This blog has been on my mind for a long time now. I thought I’d share part of the analytical process with you that I performed to determine both the offer and sale price of my own home, by far my most expensive investments. Being the extreme data addict that I am, I enjoyed reviewing the local real estate market trends in a variety of data discovery tools including Microsoft Power BI, Tableau, Spotfire, QlikView 11, SAP Lumira and Microstrategy Desktop. What was interesting about this exercise is that I took the same tiny data set and rudimentary use case, simply ran it in various tools, capturing the differences, strengths and weaknesses of each tool. Granted this specific use case does not ideally showcase a few of the data discovery tools in this mix, it was quite revealing.
I began looking for a new home about six months ago after one of my recently retired neighbors made too much noise on a daily basis. Since I often work from home, I need a quiet environment. I also love nature and dreamed of a serene yard filled with lush trees, tropical flowers, birds, butterflies, tree frogs, crickets and other critters. I did find that home and moved into it in late July. As I settle in, I am now anxiously awaiting the inspection, appraisal and closing of my other home. Throughout this process I had a few skeptical conversations with realtors. Local realtors were advising me to sell my home for less than $100 per square foot while at the same time telling me to make offers of over $140 per square foot for other homes in similar condition in my same general neighborhood. I am not an exceptionally trusting individual and thus gathered the raw data myself from realtors, Zillow, Trulia and Realtor.com for home pricing analysis. My final comparable list ended up being a small set of 17 recently sold homes that I entered into an Excel 2013 file with approximately 20 variables.
If you ever purchased or studied real estate, you most likely have learned how real estate appraisers reference comparable properties to estimate the value of a home. It is straightforward but also subjective based on many factors from location, yard, interior materials, age, and so on. To keep this blog and the following data discovery tool comparison super easy to follow, I will merely use sold price per square foot of comparable homes as the key metric for pricing analysis. In reality, you would absolutely include appraiser level variables in home pricing decisions.
Microsoft Power BI
Since I was already in Excel 2013 Professional Plus, I formatted my comparable list as an Excel table and then inserted a Power View sheet. In a mere few seconds Power View easily mapped the comparable home, address level data, without requiring me to geocode the latitude and longitude. Address level mapping was not available in any other data discovery tool that I explored out-of-the-box per se. To get a map of the comparable homes with all the other tools, I had to manually look up latitude and longitude.
After mapping comparable homes data, I wanted to assign price range bins or categories so I could see them more easily on the map. Using Power View I was able to apply colors by sold price per square foot bin and also size the comparable home markers by sold price in a few clicks. Most of the other tools reviewed also provided features for this kind of analysis.
To review pricing trends over time, I was able to add a line chart by home sold date. Unlike several other tools reviewed including Tableau, Spotfire and SAP Lumira, I was unable to view future trends on the desktop Excel Power View tool or see the statistical trend ranges. Power BI sites in the Office 365 cloud offer linear trending/linear forecasts but not local Excel Power View. The Power View chart options currently are not as deep as the other tools in analytical features. I was also not able to assign point level annotations for foreclosures and short sales (pricing exceptions) like I could with several other tools. For annotations, I could add a text box to the Power View canvas.
Lastly I wanted a nice detail list that included a picture of the comparable home. In Power View, adding a picture can get a little tricky. Today you need to use image blobs and embed them with SQL Server load code in Power Pivot, or you can use a URL and upload images to a web site. Since I am also a hobbyist web site developer, I uploaded the images to my own web servers. You could alternatively store images to an Office 365 SharePoint site for usage in Excel Power View shared via Power BI on Office 365. If you don’t embed or store them, you get ugly error warnings. Bizintelligist recently covered this specific “using images with Power BI” topic on his blog.
To pretty up my findings, I applied a theme, added a title and a header image. I totally love Excel 2013 Power View’s capability to place content anywhere on the canvas. Power View does not suffer from what I call dreaded “BI boxy syndrome”. “BI boxy syndrome” is a terrible limitation that forces dashboard designers to pick a box layout that does not allow content overlapping or exact content placement. The term may seem funny to you but “BI boxy syndrome” is a serious bummer to me…and it is still an issue with most modern BI and data discovery tools that I review! I crave creative design freedom to make dashboards and analytics beautiful and useful. Only Power View and Tableau truly delivers on this dashboard design need today in the mix of above tools.
I was disappointed in Power View’s inability to control title text color and many other granular dashboard settings. I do hope in the future that there will be more control over the little things in Power View like there are in many other tools. Aside from the image hoop-jumping exercise in Power View, I was able to quickly analyze home comparable prices and learned that my realtors were not too far off in their home offer advice. Homes that were a few blocks away in Tampa Palms were indeed selling for $120 to $140 per square foot. The realtor home selling advice was off a bit. Homes in my immediate neighborhood were selling at $105 to $115 per square foot. Bottom line, Power View shined brightly for this simple use case. It was easy, fast and fulfilled my basic analytical needs.
