Using Excel With External Data – What’s the Right Tool?

Excel has been used with external data for… well, as long as I’ve been using Excel. So why would anyone bother to write a blog post about this given that the capability is so mature? In recent years, Excel has adopted a number of new, and frankly better mechanisms for working with external data, while retaining the old. Given that there are now multiple tools in Excel for working with external data, it’s not always clear as to which one is the best, and unfortunately there is no single tool that wins over all, although I believe that that will be the case soon.

The answer, as always is, “it depends”. When it depends, the important thing is to understand the strengths and weaknesses of each approach. With that said, let’s have a look at all of the options.

ODC Connections

ODC (Office Data Connections) are the traditional method of accessing data in Excel. You can create or reuse an ODC connection from the Data tab in the Excel ribbon.

When using an ODC connection, you establish a connection with a data source, form some sort of query and import the resultant data directly into the Excel workbook. From there, the data can be manipulated and shaped in order to support whatever the end user is trying to do. The one exception to this behaviour is the connection to SQL Server Analysis Services (SSAS). When a connection is made to SSAS, only the connection is created. No data is returned until an analysis is performed (through a pivot table, chart etc), and then only the query results are retrieved.

When the workbook using an ODC connection is saved, the data is saved within it. In the case of an SSAS connected workbook, the results of the last analysis are saved along with it. For small amounts of data, this is just fine, but any large analysis is bound to quickly run into the data limits in Excel which is 1,048,576 rows by 16,384 columns in Excel 2013. In addition such a file is very large and extremely cumbersome to work with, but even as such, Excel has been the primary tool of choice for business analysts for years.

Data loaded into the workbook can be refreshed on demand, but it can also be altered, shaped, mashed up, and as is too often the case, grow stale. Workbooks such as these have become known as “spreadmarts” and are the scourge of IT and business alike. With these spreadmarts, we have multiple versions of the same data being proliferated, and it becomes harder to discern which data is most accurate/current, not to mention the governance implications.

SharePoint has provided a way to mitigate some of the concerns with these connections. SharePoint itself supports ODC connections, and therefore users can access these workbooks stored within SharePoint and it also allows them to refresh data from the source either on demand or on open. A single point of storage along with a measure of oversight and browser access helps to restore a modicum of sanity to an out of control spreadmart environment, but the core issues remain.

In order to help with the core issues, Microsoft introduced PowerPivot in 2009.

PowerPivot Connections

Created in PowerPivot

PowerPivot was originally (and still is) an add-in to Excel 2010, and is a built in add-in to Excel 2013. PowerPivot allows for the analysis of massive amounts of data within Excel, limited only by the memory available to the user’s machine (assuming a 64 bit version). It does this by highly compressing data in memory using columnar compression. The end result is that literally hundreds of millions of rows of data can be analyzed efficiently from within Excel.

You can see that compression at work by comparing the same data imported into an Excel workbook directly, and into a PowerPivot model with a workbook. The following two files contain election data, and represent the maximum number of rows that Excel can handle directly (1,048,576) and 25 columns.

Getting data into the model was originally (and still can be) a completely separate process from bringing it into Excel. PowerPivot has its own data import mechanism, accessed from the Power Pivot window itself. First, click on the PowerPivot tab in Excel and then click manage. If you don’t have a PowerPivot tab, you will need to enable the add-in. If you don’t have the add-in, you have an earlier version of Excel – you’ll need to download it.

Once the PowerPivot window opens, the “Get External Data” option is on the ribbon.

Once the appropriate data source is selected and configured, data will be loaded directly into the data model – there is no option to import that data into a worksheet. Once the data is in, pivot tables and pivot charts can be added to the workbook that connect to the data model much like when creating an ODC connection to Analysis Services. In fact, it’s pretty much exactly like connecting to Analysis services, except that the AS process is running on the workstation.

Created in Excel

PowerPivot, and more importantly the tabular data model was included in Excel 2013. With that addition, Microsoft added a few features to make the process of getting data into the data model a little easier for users that were a little less tech savvy, and may be uncomfortable working with a separate PowerPivot window. That’s actually part of the thinking in leaving the PowerPivot add-on turned off by default.

When a user creates an ODC connection as outlined above, there are a couple of new options in Excel 2013. First, the “Select Table” dialog has a new checkbox – “Enable selection of multiple tables”.

When this option is selected, more than one table from the data source can be selected simultaneously, but more importantly, the data will automatically be sent to the data model in addition to any other import destinations.

Even if the multiple selection option wasn’t chosen, the next dialog in the import process, “Import Data” also has a new check box – “Add this data to the Data Model”.

