Excel with Excel Master

I just came across a couple of pages on hbr.org that you need to know about and forgive me if these are behind a paywall: I subscribe to the HBR so I went straight to every page.

Firstly, 10 Excel Functions Everyone Should Know by Adam Lacey and Deborah Ashby. I prefer DGET to their INDEX-MATCH function but otherwise, this is a good start to learning the top 10. Top 10 Excel Functions

Then there is a REALLY useful PDF file The Definitive 100 Most Useful Excel Tips at https://excelwithbusiness.com/blog/definitive-100-most-useful-excel-tips/

Finally and following on from the 100 must useful Excel tips, Marc Zao-Zanders gives us this, A 2×2 Matrix to Help You to Prioritise the Skills to Learn Right Now, here: https://hbr.org/2017/09/a-2×2-matrix-to-help-you-prioritize-the-skills-to-learn-right-now This 2×2 matrix puts the 100 tips on a grid/graph to illustrate four quadrants:

• Decide whether you need to learn it
• Learn it as the chance arises
• Schedule a block of time for learning it
• Learn it right away

Marc says this, “You’ll find the quickest wins in the bottom-right quadrant, which we’ve labeled “Learn it right away.” In here we have time-saving shortcuts that can be applied frequently, like Ctrl-Y (redo) and F2 (edit cell) and a nice combination formula that cleanses your spreadsheet of errors (IF(ISERROR)).” Of course, IF(ISERROR) is way out of date and should be replced by IFERROR()

Duncan Williamson

22nd December 2018

Introduction

If you have stumbled across this page hoping to find a cheap Business Class seat on Singapore Airlines (SIA), sorry! What you have arrived at is a page that explains how I found a superb result to my analysis of the last 19 years of SIA’s financial results.

The Data

As I was using Power Query to help me to analyse SIA’s financial results, I downloaded their last 19 years’ worth of Annual Reports and Accounts and from there I downloaded everything you see here:

The question is, what can I do with all of that? What I often do as part of my financial analysis is to create a Net Income model. That is, I make Net Income the dependent variable and have one or more independent variables.

The First Model

My first model comprises Y = Net Income  = All Other Variables in the above table, all six of them. I used the Data Analysis ToolPak for this and here is my output:

A Correlation Matrix of all variables and I have conditionally formatted the results to highlight the extreme values and the mid range values. For example, Revenue and Fuel Costs are very highly correlated as are Passengers and Rev Pax Km (Revenue Passenger Kilometers). Rather oddly, there is virtually no correlation between Staff Costs and the Number of Employees.

The Regression output are as follows:

Let’s note that the Adjusted R Squared value is high at 0.6755, the F statistic is significant at 7.25 but of the six independent variables only Staff Costs are significant, with a t statistic of 3.3287 and a P Value of 0.0060.

I think this is a superb result: Staff Costs being the only significant variable as its coefficient shows that a unit change to Staff Costs leads to a 2.5913 million SGD increase in Net Income.

The residuals plots show that everything is probably fine although we are only dealing with 19 data points or years of data:

Conclusion

On the one hand, this model fails because only one variable is significant and using just Staff Costs to predict Net Income is not rational. This says that we need to refine or replace the model to find something better. That being true, why do I think I found something superb?

SIA is famous for its recruitment of and investment in high quality cabin crew and other members of staff. Whilst they don’t pay the highest salaries, they do reward well overall, training is high level and frequent, they use more cabin crew per flight than most, if not all, of its competitors to maximise customer satisfaction.

Out of all of that, the data contain the relationships that I just mentioned and regression analysis has brought them out! That is superb in my opinion!

I want you to replicate my work here so, whilst there is a spreadsheet to download, it only contains the data I have extracted and used. Feel free to ask for advice and guidance here but I am not providing the full spreadsheet file.

Download the file of data from here sia_data_blog

Duncan Williamson

14th August 2018

Tuesday 25th October

This is on its way … as promised … I will upload it later today or early tomorrow.