Working through the same exercise with Tableau (mapping homes, viewing trends over time, applying analytical features, adding images and making my analysis look good with design and theming), I noted the following differences in experience. For mapping comparable homes with Tableau, I had to first manually look up latitude and longitude. If I did not provide latitude and longitude in my data set, the address data all overlapped on the same zip code and was useless. After I geocoded the data, I was able to create a nice map with many different display options. Tableau has more map display options, backgrounds, layering, WMS and so on than most other tools reviewed.
Next I plotted the price trend line and was able to add additional trend and trend range analytical features. In Tableau I could also see if the trend was statistically significant and review the algorithm details. I totally love that feature.
Creating the list with home images using Custom Shapes in Tableau was easier than Power View but a bit tedious. Out of all the tools reviewed, Tableau’s approach and ability to use local images did seem the most straight-forward. Spotfire and QlikView also had comparable image in a list display capabilities. SAP Lumira and Microstrategy Desktop did not have that feature as far as I could tell – or – I gave up trying to find it in those tools.
As far as design and theming goes with Tableau, it is by far the best in this mix of tools. Tableau does not suffer from “BI boxy syndrome”. Tableau, much like Adobe Photoshop, allows you to granularly control dashboard content placement, font styles, theme colors and many other details. You can even create your own color palettes to exactly match your branding colors. Rock on Tableau engineering – I do appreciate the huge world of analytical creative freedom that you allow me to enjoy!
To be 100% transparent, I do like TIBCO Spotfire a lot. I believe Spotfire is a genuine analytic gem that does not seem to reach the right target audience effectively in all the data discovery noise. If you love analytics, you will most likely also appreciate the beauty and depth of Spotfire. Much like Microsoft, SAP and Microstrategy, TIBCO struggles with an enterprise-heavy, IT led sales model. Tableau and Qlik are much better at reaching business users and ironically suffer from the reverse issues in expanding to become enterprise IT standards.
For mapping data using Spotfire desktop tools, I did need to provide geocoded address level data. Base Spotfire comes with a few geocoded data sets to cross reference and also has a fancier mapping offering that I do not have in my installation.
When it came to the price trends, Spotfire had excellent analytical feature sets for forecasting with confidence interval bands and many other options. This is an area where Spotfire is far better than all the other data discovery tools reviewed.
When it came to adding the comparable home images, Spotfire was able to display them in the list. The process to create a matrix with the image URL took me a while to figure out. That process seemed comparable to the Power View process. It was not as obvious as Tableau or Qlik’s image data type assignment.
For design and theming, Spotfire is really flexible with regards to customizing colors and fonts. It also has fantastic granular level control on charts and chart options. Where Spotfire stumbles a little is in the area of “BI boxy syndrome”. Although I was not limited to box layouts, I was not able to have fine control on dashboard content placement. I was frustrated with the header image default display and other defaults that mess up the page. I’d rather not see all the chart options on by default. It gets tedious clicking and hiding unused options each time you add a chart. Spotfire was capable of meeting and exceeding my basic analytic needs for this home pricing analytic exercise.
At the time of analysis, Qlik Sense was not yet available. I will have an upcoming blog on it and expect that new Qlik Sense will be a much nicer development experience than Qlik v11. For this specific use case, Qlik v11 was used and it did not shine brightly. I was not able to get the mapping to work at all. It looked like mapping is more of a developer level experience to get functional. I did get comparable home images to display in a list fairly easily along with a simple line chart of sold prices by date. I did not test line trends or expressions in Qlik but noted that more sophisticated analytical expressions, custom colors, font controls and theming features were indeed available. Qlik’s development experience was wizard-heavy.
All in all, the Qlik v11 development experience paled in comparison to Power View, Tableau and Spotfire. To be fair to Qlik, I am a newbie and I may not have known what I was doing. Since my Qlik work was not at all lovely, I intentionally did not include a picture of it here. I will dig deep and learn Qlik Sense that has already impressed me in demos and my preliminary playing. As my readers know, I have been rough on Qlik in the past. So yes, I am openly admitting that Qlik Sense is nice. They have completely redesigned the development experience and also are better targeting Qlik Sense for the business user. Qlik v11 was really for a BI/IT developer user despite the Business Discovery marketing, positioning and messaging. I do recognize that Qlik Sense will be a data discovery force to be contended with this coming year or longer.
