Making comparisons
The pros and cons of your list of pros and cons
- How people often compare alternatives and why it isn’t effective
- Five better ways to compare alternatives and which to use when
- Decision trees – for weighty strategic issues, not just admin procedures
- Why you shouldn’t work on the presentation as an afterthought
Sally: But I’d like the pie heated and I don’t want the ice cream on top, I want it on the side and I’d like strawberry instead of vanilla if you have it; if not, then no ice cream, just whipped cream – but only if it’s real; if it’s out of a can then nothing.
Waitress: Not even the pie?
Sally: No, just the pie, but then not heated.
Meg Ryan, When Harry Met Sally
To help managers reach decisions, they are given notes, reports and presentations that compare alternatives. The alternatives could be whether to outsource IT or keep it in-house. Or different ways to enter a new territory: do a joint venture, set up an overseas office or acquire an existing competitor in that country. The presentation might be an internal one, done by the managers’ staff. Or it might be done by consultants comparing alternatives for their client.
Regardless, they all have the same objective: to help decision-makers reach informed decisions (ignoring, of course, people’s hidden agendas). Yet – as we shall see – many fail to do this effectively. Decision-makers struggle to have informed discussions about the alternatives and often don’t even realise it.
This chapter shows how to dramatically improve decisions by bringing clarity and comparability to your analysis. Firstly, though, here’s a quick warning – this chapter doesn’t look at any of the following:
- How managers make decisions, e.g. are they risk-averse or risk-seekers?
- How managers take decisions, e.g. behind closed doors or seeking a consensus?
- How accountants quantify outcomes, e.g. should they compute an ‘NPV’ or an ‘IRR’?
Such topics are important but a host of other management books cover them. This book is about clarity, so this chapter is too. It looks at the common way of presenting comparisons in business – bullet point lists of pros and cons – then shows five better ways to do it, and which to use when.
And by the end, you will be able to make more informed decisions about how to compare alternatives so your readers can make more informed decisions.
Did you get that?
A table of ticks and crosses
Your managers have asked you to help them buy a camera. They believe there are three alternatives – a cheap compact, a medium-priced one and a pricey fancy one – and they have asked that you do a presentation comparing them.
This scenario is played out regularly at work, albeit not with cameras (at least, not in my experience). Today, though, your managers have asked about cameras, so you do some research. You analyse the pros and cons of each alternative. Is the camera easy to use? Does it give you control over how you can take a picture? Is it compatible with lenses you already own (it would be good if it is, since it saves you having to buy new lenses)? Can it take videos as well as photos? And so on.
What happens next is all too common. In your presentation, you show three slides that look at the pros and cons of each of the three cameras (Figure 4.1): slide 1 on a cheap compact, slide 2 on a medium-priced one, slide 3 on a pricey fancy one.
But because the pros and cons are listed over a series of slides, there is a physical separation between each camera’s details; direct comparisons aren’t easy. People are unable to compare alternatives unless they have a photographic memory (excuse the bad pun).
Often people do the same in written reports – the pros and cons of the cheap compact are on page 8 of the report, pages 9 and 10 are on the medium-priced one, and page 11 is on the pricey fancy one. Again, readers can’t easily make comparisons.
Instead, do a table of ticks and crosses (Figure 4.2). There are even a few double ticks and double crosses for seriously good or bad items.
There are many benefits to the table of ticks and crosses:
*:
You need to ask your questions properly to get all these benefits. If you were to ask about the cheap compact ‘Is it easy to use?’ and ‘Is it expensive?’, you would get a tick for the first question and a cross for the second, even though both are good news for potential purchasers. Readers would struggle to get a visual overview from the patterns of ticks and crosses. Instead, ensure you phrase all your questions as positives, e.g. ‘Is it cheap?’.
Before moving on from cameras, let’s look at two other layouts we could have done – a table of pros of cons (Figure 4.3), and a table of numbers that rank (Figure 4.4).
Firstly, the table of pros and cons is all on one page so is certainly better than comparing cameras over three slides. But it isn’t ideal – check out how good each camera is at enlargements. The eye has to float around the table trying to find references to ‘enlargements’. It isn’t easy. Comparable items aren’t aligned. In the table of ticks and crosses, they are.
And other problems remain too. The table of pros and cons is repetitive, offers little granularity (just a ‘pro’ or ‘con’) and has no underlying structure to impose a consistency or completeness.
Then there’s a table of numbers (Figure 4.4) that gives scores from 1 to 5 rather than double ticks to double crosses. But take care. Firstly, numbers are not as intuitive – readers have to think more about them than about ticks and crosses. Secondly, with numbers people may be tempted to add them up, and whilst this may help focus the discussion, people might incorrectly read too much into the total scores. The most appropriate camera depends on people’s purchasing priorities, not a mathematical sum of the answers to the questions. The box below explains why.
