An in-depth first experience with SAP Visual Intelligence
Given my blog and passion is interactive data visualization, you can imagine how excited I was when I saw Visual Intelligence for the first time. I have purposely waited to pick it up for the first time until I had a legitimate business case for it and carve out the time to document my experiences.
THE BUSINESS PROBLEM
I wanted to explore how our customers are utilizing GMaps Plugin page views, so we can engage heavy usage customers to gather feedback, offer help to low usage customers, and identify any problematic trends that may exist.
Here are more specific business questions I came to the table with as I discovered Visual Intelligence (VisI) for the first time
- Who are my top users this year and in the past few months?
- Who are trending upward quickly to become a top user?
- Are there any sudden spikes in usage where there could be a problem?
- Who is not making good use of their page views?
With any new insights learned from Visual Intelligence, I wanted to share my findings with my sales director so she could take necessary action.
I had one data set with over 50,000 records that report page broken down by license and authentication keys. This data was downloaded from our page view data database.
The second data set came from our Salesforce.com CRM, which contains all of the accounts and developer licenses.
MY FIRST GUT REACTION
The first time, I opened Visual Intelligence, the interface was pretty basic, but as I would find there is a lot of power to be unlocked within this tool.
IMPORTING AND MERGING MY DATA
Importing data was very simple since it originated from a CSV file. Merging was also quite simple as VisI actually tries to figure out which columns to merge by. An analyst or business user who is not intimately familiar with all of the data columns, would definitely benefit from this built-in intelligence.
Once the data is merged, I can see everything together in one table. A very good start!
Initially, VisI does not automatically convert columns to measures, so I can go through and pick out my measures on my own. Otherwise I was able to click on a button titled “Enrich Data.” VisI will figure out all of my measures and add them to a list in the left pane.
One thing I wish the VisI team would enhance: The attributes list does not indicate which data source an attribute originated. For some users that is not important, but in my case it would have been nice to know so I chose the correct measure.
WORKING WITH TIME DATA IN VISUAL INTELLIGENCE
With all of my raw data ready to go, I wanted to get myself to a monthly or yearly breakdown, rather than daily. I went to the dates column looking for a way to summarize the data by month or week. It was not completely obvious how to work with dates. After clicking around for a while, I found the Manipulation Tools window on the right side of the screen. From there, with my date column selected, I figured out how to duplicate my date column, and then convert them to months, years, and even quarters.
I also figured out that a DateTime fields offers an option to quickly add new columns with day, month, year, etc. I could not figure out why my basic date field did not offer these options forcing me to manually duplicate my fields. I would never expect a business user to ever figure this out without some instruction or training, yet summarizing data by date is critical functionality. I have seen other reporting and dashboard tools deal with dates more elegantly.
With my month and year attributes now available, it opened up an option to create a time hierarchy Now we are cooking!!!.
BUILDING MY FIRST VISUALIZATION
At this point I am only 10 minutes in even with a learning curve. I think I have my data setup, so I go to the “Visualize” option to see if I can start to answer my questions.
I drag one measure (page views), and then I dragged my time hierarchy into the YAxis dimension. Finally I added my Account name in the “Legend Color” Dimension placeholder to get the following visualization. I tried applying a rank but it botched the date order so I could not quite figure out a way to get the top 10 customers with the following chart view.
With my mouse, I clicked on the latest month quarter and year that I wanted to see. In the upper right portion of the chart, I can exclude or filter to see just those data points. When you filter or exclude, the Filter option in the lower right corner changes so you can modify or delete your selections.
Now I have visualization displaying a single month with all of my customers. However, the visualization is not useful as a single point, so I switched to a bar chart. Because I am concerned with top page view consumers, I sorted my page views in Descending order. This task is easily accomplished by clicking on the measure in the left pane. At this point, I see something that I don’t like. My third highest user of page views for last month is an un-identified account. I select this single point and refine my filter even further.
To navigate your data, you will rely heavily on the left pane with your measures and dimensions. As you filter and drill, you can remove these filters to get back to a point where you can see multiple months, quarters, or years. In my case, with one single unrecognized account selected, I wanted to see what specific keys belonged to this account. To do this, I added a second dimension and cleared out the quarter and month filters to get historical context to all license keys that belong to the unidentified account.
IMPORTANT NOTE: Navigating around takes a little time to get used to.. I made a huge mistake of getting to a point and then hitting “Reset”, which completely cleared my work and killed my thought process. I had to start over to get back to the thought plateau I had previously reached. It is easy to do but a real headache after doing it a couple of times. I hit the reset button because the user interface does not give me a clear way how to back out of where I was at.. A simple back button would be great, especially when you drilling a time or geographic hierarchy.
MY EXPERIENCES TRYING TO ANSWER MY ORIGINAL QUESTIONS
VisI is fairly easy to get used to and is not difficult to learn. However, since this is being pitched as “business user tool” I wish the experience was more fluent. For example, there are buttons and menus all over the screen so I found myself on multiple occasions having to re-teach myself how to use the tool even when I came back to gather my screenshots and re-trace my steps for this article.
One of the questions I came to the table with was what customers are moving up in page views the quickest? I was not able to come up with a good way to answer this question because it required me to obtain growth figures which does not exist in my data. Perhaps with some formulas this is possible, but I was not sure where to start because the documentation is barely existent and Google turned up with nothing. Other trending capabilities would be nice.. Typically in addition to visualizing data, I want to know what the growth or delta over time is. Perhaps best-fit lines or accumulated growth functionality would enhance Visual Intelligence past visualizing your data.
I could have gotten the same answers in Web Intelligence that I did in VisI. The big difference is that VisI allows me to manipulate, merge, and visualize data without a Universe, and most importantly without Excel. The basic questions I answered in VisI with 15 minutes of work would have taken me 30 minutes or more in Excel, and that is coming from an Excel guru. The more I want to dig, the more time I save, which is a powerful proposition for someone like me who wants to explore the data in great detail.
After my first pass, I merged a third file containing the Address data for our GMaps Plugin accounts. While it was nice to get them at the City and Country level, the use of maps in this visualization medium was not useful for communicating meaningful information. Geo-visualization is a deep rabbit hole as I know from personal experience, so I will be curious to see how the SAP team can make this visualization more powerful.
Publish to HANA, Streamwork, or Email are your export options right now which had me wanting more. I can only imagine that this is an area that SAP is working on for improvement along with connectivity to the BOE platform. A tool like VisI should allow users to share models more easily. For example, I wanted to share my VisI model with my sales director and had no simple way to do so without emailing screenshots.
Another example how sharing and output came up short, is when I did find a problem, I was unable to capture the 24 digit authentication keys at once. Instead, I had to click on each point one by one and paste the keys from the legend into notepad.
WHO IS THE USER?
My biggest question is who exactly is a VisI user? A business user is going like this, but I can see the novelty wearing quickly as they have to re-teach themselves how to use it every time. An analyst is going to want more power for finding trends and exporting the data back to Excel. I can see a power business user picking this up, running with it, and loving it. As a technologist, business owner, and data visualization guru this is the perfect tool in my arsenal.
MY FINAL RECOMMENDATION
While evaluating VisI, I actually did find a legitimate business problem that I knew existed but didn’t realize how bad the severity was. When I started playing around with the visualization and looking at its origin and growth, I was able to actually gain new insight I never had before… If that is not a smashing success, I don’t know what is!
SAP is offering VisI for a free evaluation as a part of their data geek promotion. Check it out for yourself and let me know what you think.