When you right click something in your chart you see the usual right click menu as with Excel 2007 but now you also get the floating mini-bar above it as shown below (click for larger version)

On the right hand end of this you can see the selected element (Series 3 in this case) which you can then directly format, but even more usefully, you can use the drop down here to select another element, format that, then another and so on, and the mini-bar stays on top throughout.

Now of course you can achieve the same thing by selecting elements on the Chart Tools|Layout or Chart Tools|Format ribbons, but for some items like labels and axes you want to change font properties and that means going from one Ribbon to another. If you want to rapidly change the colours or borders of various things, make the gridlines and plot area less obvious, alter the fonts of data labels or axes and other quick changes, you can do all of this without ever leaving this mini-bar area

In the screenshot below you can see I have made all sorts of changes, including making the borders of the hidden series visible, so you can see that this waterfall chart is basically a stacked bar chart with no gaps between series, and did this without going up to the Ribbon once. (I am not suggesting the formatting of the final version is better, in fact the opposite is probably true, I simply wanted to make several changes to demonstrate the new feature).

While this is not quite as instantly available to use as the tear-off colour palettes used to be in Excel 2003 and earlier, it does go a long way to replacing that functionality, and has the slight benefit of remembering the last choice made for most items so you can easily apply the same colour to axes and gridlines, for example, or the same fill colour to two related series (font options show the current value for the selected element, rather than the previous choice, and this makes good sense to me intuitively).

One minor niggle with this mini-bar is that you can’t move it, so it can obscure the things you are changing, so you should right click in the chart area near the top edge to avoid it getting it the way.

Helpfully, formats which don’t make sense get greyed out, so you can’t use fill colour for lines, which reminds you that the formatting option for “outline” applies to any element which consists only of lines, such as the axis, gridlines and so on.

Have you used this new feature already? Have you discovered other small but valuable features of Office 2010 (or earlier versions)? What are your favourite time-saving tips?

(Aside: the same applies to a series of posts about 6 related features of some software or comparing 5 alternative products which would have made much more sense written as a single cohesive article, but failed to tick the box for 10 blog posts per month. You get what you measure, or WYMIWYG)

But today, I though this was important enough to just say – have you seen this post on the Excel Team blog about improvements to functions in Excel 2010? Now that one’s a week or so old now, but today there was an even more important post with a link to a draft whitepaper with more information.

I have linked to the post , not the pdf file directly as it is only a draft, and hopefully the post will get an edit or at least a comment when a newer or final version is available. This document is for those who like to understand the details, and for any sceptics who might say “well, they said it was accurate last time, how do we know it’s any better now?” – it does sound a bit like washing powder ads who always tell us that this time round it will get things even whiter and brighter and cleaner than ever (just like they said before).

There are a couple of typos in the draft (the floor.precise function for example has an obvious chunk of copy and paste from the ceiling.precise function for example), and there are some things not made very clear (for example in most cases it only describes the new behaviour, not the old for comparison or for explanation of the difference and why the new way is more accurate).

It lists the MOD function in the section on functions whose accuracy has been improved. Did you know that MOD gave inaccurate results on older versions of Excel? I didn’t. In fact, it doesn’t – it gives a completely accurate result or fails with a #NUM error if the divisor goes into the number more than 2^27 times as described here. I think this is distinct from being inaccurate in the way the statistical functions have often been criticised.

Similarly the RAND function is listed but its problems not described in any detail – I can only assume that the function is flawed in that it does not give a perfectly even distribution of results and is therefore “weighted” to some extent. Since I only ever use it to produce dummy data for examples used in my Excel training courses, it does not really matter to me if it not truly statistically pure, but I am sure to others it is vitally important what algorithm is used to generate the results (it’s now the Mersenne Twister, for those who care, but this fact is not from the whitepaper, it’s in a comment to the original post made by Jessica Liu).

Anyway, the bugs and inaccuracies that are discussed in the whitepaper are all now fixed (but it does not say this in the whitepaper, merely leaves it implied), and already works in the Technical Preview. Some of the other changes came too late for TP but should be in the public Beta when that gets released.

I expect the changes to naming conventions will also help people who use the statistical functions a lot (I’m not one of those) or have to make sense of others’ work. The convention of .precise added to a function name seems to mean “according to a precise definition” rather than “inherently more accurate”. I would have though .strict might be less ambiguous and similar to the use of the term in other fields (eg web design using XHTML versus XHTML strict).

Has anybody had real issues with these inaccuracies in the past? (and had you even noticed?)

