Alex is founder of Data Driven Consulting, an independent consulting company that focuses on obtaining high quality data to facilitate effective business decisions. Data Driven Consulting specializes in a number of areas, including market research, dashboard and data presentation, and authoring of white papers and e-content.

One of the more anticipated functions of Excel 2010 is Microsoft’s implementation of sparklines (as defined by Edward Tufte – data intense, design-simple, word-sized graphics). Although the recent patent application made by Microsoft would seem to imply that they are the inventors of in-cell sparklines, both Edward Tufte and the software vendors who have offered Excel sparkline solutions would beg to differ.

One of these solutions, Sparklines for Excel (SFE), is favored by many, is free and open-source, and works well. There are versions for Excel 2003 (and before), and Excel 2007, with the latter version adding a new section to the ribbon. Below are screenshots of the implementations – click on these reduced images to open a new window showing the full size sparkline tabs.

Excel 14 Native Sparkline ribbon tab

Sparkline For Excel Add-In Ribbon Tab

Given the anticipation of Excel 2010’s sparklines it makes sense to compare the two offerings. This table compares the functionality of the two options.

	Sparklines for Excel			Excel 2010
Types of cell charts	14 and growing: Line, Pie, Pareto, Bullet, XY, Scatter, Bar, Column, Stripe…			Three: Line, Column, Win/Loss
Overall look (sparklines)	Some blocking, not noticeable at normal zoom levels			Smoother
Features on individual chart types	Excellent, whole range of markers like hi/low lines, max/min points			Limited to some basic additional data indicators, but they work well
Ease of deployment	Have to distribute macros with file, some VBA crashes, mostly during creation. Win today, as no built-in option			No extra package, but Excel 2010 needed. Win in 2-3 years when 2010 is ‘standard’.
Ease of use	More complex, for  example looking up meaning of formula variable, color chart			Easier, native to the product, so formatting is quick
Code Customization and Stability	VBA macro based (non-locked) so customization possible (and legal)			Not VBA based – more stable, but no customization
Grouping	Can set max and min so that multiple charts scale the same			Adjacent charts are grouped automatically making for easy changes of formatting/scales
Ease of editing	Charts are shapes created on the basis of formula in the cell			Effectively separate charts, edited with own section on ribbon
Future additions	Constant addition of new chart types/upgrades to existing			Maybe some fixes, but will likely have to wait until Excel 20XX for new chart types
Cost	Free, but you need Excel of course			Free, at least after you’ve paid any upgrade costs

While there are equal numbers of ticks in both columns, there are clear winners depending on your required functionality and how you intend to share the information. For creating dashboards that benefit from different types of in-cell charts and line markers, SFE is still a clear winner – the range of chart types, the continued development, along with the fact that the sparklines are formula based makes for a powerful tool. Distribution is complicated by the need to load the add-in file, but for dashboards that shouldn’t be a barrier.

On the other hand, for a quick summary of data that is emailed around, the native version in Excel 2010 is clearly simpler. In the near term (until 2012 I would guess), SFE is still really the only option anyway as sparklines created in Excel 2010 simply won’t show on earlier versions.

It’s great that Microsoft has offered sparklines – hopefully end-users will appreciate their value more now they are a native part of the program, and become used to sparklines’ inclusion in their data lives.

Peltier Technical Services, Inc., Copyright © 2011.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

I decided a comparison was in order between Excel 2003 and Excel 2010. I have Excel 2003 SP3 and Excel 2010 TP installed under Windows XP Pro SP2, on a three-year-old Presario laptop (which has miraculously recovered from a bad crash) that has 2MB of RAM and a 1.79 GHz AMD single-core processor.

I used the following macro to automatically change the chart’s data range, and measure how long it took to redraw the chart. I started at -2 just because the code seemed to run slowly the first loop or two, and I wanted it to warm up. I adjusted the code to account for the much smaller limit of points per series in Excel 2003.