Its purpose is pretty self-explanatory. It should be noted that if you choose this option, and also choose “Only Create Connection”, the data will ONLY be added to the model, nowhere else in the workbook. This is functionally equivalent to doing the import from the PowerPivot window, without enabling the add-in.

Power Query Connections

When Power BI was originally announced, Power Query was also announced and included as a component. This was very much a marketing distinction, as Power Query exists in its own right, and does not require a Power BI license to use. It is available as an add-on to both Excel 2010 and 2013, and will be included with Excel 2016.

Power Query brings some Extract, Transform and Load (ETL) muscle to the Excel data acquisition story. Data can be not only imported and filtered, but also transformed with Power Query and its powerful M language. Power Query brings many features to the table, but this article is focused on its use as a data acquisition tool.

To use Power Query, it must first be downloaded and installed. Once installed, it is available from the Power Query tab (Excel 2010 and 2013).

Or from the data tab, New Query (Excel 2016)

Once the desired data source is selected, the query can be edited, or loaded into either the workbook, the data model, or both simultaneously. To load without editing the query, the load option at the bottom of the import dialog is selected.

Selecting “Load To” will allow you to select the destination for the data – the workbook, the model or both. Selecting Load will import the data to the default destination, which is by default the workbook. Given the fact that the workbook is an inefficient destination for data, I always recommend that you change their default settings for Power Query.

To do so, select Options from the Power Query tab (2010 and 2013) or the New Query button (2016), click the Data Load section, and then specify your default settings.

Data Refresh Options

In almost every case when external data is analyzed, it will need to be refreshed on a periodic basis. Within the Excel Client, this is simple enough – click on the data tab, and then the Refresh All button, or refresh a specific connection. This works no matter what method was used to import the data in the first place. Excel data connections can also be configured to refresh automatically every time the workbook is opened, or on a periodic basis in the background.

However, workbooks can also be used in a browser through Office Web Apps and Excel Services (SharePoint and Office 365) or as a data source for Power BI dashboards. In these cases the workbooks need to be refreshed automatically in order that the consuming users will see the most up to data when the workbooks are opened. The tricky part is that not all of the connection types listed above are supported by all of the servers or services. Let’s dive in to what works with what.

SharePoint with Excel Services

Excel Services first shipped with SharePoint 2007, is a part of 2007, 2010, and will be included with 2016. From the beginning, Excel Services allowed browser users to view and interact with Excel workbooks, including workbooks that were connected to back end data. The connection type supported by Excel Services is ODC, and ODC only.

Excel Services has no mechanism for maintaining data refresh. However, the data connection refresh options are supported which means that the workbook can be automatically refreshed when opened, or on a scheduled basis (every xxx minutes in the background). Unfortunately, this can come with a significant performance penalty, and once refreshed it is only in memory. The workbook in the library is not updated. The data in the workbook can only be changed by editing the workbook in the client, refreshing it, and re-saving it

Workbooks with embedded data models (PowerPivot) can be opened in the browser, but any attempt to interact with the model (selecting a filter, slicer, etc) will result in an error unless PowerPivot for SharePoint has been configured.

SharePoint with Excel Services and PowerPivot for SharePoint

PowerPivot for SharePoint is a combination of a SharePoint Service application and Analysis Services SharePoint mode. When installed, it allows workbooks that have embedded PowerPivot data models to be interacted with through a browser. The way that it works is that when such a workbook is initially interacted, the embedded model is automatically “promoted” to the Analysis Services instance, and a connection is made with it, thus allowing the consuming user to work with it in the same manner as with a SSAS connected workbook,

The PowerPivot for SharePoint service application runs on a SharePoint server and allows for individual workbooks to be automatically refreshed on a scheduled basis. The schedule can be no more granular than once per day, but the actual data within the model on disk is updated, along with any Excel visualizations connected to it.

When the refresh process runs, it is the functional equivalent of editing the file in the client, selecting refresh all, and saving it back to the library. However, there is one significant difference. The Excel client will refresh all connection types, but the PowerPivot for SharePoint process does not understand Power Query connections. It can only handle those created through the Excel or PowerPivot interfaces.

Power Pivot for SharePoint ships on SQL Server media, and this limitation is still true as of SQL Server 2014. At the Ignite 2015 conference in Chicago, one of the promised enhancements was Power Query support in the SharePoint 2016 timeframe.