On to SAP’s constantly updating Lumira offering. I will share that I was shocked to see infographics and reports in SAP Lumira recently. SAP is seriously moving fast, delivering new SAP Lumira features every month. This solution has come a loooooong way since May 2013.
For this simple exercise, SAP Lumira did not perform highly. I was not able to create a map with the address level data even though I geocoded it. I ran into issues trying to assign latitude and longitude to the comparable home geocoded fields. UPDATE August 8, 2014: The mapping issues in SAP Lumira were really bothering me so I tried one more time. This time I did get the default free mapping that comes with SAP Lumira to finally show the comparable homes that I geocoded but it was not at all usable. See image below of the lowest level that I was able to zoom into with default SAP Lumira mapping. Thus I started up a trial of the paid mapping ArcGIS ESRI service that you can set up in SAP Lumira preferences to give that a whirl. The ArcGIS ESRI paid add-on maps were much better for visualizing address level data for this use case.
I also was not able to get an image to display in a list quickly. I could easily add a line chart and also review an automatically generated forecast. SAP Lumira did have ok control of colors and fonts but not completely granular control like Tableau, Spotfire and Qlik.
I did not care for the SAP Lumira dashboard design and layout experience. SAP keeps changing it and quite frankly it feels like the dashboard design is not in the right place now. It just feels odd to me…jumping back and forth between screens. I am guessing that the next time I open SAP Lumira that annoying design experience could progress. In the mix of tools I played with for my personal real estate use case, SAP Lumira ranked the worst. Keep in mind that there are many other really cool SAP Lumira improvements that are not at all covered here including valuable data prep. This elementary test of SAP Lumira and poor results should not prevent you from giving it a try. It does have visualization API capabilities and many other features that I have not touched on at all in my scenario.
Oh suffering Microstrategy…I feel bad covering them in this mix but someday they could reign supreme again like they did in the good ol’ days of enterprise BI dominance. For my exercise, I used the free Microstrategy Desktop. I will share that even though I did not review data source connectors, I noticed Microstrategy actually has a really awesome array of data sources out-of-the-box including data virtualization vendors that I have not seen as a data source anywhere else.
Back to my basic use case. For the mapping, Microstrategy was able to render the comparable homes on a geospatial map. It looks like they use ESRI for mapping, a highly respected mapping vendor. For the price trend line chart, that was simple to create but I did not notice the ability to add a forecast or other analytical features to that line chart. I also was not able to get images to display in a list. I might have missed something but I did search, reviewed help docs and could not figure it out.
Microstrategy Desktop has limited design and theming capability. It also suffers from “BI boxy syndrome” with a miserable case of Flash rendering. I don’t know what the future holds for Flash but most groups have already shifted to using HTML 5 for rendering dashboards in a mobile-loving world. Microstrategy used to be the leader in Mobile BI. I don’t know how this personal Flash UI asset can be used with Microstrategy Mobile BI in a design once and deploy anywhere BI strategy. If I got this part of the assessment wrong, please correct me and I will add the changes. In the meantime, I’d say Microstrategy Desktop fared better than SAP Lumira and Qlik v11 but worse than all the others in this mix for my simple use case. Again, don’t let my test prevent you from playing with this specific analytic tool. I am merely sharing my findings.
As you probably know, there are a bazillion other data discovery tools that I could also review in this mix like IBM Cognos Insight, Targit, DataWatch, Antivia, Panorama Necto, Advizor, Omniscope, Pentaho, Birst, Good, Logi, etc. Many of the final dashboards often do look the same to business users. People ask me all the time what makes a specific data discovery, BI or analytic tool better than another tool, when to use which tool, and on and on. These tools seem like commodity offerings …until you test drive them! To understand the vast differences in data discovery solution capabilities and development user experiences, you do need to evaluate them in a hands-on manner with the same evaluation scenarios. Vendor demos do not reveal the critical differences. When you do test, be sure to include a wide variety of use cases from basic to complex to find the tool limitations…not a tiny data set as I did for my home pricing analysis. Consider involving IT in the process to ensure that you make a great decision, don’t end up with a crummy free analytical tool and an expensive analytical mess after a few months of using it.
If you do want help in evaluating these kinds of tools, please don’t hesitate to reach out to me. I love reviewing analytic solutions, understanding where they best align and I do genuinely appreciate the splendor in many different vendor solution designs. Most importantly if you are buying or selling a home, take a few minutes of time to gather your own comparable home data and run it through one of these data discovery tools to literally save or earn yourself thousands of dollars. Keep in mind that for a smaller 2,000 square foot house, the difference in $5 per square foot means a $10,000 difference to you and only $300 or $600 variance to a realtor that has typical 3% or 6% sales commission motivations.