Are you using other software tools to avoid the problem in Excel, and will these changes allow or encourage you to switch back?

Along the way I’ve been finding out loads of cool things about major new features such as sparklines and slicers (more on that in a future post, as promised), and the ability to customise the Ribbon easily without writing code. There are also lots of tiny changes as well, which are easy to miss and may get drowned out in the sea of other news about the next version, so I thought I would mention a couple of them here – the status bar summaries and filters in Tables.

Subtotals ignore previous subtotals – so does the status bar in 2010

As you might already know, if you use a SUBTOTAL function to calculate a cell, and then use another SUBTOTAL to calculate cells including the first one, the original subtotal is excluded from the second one, which makes sense to avoid double counting / summing.

If cells A1,2,3 contain 1,2, and 3, and A4 is =SUBTOTAL(9,A1:A3) then A4 will have the sum: 6 (‘9’ is the sum sub-feature of the SUBTOTAL function). If cells A5,6,7 contain 5,6, and 7, and A8 contains =SUBTOTAL(9,A1:A7) then this will have the sum of A1,2,3, 5,6,7 and skip the value in A4 as it realises this would give a wrong result. This works even if you use different functions for the totals, for example you might use counts (function 2) for most subtotals, then use an average (1) and a sum (9) right at the bottom or top of the column.

Now, you can select the same 7 cells and take a look at the status bar – this should show you the sum, count, average or whatever functions you have chosen. In 2003 and earlier you can select one at a time, in 2007 and later you can choose to see multiple summary results at the same time – in all cases, right click where it has a summary at the moment (eg “Sum: 123”) and choose the functions you want to see.

If you have 2007 or earlier, the status bar shows the ‘wrong’ result of 30 – the sum of all 7 cells including the subtotal. While this is a correct sum, it is unhelpful, particularly if you select a large number of cells or a whole column, and may not realise there are one or more subtotals in there. Selecting cells A1:A8 is even worse, you now have two subtotals included for a whopping error of 54.

In Excel 2010 this behaviour has changed and these functions in the status bar ignore SUBTOTAL formulas in cells just like the function itself does, so the count shows 6 and the total 24. You can see the difference in the status bar of the two screenshots below, for Excel 2007 (left) and Excel 2010 (right) (click for larger images). The other features like Average and so on do the same, but I could not show too many at once on the same screenshot.

Now, on the one hand this is a great idea because in general I think this is the result most people would actually want and even expect. But on the other hand, there are thousands of existing users who already know and understand how this works, and might find the change undesirable, if for no reason other than the fact that it is not obvious. It does save time because you don’t have to select all those cells separately using CTRL-clicks or a GoTo > Special (if the numbers are really numbers and the only formulae are the subtotals).

Maybe a better way to offer this functionality would be to add “Subtotal” to the list of status bar features you can display, with the expectation set that this does the same as the function of the same name (unlike count, which is more like COUNTA; you have to use “Numerical Count” to get the same as COUNT. Don’t get me started on that one!).

Tables in Excel 2010 show filters when the headings row goes off the screen

If you have used Tables in 2007 at all, you may have been pleased at how the column labels A, B, C… get replaced with the headings from your table if you scroll down so the headings are no longer visible while you still have a cell in the table selected. Frustratingly though, the autofilters which are very helpfully added to Tables automatically are not available. you have to scroll all the way to the top to put a filter on, then scroll to take a look at your data and perhaps the totals, then up, down, until your scrolling finger aches.

Well, ache no more – in Excel 2010 the filters are also shown in the column headings so you can filter wherever you are in your table, as shown in the screenshot below. A very simple, but very useful additional feature this one, I am sure you will agree.

In the context of these two features, I should also point out for anyone not already aware, the Total row in a Table uses the Subtotal function rather than a sum, count, average etc, so this means it will show the total for the visible rows if you filter it (which makes sense), and will get ignored in the status bar in Excel 2010 but not previous versions. In the example above I have selected the whole of column C (sales) to include the total row, but the status bar shows the more useful result of 71,261, rather than double this as would be seen in Excel 2007.

Anyone else found any little nuggets that have changed? What are you favourite improvements to Excel 2010 (or the other Office apps, come to that)? Anything missing that you wish had made the cut this time round?

As more information emerges, I’ll write in more detail about some of the new features. Today let’s have a look at probably the most eagerly awaited extension to Excel’s data visualisation capabilities – sparklines.

