said.

The IT department also faced an adjustment. “We’re not a business intelligence group by core competency,” he said. “What eQuest is core at is transactional, so growing those skills has been some of the challenge, too.”

EQuest went with QlikView after an unsuccessful implementation project with a competing vendor’s visualization software. The rival application turned out not to support eQuest’s reporting requirements or high volume of data and creating visualizations required too much of the IT department’s time, according to Bernstein.

How to Manage Change That Visualizations Can Bring

Some organizations investing in data discovery and visual analytics applications may do so because they want to shake things up, but many realize after the fact that greater access to data can change how they do business on individual, departmental and company levels. Elissa Fink, chief marketing officer at Tableau Software, and Kevin Spurway, senior vice president of marketing at MicroStrategy, summarized lessons learned by businesses implementing these applications:

Connect with users based on individual needs. More than changing entrenched business processes and policies, visual data discovery software can provoke resistance by exposing in real-time the behavior and decisions of individuals and departments and factoring in their effects on others. To counter this, spell out the benefits of the software to each user.

Acquire top-level champions. Senior executives need to actively promote the new software and procedures, rather than issue an edict.

Establish realistic goals. In setting project goals, don’t overreach.

Go for quick wins. The overall will benefit from quick wins, even if small, to show progress.

Count what you see. ROI can be hard to measure, but you can tally queries and reports created, questions generated based on the data.

One plan for IT, another plan for users. Projects work best when there’s a plan for IT and a plan for user education and training. Foster a user community. If possible, dedicate a position to education and hands on training.

Don’t stifle the innovation that’s happening at the lower levels if it doesn’t come in the normal channels – new ideas can emerge from the bottom-up.

Expect to make organizational adjustments. Plan for redefining relevant jobs based on the new analytical tools, and support staff in making the transition.

Make sure the visualization software you choose can handle your data volumes.

Have a plan for maintaining data security, data integrity and IT governance rules. Decentralization of data discovery systems can lead to “multiple versions of the truth,” whether by generating disparate analyses and reports from the same data, or by creating inconsistencies in metadata. Organizations need to account for the effect of these tools on their overall data architecture.

Pilot Project Wins Support with Views into Sales and Operations Data
Dan Boyce, director of demand planning at poultry and pet food company, Simmons Foods in Siloam, Ark., turned to Tableau out of frustration. Using Microsoft Excel to generate reports from Oracle E-Business Suite, he and his team of business analysts found they were unable to produce the volume, quality and timeliness they were after.

The group, which provides sales support and analysis, and sales and operations planning, began in the summer and fall of 2012 with a small demonstration project providing business intelligence dashboards to a few senior executives.

When Boyce’s group started using Tableau to create BI dashboards and reports for the executives, they began asking more substantive, causative (“why”) questions of more departments.  “As long as we had the data loaded, we could answer questions all day long, but the rest of the functions within the organization were a toolset behind us. So, when they went to answer these follow-up questions, it would takes days, and weeks sometimes, to answer,” he said.

The executives saw the value of the software and advocated for the $40,000 to $50,000 investment in the desktop and server versions. So far, the company has rolled out the software to operations planning and supply planning in both poultry and pet food groups.

“We saw need to integrate data so we’re answering questions in an integrated manner,” Boyce said. Simmons has a total of about 40 users, some using the desktop version to build data views and others connecting to the server from desktops or tablets, consuming the views and using the system as their primary information portal, according to Boyce.

The BI dashboards are mostly for executives and senior sales management, providing “better ways to interact with lots of different data sets to manage teams,” he said. “In sales and operations planning we have taken operational data and given it to our planning group and interact with that on a regular basis.”

In part, it’s about variables, Boyce explained. “We grow chickens and lots of things can affect how big they are.  We make the most money when we have a very consistent size of chicken coming in the back door. So, there’s a lot of variables that we’ve been able to throw in these models.” This allows the company to quickly respond to changing business conditions and other factors that affect production, including the weather and corn prices.

For example, a model projecting the cost of production needs to account for ethanol subsidies, which influence the price of corn. “If our corn prices go down, that’s a good thing for us, but the rest of the market’s going to react, too. We can do a lot better at modeling these things and react to them on the ground when we have a consolidated place to pour all this data,” he said.  

In piloting the software, Boyce and his group were careful to set expectations. “The executives expected to be able to get an answer very quickly with the tweak of a view or to pull in a different aggregation – units, or currencies, for example,” he said.

The analysts were able to provide answers often in minutes and “never less than a day, even for a very complex request,” Boyce said. They also went from running sales and operations planning reports once a month (twice, if pressed) to daily, with little manual input,” he explained.

This provided better insight into the smaller daily fluctuations throughout the enterprise, allowing the company to wring many small savings out of operations, which added up, according to Boyce. He convened daily meetings to identify follow-ups, whether scheduling changes or calls to customers, even changing the speed of the production line. Overall, the company has been able to increase on-time shipments and trim inventory by nearly 25 percent. Boyce chalks up a significant portion of those improvements to visual data discovery.

“It’s not just the [visual data discovery] product itself but the mentality of taking data and putting it in a format that the decision maker needs to either validate a good decision or give them enough information to tweak their behavior,” he said. “And when I say decision maker I mean all the way down to the person on the production line.”

While executives and managers welcomed the analytical capabilities, Boyce said it took about six to eight months to convince the IT department of the software’s benefits.

Ted Smalley Bowen is a freelance writer and editor based in the Boston area Reach him via email at

Tags: , , , , , ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>