Tallan Blog

Tallan’s Experts Share Their Knowledge on Technology, Trends and Solutions to Business Challenges

Observations on July 2015 PowerBI Release

The following observations come from the NYC SQL Server Meetup 9/1/2015 presented by Devin Knight of Pragmatic Works, augmented with my own discoveries. They assume the reader has basic familiarity with the prior PowerPivot, PowerQuery and PowerView products.

• This release debuts PowerBI Desktop (PBID). The significance of this is that PowerBI is no longer tied to Excel or Sharepoint, thus it no longer requires IT involvement (aside from giving access to data sources). It is a stand-alone app ala (but not part of) Office 365.  It is basically a bundling, divorced from Excel, of the prior PowerQuery, PowerPivot and PowerView products. Look and feel is very similar to these. It can be downloaded for free from www.powerbi.com

• The disconnect from Excel means it can have its own release cycle and not be constrained by having to render through Excel. There have reportedly been PowerBI problems that are really due to bugs in Excel, which did not get fixed because that team had other priorities. Such will no longer happen. The PowerBI team says they are aiming for a release every six weeks.

• The PBID Motto is “Five minutes to WOW”.

• Partnerships with other vendors for data are emphasized, e.g. with Salesforce.

• You can create dashboards in a similar manner to PerformancePoint, i.e. drag and drop of charts onto a palette, resizing and rearranging them; hooking up filters to them and so forth.

• PBID is claimed to be mobile-ready on ixxx, Android and Windows. You can pick which parts of a dashboard you want to see. It also supports setting up alerts and adding annotations. Just how well this really works remains to be seen.

• There are some new chart types, including a map chart.

• There are some new data sources, including Facebook and web pages (where HTML tables are scraped).

• Competitors include Tableau, QlikTech and Spotfire.

• SSIS in SQL 2016 will in some way integrate PowerQuery.

• At present PBID does not support hierarchies or KPIs. It is expected it will at some point.

• PBID includes a functionality called “Merge Table”. Suppose you have imported tables A and B. For convenience, you just want to deal with table A, as it has most of the columns you want, but B has one or two. You can do a “merge table” on A and define the join between A and B, selecting which columns from B you want. From then on A has the columns from both and you can ignore B. There is even a new language “M” (for Merge) which it generates to describe these operations. You can edit this code for special cases. You could do a similar thing in PowerQuery by adding calculated columns to a table that are derived via relationships from other tables, but this new UI is much more intuitive.

• PBID can do table pivot/unpivot operations in an impressive drag/drop fashion.

• Sharing of results built in PBID has various possibilities and limitations:

  • It saves as a file type analogous to .xlsx and this can be shared like any file – but consumers also need PBID to view it. It cannot save as PDF.
  • Amazingly you cannot export data as CSV or Office formats such as Excel and PowerPoint.  You can cut/paste, but this is cumbersome.  A glaring omission.
  • Rather than Sharepoint as previously, the work can be uploaded to the cloud. Powerbi.com acts as the host for this. You then can share with others in your corporate domain via their email address. You cannot share outside your corporate domain. This is the extent of security/entitlement. Doubtless not all clients will be happy with this collaboration model.
  • Eventually it will be integrated with Sharepoint again, adding that as an option allowing more generalized sharing and access control.
  • You cannot embed PBID dashboards in arbitrary web pages.
  • There is another artifact called “Content Packs” which are a form of sharing.

• PBID comes in free and paid ($9.99/mo) flavors. The main differences between free and paid relate to:

  • The maximum size of dashboard+data you can upload to the cloud.
  • The rate at which data uploads.
  • How frequently the data can be refreshed once in the cloud.

• PBID includes Q&A – Microsoft Research’s natural language query language. You can type in things like “Revenue by Year as Barchart” and it just does it. Interesting but still seems like something of a toy.

• PBID supports M:M relationships, although non-intuitively to data professionals: you first define 1:M relationships in both directions and then set a dropdown to “Both”. This probably reflects its self-service orientation.

• There is no integration with Reporting Services, although RS is getting a big overhaul in SQL 2016 which may include some.

• Drilldown is not really supported – not via hierarchy as previously noted, and drillthrough to detail, as understood in Analysis Services, is not supported.

• PBID includes the PowerQuery Excel add-in, which can implement “query folding” – built-in smarts which, when possible, render the native queries that are generated in an intelligent fashion to minimize the data that needs to be downloaded to the tool – for instance, SQL with constraints when going against relational sources.  This can be an important optimization but there are many limitations to when it can be used.

• When getting data, PBID’s primary method is Import – which copies the data source in to PBID.  In this mode the data set must be recopied into PBID when many actions are performed, which can slow things down with large datasets.  For a limited set of relational sources you can also choose Direct Query, which means a direct connection to the data source is created, and real-time queries are issued as you interact with design and consumption, rather than a copy of all the data being first brought to PBID.  This is useful if the data must always be current – i.e. real-time.  However, this mode also has serious limitations, including an inability to create calculated columns or measures !

• Ability to customize graphics is better than previously but remains woeful – e.g. chart legends can be turned on and off, some colors changed from the default, etc. You still cannot change fonts, placement, format or much of anything else. It is very reminiscent of Performance Point dashboards in this regard. Reporting Services has its limits in these regards also but is much more customizable.  PBID still does not support sparklines, and what most would recognize as heat maps are called “tree maps”.

• DAX is still supported for measure definitions.

• The PowerPivot side of PBID defaults all numeric columns to SUM aggregations, including nonsensical ones like surrogate keys and years. This means all such appear with sigma symbols in the pivot table field list, which is silly-looking at best and confusing at worst. From a self-service perspective this is unforgivable.


  • PowerBI Desktop certainly represents an improvement over its predecessors. It should be kept in mind that its target audience is self-service BI, thus end users, and that its lack of many capabilities which seem ludicrous may in fact be deliberate and desirable, to keep things simple.
  • The ability to stitch disparate data sources together without IT help, which has been present since the original PowerPivot, is very compelling for that use case.
  • The graphics options continue to be very substandard, almost all with very low information-to-square-inch ratios. Microsoft seems determined to ignore the no-longer-new recommendations of Tufte and Few for effective information graphics – which demand not only particular types of graphics (e.g. sparklines) but fine-grained customization capabilities. I have yet to see a PowerBI demo where the graphics are in any way compelling, and that hasn’t changed. Instead, clunky is what I usually think of.
  • The lack of hierarchy support is a serious limitation we can only hope gets rectified soon.
  • The inability to export to standard Office and other formats is a serious limitation, likely a show-stopper for many users, and an Achilles Heel versus competitors.
  • The limitations imposed by using Direct Query to connect directly to a (supported) relational source may seriously impede its usefulness.  But the alternative of importing a file-based data set will not be practical in many cases.
  • Microsoft’s “Five Minutes to WOW” motto for PBID is laughable.  For all that has been improved, the tool still has serious limitations that will rule out its usefulness in many scenarios.  While I am not an expert in competing tools I believe they do not share many of the most serious limitations.
  • Nonetheless, PBID probably has real value to developers as a prototyping tool for envisioning and even for quickly determining whether what a client wants analytically is feasible (e.g. whether the necessary data exists).
Share this post:

No comments

Leave a Reply

Your email address will not be published. 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>