Sparklines in Excel 2010

Sparklines are very dense microcharts used to display simple information, usually showing historical values to give context to the current data. The term was coined by information visualisation guru Edward Tufte and discussed in a whole chapter in his book Beautiful Evidence, which describes sparklines as “intense, simple, word-sized graphics”.

Examples might include past share values, commodity prices, exchange rates, or internal business key performance indicators. Lines may have key points highlighted (high value, low value, last value), show a trendline or normal band, but otherwise will be deliberately uncluttered to aid easy interpretation.

The example to the right shows the previous 12 months sales and target, with the highest and lowest sales figures highlighted. Some people might prefer to show the figures to the right of the sparkline as they relates to the final data point and therefore the right hand end of the plotted line, but this is a matter of personal preference. Although this example is show here in quite a large screenshot, the trend is very clear at much smaller sizes too.

While lines are by far the most common choice for sparklines, especially to show changes over time, other formats may be found – columns to show breakdown of a total by category for comparison, for example. Another popular use is to track success and failure, such as wins and losses for a sports team, a technique described well by this article on sparklines at Bissantz, the creators of SparkMaker, an add-in for Excel.

Now Excel 2010 does sparklines straight out of the box, no add-ins required, which is a testament to how popular these kind of dense data visualisations have become, especially in the BI and business dashboard community. Of course, you could already do this with add-ins or just a lot of patience to customise normal charts to suit the restricted size of a typical sparkline. The way these are shown in the videos is that they are much easier to apply to a range of cells rather than having to build then copy and paste for multiple rows, adjusting data ranges and so on. So you build one in a cell, choosing the data range that it relates to, and then you can use the usual “fill down” technique by dragging a cell handle to create more sparklines for subsequent rows of similar data. Various formatting tools can then be used to quickly change colours, show or hide markers and highlight highs or lows, first and last point and so on. Whether these sparklines can be made to be dynamic to accommodate new columns or rows of data (using dynamic named ranges for the source, for example) is not clear at this stage, but hopefully this is covered. Currently the three formats which are built-in are lines, columns and win/loss blocks. Thankfully no-one tried to build a spark-pie-chart!

Sparklines are good news, if used well

These could be a really powerful addition to many pivot tables or other reports. Let’s just hope their convenience is matched by sensible controls so that first-time users do not end up with pretty but meaningless charts. The screenshot of sparklines shown on the Microsoft PressPass site is a fairly poor example – the lines and bars show the same data (acceptable from the point of view of a demo shot showing both formats are supported), and they actually chart all the visible data. Part of the point of a sparkline is to show additional data in a very dense way – for example you might display the current value of some measure, and a line to show the past twelve months so you can see what has happened to get to that position. Showing monthly figures as well defeats the point of the density possible here, and the principle of being able to visualise the information available rather than wading through figures to try and interpret the patterns.

Related news – data bars are being improved

I haven’t seen many details, but the data bars and some other conditional formats introduced in Excel 2007 are being improved in this release. A key feature being mentioned is the possibility to show negative values with data bars. For those of you unfamiliar with the strange implementation in 2007, the lowest value was displayed as the shortest bar, highest as longest. So if you had a mix of positive and negative numbers, a zero value would appear as a bar of an intermediate length, and there was no really useful meaning for the lengths of the bars. For a set of all negatives, the highest value (nearest to zero) had the longest bar, and the lengths do not have a proper relationship to one another, as seen below.

In this example I have highlighted some figures – two figures for the variation between actual sales and target in monetary terms (2,280 and –2,280) and two for the variation in percentage terms (9.0% and –9.0%). Notice that the lengths of the bars for these pairs of figures have absolutely no relationship to one another. A shortfall of sales by 2,280 should surely be shown using the same length of bar as an overachievement by the same amount? As for April – sales target was met exactly, this is a completely neutral achievement but the bar uses a substantial amount of “data ink”. This amount also varies depending on the magnitude of the highest and lowest values in the series!

Hopefully the new implementation for negatives in data bars in Excel 2010 will give us a zero-length bar for zero, and bars extending left for negative and right for positive (optionally with colour coding to highlight negatives even further). Additionally, cleaning up the bars to lose the graduated fade-out which makes the ends unclear, and allowing the shortest bar to have zero length rather than a small minimum size would be better (especially so that a zero value is encoded as a zero-bar).

Next time – data slicers

In my next post I’ll take a look at the new data slicers feature to try and work out what it does and if it really introduces new capability or is just a new way of presenting some existing tools like data filters.