Sub point_timer()
  Dim i As Long
  Dim i10 As Long
  Dim t As Double

  t = Timer
  For i = -2 To 12
    i10 = IIf(i < 0, 1, 10 ^ (i / 2))
    ActiveSheet.ChartObjects(1).Chart.SeriesCollection(1).Values = ActiveSheet.Range("B1")
    ActiveSheet.ChartObjects(1).Chart.SeriesCollection(1).XValues = ActiveSheet.Range("A1")
    DoEvents
    t = Timer
    Application.ScreenUpdating = False
    ActiveSheet.ChartObjects(1).Chart.SeriesCollection(1).Values = ActiveSheet.Range("B1").Resize(i10)
    ActiveSheet.ChartObjects(1).Chart.SeriesCollection(1).XValues = ActiveSheet.Range("A1").Resize(i10)
    Application.ScreenUpdating = True
    DoEvents
    Debug.Print i, i10, Timer - t
  Next
  ActiveSheet.ChartObjects(1).Chart.SeriesCollection(1).Values = ActiveSheet.Range("B1")
  ActiveSheet.ChartObjects(1).Chart.SeriesCollection(1).XValues = ActiveSheet.Range("A1")

End Sub
 

Here are Excel 2010 and 2003 charts showing my data (X = Row and Y = 1/Row) for 10 points:

Excel 2010 chart with ten points

Excel 2003 chart with ten points

Here are Excel 2010 and 2003 charts showing their respective maximum number of points:

Excel 2010 chart with 1,048,576 points

Excel 2003 chart with 32,000 points

I’ve just noticed that the chart export feature has stripped off the axis labels in the Excel 2010 chart.

The timed redrawing results are shown in this table, seconds to redraw vs. number of points:

This is a blog about charting, so I have charted the performance here:

At the low end, both versions show a plateau where redraw times are insensitive to the number of points. Excel 2003 is faster than Excel 2010, but by a factor of two or less. At the high end, both versions converge on the same redrawing vs. points line. The ultra long redraw times I observed last week obviously was not due to inferior redrawing, but to superior amounts of data.

This is good news. It means that the work Microsoft has undertaken to improve charting performance, particularly chart redrawing performance, has paid off.

Peltier Technical Services, Inc., Copyright © 2011.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

In Excel 2007, if the Visual Basic window is open, you still get this dialog. In fact, the dialog shown above is the Excel 2007 version. However, if the VB window is not open, you don’t get the dialog. Instead you get a notice embedded in the UI. It’s a non-modal message, with no beep, and it doesn’t make you respond to it.

See the notice, tucked away between the ribbon and the formula bar? I didn’t think so. Half the time I don’t notice it, and I run around wondering what’s wrong.

I can’t count how often I’ve loaded a workbook with macros, and not known why they don’t work. Or at least I’m annoyed by the next message.

Excel 2010 works like 2007. If the VB editor is open, you get the blatant dialog that forces you to respond. If the VB editor is closed, you still get the sedate little notice in the UI, but it’s no longer camouflaged to blend in with the ribbon.

See it there, between the ribbon and the formula bar? Think you could possibly miss it? Good.

This is one bit of gaudy formatting that I welcome. I always see the notice immediately every time it appears.

Peltier Technical Services, Inc., Copyright © 2011.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

Never mind now long it takes to redraw.

Number of Points in a Series

Despite the limited utility of cramming data points on top of each other, I thought I’d investigate the rumored new limits. Here’s an Excel 2010 chart with 100 points; I’ve used the row number for X and 1/(row number) for Y.

Let’s skip 1000 points, here’s a chart with 10,000 points

Keep piling on. Here’s a chart with 32,000 points, the maximum allowed in a series in earlier Excel versions.

100,000 points, anyone?

Okay, I’ll bite. Here’s a chart with all of column A as X values and all of column B as Y values. 1,048,576 points. And it didn’t choke. It took a while to redraw, of course.

Check out this unique chart series formula. It uses the entire columns in the definitions of X and Y.

Number of Series in a Chart

I tried the same thing with the number of chart series. This is what 100 series look like in a chart. The X values are 1 through 25. Series 1 uses Y values 2 to 26, Series 2 uses 3 to 27, and so forth.

Here’s the same chart with 255 series, the maximum number of series in a chart in prior versions of Excel.

I tried adding more, and learned I’d exceeded the limit.

Summary

The number of series allowed in a chart in Excel 2010 has remained the same as in earlier versions, 255 series, but the number of points in a series has increased dramatically, from 32,000 to 1,048,576, so that entire columns in the larger grid of Excel 2007 & 2010 can be used as the source data of a chart series.

The total number of points allowed in a chart has obviously increased, since the old limit of 256,000 points is smaller than the new limit on points per series. Since drawing one series with a million points took over a minute, I did not have time to explore this limit. (Hmm, I also didn’t check whether the help files have yet been updated with the new limits.)

Peltier Technical Services, Inc., Copyright © 2011.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

I spent my first couple hours in Excel 2010 just driving around, seeing what the terrain looked like. Then I experimented with a new feature, Sparklines, and with a feature that was introduced in Excel 2007 but made useful in 2010, Data Bars.