Duncan Williamson

Someone asked this question in Quora and here is my answer which I think many of you will find useful:

If you use Excel on a Mac the chances are that you are not running Excel 2016 for the Mac and that your Mac does not have the ToolPak at all … I know, older versions have it and I know you can get alternatives!

In that case, I often demonstrate to Mac users how to create and automate the functions in the ToolPak: correlation matrix, regression analysis, moving averages, descriptive statistics … the others as well!

Descriptive statistics, for example, could be, for data in column A:

=AVERAGE(A:A)

=STDEV(A:A)

=KURT(A:A) …

=SKEW(A:A)

and so on.

Other answers have mentioned statistics software packages and that’s fine except they might not be free! Yes, if you are a student, your college or university is likely to have statistics software free for you to use.

How about R and R Studio, however? Open source, free, with massive amounts of support? Of course, it takes time to learn R but here is the code for some descriptive statistics using the psych package in R:

describe(order_sales_profit$Sales)

That’s it! This is what I get from my current data set, sales values: not exactly the same as the ToolPak but my point is, it is very easy to replicate. Look at the screenshot of the output from R.

By the way, as a novice or beginner level user of Excel, there is a lot to learn from manually automating what’s in the ToolPak. Moreover, if you take my next learning point, use this opportunity to set up templates for you to analyse your data sets: that means, you automate the ToolPak elements once and that is it!

Finally, many elements of the ToolPak return non volatile results which means that if you change your data, you have to run the ToolPak again. If you automate it yourself, the formulas you create will all be volatile: change the data, change the answers!

Duncan Williamson

I hope you can see this photo.

DW

19th August 2016

I have just completed another very successful Financial Modelling course and as you know, at the end of such courses, I come here and offer something new: a new topic, a new file or some advice. In this case, it is advice: things that you really need to think about when you create and work on any Excel file.

• Tab/Sheet Names
• Links
• Dead Cells 125,433 rows created but only 831 needed/active … files that balloon to many Mb for no real reason

Ever seen a tab name like this: FBU or OPT? I bet you have: short and sweet and probably mnemonic so easy to read and remember. How about L_P_Obasange_receivables_dont_forget_to PRINT_it_out? You think I am joking? I am serious! Just imagine you are working on your file with the large tab name and you want to link to a cell on that tab from another one: this is what will appear in your formula, by way of an example … =IFERROR(AND(A15=45,D26=”Jack”,L_P_Obasange_receivables_dont_forget_to PRINT_it_out!BA154 …

I am sure you see the point now. Keep tab names short and simple! More than that, if you do feel the need to use tabs to give instructions, colour code them to pass such messages: there are many colours to choose from so do that. Have a table of contents too. Give everyone a chance for a simple life!

Links

If you share a file with someone, make sure any links in your file are either live or delete them. If you receive a file with links that you cannot use or update, you know how frustrating it is. Think of the user before you send linked files.

Dead Cells

It is the easiest thing in the world to create a worksheet and as you work and improve what you are doing, to delete cells and ranges. We all do that. We create new ranges too, don’t we! Check your work now and again though and if these happen, take a break and check your file:

• it takes 30 seconds 45 seconds or even longer for the file to open
• what seems like a small file in terms of content and complexity has ballooned in size to 20 or 30 or more Mb
• saving the file takes an age too

If these things happen, go to a worksheet and press Ctrl+End and see where that takes you. You work only in the range A1: CD831 but Ctrl+End has taken you to CG125433 … what? How did that happen?

Even if there are not as many as an additional 1.8 million cells but just 500,000, look in those cells for formulas  that are trying to find something from somewhere that is not there … in some of these extra cells for example. Delete all of these extra cells. I did that this week: an extra 1.8 million cells in TWO separate worksheets complete with formulas. File size down from 28 Mb to 0.8 Mb, opening time just seconds, recalculation time hardy noticeable.

They were just some of things to report on from this week. Otherwise, this group of delegates really enjoyed the work and their end of course presentations were interesting and showed that significant learning had taken place!

Duncan Williamson