Office 365

Office 365, or more precisely, SharePoint Online supports Excel workbooks with ODC connections and PowerPivot embedded models in a browser. These workbooks can even be refreshed if the data source is online (SQL Azure), but they cannot be refreshed automatically. In addition, only ODC and PowerPivot connections are supported for manual refresh. Power Query connections require Power BI for Office 365. In addition, Office 365 imposes a 30 MB model size limit – beyond that, the Excel client must be used. In short, the Office 365 data refresh options are very limited.

Power BI for Office 365

Power BI for Office brings the ability to automatically refresh workbooks with embedded data models. Data sources can be on premises or in the cloud. On premises refresh is achieved through the use of the Data Management Gateway. It also raises Office 365’s model size limit from 30 MB to 250 MB. With Power BI for Office 365 both manual and automatic refreshes can be performed for both PowerPivot and Power Query connections, however Power Pivot connections are currently restricted to SQL Server and Oracle only.

The automatic refresh of ODC connections is not supported. A workbook must contain a data model in order to be enabled for Power BI.

Power BI Dashboards

Power BI Dashboards is a new service, allowing users to design dashboards without necessarily having Office 365 or even Excel. It is currently in preview form, so anything said here is subject to change. It is fundamentally based on the data model and it works with Excel files as a data source currently, and it is promised to use Excel as a report source as well. The service has the ability to automatically refresh the underlying Excel files on a periodic basis more frequent than daily.

In order for a workbook to be refreshed by Power BI, it must (at present) be stored in a OneDrive or OneDrive for Business container. It also must utilize either a PowerPivot, or a Power Query connection. At present, the data source must also be cloud based (ie SQL Azure) but on premises connectivity has been promised.

SQL Server Analysis Services

Another consideration, while not a platform for workbooks is SQL Server Analysis Services (SSAS). Excel can be used to design and build a data model, and that data model can at any time be imported into SSAS. As of version 2014, SSAS fully supports all connection types for import – ODC, PowerPivot and Power Query. Once a data model has been imported into SSAS, it can be refreshed on a schedule as often as desired, and you can connect to it with Excel, and share it in SharePoint. You can also connect to it in Power BI Dashboards through the SSAS connector. From both a flexibility and power standpoint, this is the best option, but it does require additional resources and complexity.

Refresh Compatibility Summary

For convenience, the table below summarizes the refresh options for the different connection types.

 

ODC

PowerPivot

Power Query

Excel Client

M

M

M

SharePoint/Excel Services

M

SharePoint/Excel Services/PP4SP

M

A

SQL Server Analysis Services Import

A

A

A

Office 365

M

M

Office 365 with Power BI

A*

A

Power BI Dashboards

A

A

M – Manual refresh

A – Both Manual and Automatic Refresh

* only limited data sources

 

The Right Tool

I started out above by saying that the selection of import tool would depend on circumstances, and that is certainly true. However, based on the capabilities and the restrictions of each, I believe that a few rules of thumb can be derived. As always, these will change over time as technology evolves.

  1. Always use the internal Data Model (PowerPivot) when importing data for analysis.

     

  2. Power Query is the future – use it wherever possible

    All of Microsoft’s energies around ETL and data import are going into Power Query. Power Query is core to Power BI, and announcements at the Ignite Conference indicate that Power Query is being added to both SQL Server Integration Services and to SQL Server Reporting Services. Keep in mind that we have been discussing only the data retrieval side of Power Query – it has a full set of ETL capabilities as well, which should also be considered.

  3. PowerPivot or ODC Connections must be used on premises

    PowerPivot for SharePoint does not support Power Query for refresh. This means that you MUST use PowerPivot connections for workbooks with embedded models. If you are already using SSAS, use an ODC connection within Excel.

  4. Power Query or PowerPivot must be used for cloud BI.

    PowerPivot connections will work for a few limited cases, but more Power Query support is being added constantly. Where possible, invest in Power Query

  5. If on-premises, consider importing your models into SSAS

    SSAS already supports Power Query. If, instead of using PowerPivot for SharePoint, Analysts build their models using Excel and Power Query, they can be “promoted” into SSAS. All that is then required is to connect a new workbook to the SSAS server with an ODC connection for end users. The Power Query workbooks can be used in the cloud, and the SSAS connector in Power BI Dashboard can directly use the SSAS models created.

  6. Choose wisely. Changing the connection type often requires rebuilding the data model, which in many cases is no small feat.

In summary, when importing data into Excel, the preferred destination is the tabular model, and to import data into that model, Power Query is the preferred choice. The only exception to this is on premises deployments. In these environments, consideration should be given to connecting to a SSAS server, and failing that, PowerPivot imports are the best option.

Advertisements