PTS Utilities in Excel 2010 – Good news!

Finally today I got up the courage to test my commercial utilities in Excel 2010. Aside from very minor things, like building the ribbon interface not just in 2007 but in any version 2007 and later, the utilities have worked just fine.

This screen shot shows the commands for my utilities happily displayed on the PTS Charts tab of the ribbon. In the worksheet below is a waterfall chart created using the unmodified Waterfall Chart utility.

If any PTS Utility customers want to try out their utility in Excel 2010, send me an email and I will reply with a link to an updated version.

PTS Utility in 2007 – Bad News

I was putting the finishing touches on a new utility, the PTS Chart Zoomer Utility. It allows a user to select a region of a chart using the mouse, and the chart axes are rescaled to show a closer view of this region.

Select a region of a chart with the mouse…

… the region expands to fill the chart. Select another region…

… and this region now fills the chart.

I built the ribbon bits for this utility, and I was going through my final testing in Excel 2007 SP2, and nothing worked. No selection rectangle appeared in the chart, no resizing of the chart region occurred, nothing happened. I checked my code: it looked fine. I reran my code in Excel 2003: it worked fine.

I was starting to think, not another Excel 2007 upgrade hassle. I’ve had three unpleasant 2003 to 2007 upgrades of client projects in the past couple of months. The unpleasantness is due to a load of unanticipated incompatibilities, and even my best workaround attempts weren’t always successful. In one or two cases, the code might have changed, so that formerly optional arguments were required (though the documentation didn’t point this out). Mostly, though, the problems were with file formats, template handling differences, and charts. Charts, ouch. Dynamic charts would lose track of the ranges they were linked to. Charts on copied sheets would end up linked to the wrong data sheets. Labels or entire charts would move or disappear.

The problem in this case boiled down to a couple chart events not working. To draw a selection rectangle as shown in the screen shots above, you need to click the mouse (the MouseDown event), drag the mouse (the MouseMove event), and release the mouse (the MouseUp event). MouseMove is working just fine in 2007, but neither MouseDown nor MouseMove are triggered by user actions.

I think I could hack my way around this deficiency if all I were doing were drawing the rectangle and rescaling the axes. It wouldn’t be pretty, but it might work. However, the Chart Zoomer utility also has features that remember and restore previous zoom settings (Ctrl+Left or Right Click), and that resize to fit all points or to restore Excel’s default axes (Alt+Left or Right Click). I can’t figure out how to capture these mouse clicks, so I wouldd have to add a bunch of controls to the ribbon tab, or float a dialog nearby, or generate a bunch of shortcut keys.

I haven’t decided whether or how to proceed with this utility: should I scrap it, release what I have as a Classic Excel only utility, or hack my brains out and release a utility that works or at least mostly works in all versions of Excel?

Peltier Technical Services, Inc., Copyright © 2011.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

Sparklines

There are numerous products which add sparklines to Excel. Two popular commercial products are BonaVista Microcharts and Bissantz SparkMaker, which work by forming the graphics using specially designed fonts. Sparklines For Excel is an open source add-in which works by drawing sets of shapes to construct the sparklines. These are full-featured sparkline programs which accommodate many chart types and styles.

Microsoft has gotten its start in sparklines in Excel 2010. The Microsoft Excel Team Blog has discussed this new feature in Sparklines in Excel, Adding Some Spark to Your Spreadsheets, and Formatting Sparklines. There are only three chart types: line, column, and high-low; and they do not have such features as baselines or axes or shaded zones. However, I think it’s a promising start.

To try out the sparklines, I loaded my blog stats into a pivot table, with months in rows and day of the month in columns. Creating the charts was easy. I selected the block of data to plot, then clicked the Sparklines – Lines button on the Insert tab.

A simple dialog pops up with two RefEdit boxes, one for selecting the data, the other for selecting the cell(s) to contain the graphics. If you’ve selected cells containing data, the Data Range box indicates this range; if you’ve selected a range of empty cells, the Location Range box indicates the selected range. These two ranges do not need to be on the same worksheet.

Click OK and the sparklines appear in the indicated position.

The weekly cycles within each month are readily apparent, but the scales aren’t quite right. The statistics begin in March 2008, about one month after starting my blog. The numbers on the first day were nowhere near where thay have been for the past few months.

The default setting is that each sparkline scales the vertical position of the data point so that its minimum and maximum fill the cell. When I selected the ‘Same for All Sparklines’ setting, the scales of all sparklines are the same, so the earlier months are pretty much flat compared to more recent months.

It’s easy to change from one sparkline style to another: just click in the range of sparklines, then click the button for the desired style.

Above are the line and column styles. A third style is Win-Loss, which plots a positive block for any positive number, a negative block for any negative number, and a blank for a zero.

The sparklines in a set are grouped together, and when you select one cell, the whole range of sparklines is outlined with a thin blue border, as shown in the sine waves below. When you apply sparkline formatting to a cell containing a sparkline, all sparklines in the group assume this formatting. You can ungroup the sparklines, and format them individually.

Although there are a limited number of sparkline chart types, within each type you have a range of formatting options. In the line charts, you can draw the line only, you can add markers, and you can format positive and negative markers differently (there is apparently no distinction between zero and positive markers). You can highlight the first and last points, and the high and low points. I added the last group to see what happened when the cells contained both sparklines and content. Apparently the sparklines decorate the back of the cell, and the text appears in front.

You can format the bar chart sparklines all the same (top left) or different positive and negative bars (top right). You can highlight the first and last bars, or the high and low bars (bottom left and center). Any text in the cell appears in front of the sparkline.

The top center group below shows how one sparkline can be ungrouped in order to format it differently. The other three have remained grouped despite not being contiguous.

Data Bars

Data Bars appeared in Excel 2007 as a way to show values visually using bars within the cells containing the values. In 2010 the capabilities of Data Bars were expanded, and the ability to make deceptive Data Bars was reduced.

Data Bars in 2007 had their base at the left edge of the cell, and they extended to the right. If a cell had a zero value, it still had a small length of bar, giving the wrong impression that the cell really did contain some value. Also, the bars started out with a good enough color at the base, but they faded to the right, so in some cases it became very difficult to judge where the bars ended.

These deficiencies were corrected in Excel 2010. Below left is the Excel 2007 representation of data bars. The first few cells contain zeros, but the cells have data bars of finite length. The bars increase in size appropriately, but they fade out from left to right, and eyes are forced to work just too hard to distinguish their ends. Finally, at the bottom the values decline asymptotically towards zero, but the bars don’t completely vanish.

In the center is the Excel 2010 version of the same data bars. Zero equals zero, and the bars have a distinct endpoint. You still can make the faded bars, but you know better. I’m not sure whether you can make nonzero bars for zero values; I don’t think so, and I hope not.

In Excel 2010 you can make your bars go right to left, as shown below right.

The unorthodox treatment of non-positive values in Excel 2007′s data bars is further illustrated below. At left, since Excel 2007 didn’t allow for negative or right-to-left data bars, the sine wave showed positive bars, even for the most negative value. The Excel 2010 data bars plot negative values in the opposite direction, optionally in a different color.

You can control the color of the axis where positive meets negative, but unfortunately you cannot change the line style. The designers have picked a dashed line, which by nature of its discontinuous dashes, draws more attention to itself than a solid line would. However, data bars have been improved so much, that a small cosmetic problem like this isn’t too important.

Backwards Compatibility

A commenter on one of the Sparklines blog posts wondered what would happen if a workbook with sparklines were opened in a previous version of Excel. I tried it and discovered:

• A workbook containing sparklines will show blank cells when opened in Excel 2007 or 2003.
• A workbook containing Excel 2010 data bars will show Excel 2007 style data bars when opened in Excel 2007, and blank cells when opened in Excel 2003.

In either case, you receive a warning about the file being created in a later version of Excel, and while your version of Excel will do its best to open the workbook, there may be some formatting which will not be faithfully displayed.

I also tried round-tripping a workbook through previous versions of Excel:

• When a workbook with 2010 data bars and sparklines is opened and saved in Excel 2003, and reopened in Excel 2010, the sparklines have vanished, and the data bars have reverted to Excel 2007 style (bars fade at the ends and all start at the minimum at the left and extend to the right, although there are no positive length zero values).
• When a workbook with 2010 data bars and sparklines is opened and saved in Excel 2007, and reopened in Excel 2010, the sparklines reappear, and the data bars retain their Excel 2010 style.

Apparently these features survive displacement by one version, but not by two.

Summary

Excel 2010 introduces a simple version of sparklines for compact visualization in the worksheet. The native Excel sparklines are not as comprehensive as existing third-party solutions can produce, but they can still be useful in many cases.

Excel 2010 also fixes conceptual problems and cosmetic issues with the data bars that were introduced in Excel 2007.

Peltier Technical Services, Inc., Copyright © 2011.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.