But do replace ticks and crosses with numbers if your audience want even greater granularity. The ticks and crosses effectively score items out of 5, but what if you want a score out of 100? You can’t do that with ticks and crosses, you need numbers. Which? magazine does this a lot. Which? is an association in the UK that sticks up for the consumer and its monthly magazine has articles that compare products. Some of its comparative tables give scores out of 100 and others just show ticks and crosses. But one thing Which? doesn’t do is make comparisons with a bullet point list of pros and cons.
With the cameras, we had seven questions we wanted answering, so a table worked well. The next section looks at comparing items if there are only two questions. As we shall see, we have more options open to us simply because a piece of paper can show something in 2-D but not 7-D.
Putting an overall score
In the table of ticks and crosses we could put a total score, e.g. if a double tick is +2 and a double cross –2, then a small camera gets a total score of –1, the medium-priced a score of 2 and the fancy SLR 0. However, a simple arithmetic score can be misleading because the most suitable camera could be any of the three – it depends on what buyers want. If buyers want either the cheapest or easiest to use, the cheap compact is best. (Though if you knew someone just wanted the cheapest, you wouldn’t do a comparative table, you’d just find the cheapest.) If buyers need one that does brilliant enlargements and all other criteria are irrelevant, the pricey fancy one is best. Or maybe it’s a mix of criteria – buyers want reasonable enlargements but aren’t fussed whether they are superb, but really want a video. The first criterion rules out the cheap compact, the second rules out the pricey fancy one, so the best is the medium-priced one.
So a simple arithmetic total might be misleading. And if you think it would be good to do a decision tree, we look at that in the last section.
A two-by-two grid
You are reviewing your portfolio of overseas offices. Some are in high-growth countries, some in low-growth ones. Some have a strong ability to compete, others are not so hot. Your report shows these findings as bullet points (Figure 4.5).
Even though each individual word makes sense, it’s collectively difficult to grasp either the detail or the overview. Study Norway. What is the country’s expected growth and our Norwegian operation’s ability to compete? Which other countries share Norway’s characteristics? The information is embedded deep in wordy paragraphs and is not easy to find quickly during a discussion.
If you were doing a presentation slide, the detail might again be in bullet points, albeit with abbreviated English rather than full narrative (Figure 4.6). And the same problems arise – it’s difficult to see the detail or overview.
Instead try a two-by-two grid (Figure 4.7). It shows both overview and detail more effectively because of its use of physical space. Shapes and groupings are easier to pick out – many countries are top right (and the double tick signals to readers that ‘top right’ is best), and only one is bottom left (bad news obviously, it’s a double cross). These ticks and crosses make the grid more intuitive – if they weren’t included, readers would have to think a bit harder: ‘Now … which is good – top right or bottom left?’.
And in the grid, detail is easy to see; it isn’t hidden within wordy bullet points.
Of course, if we wanted to score the growth prospects as ‘high’, ‘medium’ or ‘low’ as opposed to just ‘high’ and ‘low’, we would do a two-by-three grid. Or we could even do a three-by-three grid. You get my drift.
These grids have limitations though:
| They cope if three questions, but not more | If there was a third question, e.g. ‘political stability’, the grid can still cope if you use a simple typographical signal, e.g. politically stable countries are in darker fonts, unstable ones in lighter fonts (Figure 4.8). (Please don’t read anything into how I’ve shown the countries, it’s all just illustrative.) |
| If there’s a fourth question, don’t try yet another typographical signal, it gets confusing. Bigger fonts could signal tight regulatory environments, smaller fonts less tight, so a small light-grey country name means (wait for it, let’s get this right) politically unstable with a slack regulatory environment. You’re setting tough puzzles for readers to decode. If more than three questions, revert to a table of ticks and crosses. When we compared cameras, we asked seven questions. | |
 |
|
| They don’t cope well with ranges | Say we wish to give a view on the ‘ability to compete’ for our China operation – is it ‘good’, ‘average’ or ‘poor’? What if we have four offices in China and some are good, some are average? Our answer now stretches over two boxes of the grid and the strict divisions of the grid would struggle. This problem also arises when making predictions (‘Growth could be anywhere from “average” to “good” – that’s as accurate as we can say’). |
| See the next section for how to cope with ranges. |
There are three other points to make about these grids.
Quantify the answers? We could quantify our thinking and assign numbers to each country’s ‘ability to compete’ or ‘growth prospects’. We could then plot the results on an x–y graph and for each country there would be a dot on the graph that represents its position. However, unless you work for a highly analytical outfit, such an exercise may hinder, not help. To quantify an ‘ability to compete’ for all operations, you’d need to derive some weighted average score for each of them based on quality of management, their recent financial results, position in market, and so on. There is then a risk that managers would get hung up on the exact scores, weightings and criteria that make up each operation’s weighted average score. Management might spend too much time debating whether Canada warrants a score of 65% or 70% for ‘quality of management’, and too little time on the bigger picture.
Do a table of ticks or crosses instead? It isn’t bad (Figure 4.9), but isn’t as effective or intuitive. The two-by-two grid uses its physical space more effectively.
Don’t do the presentation as an afterthought to the report. Occasionally, the report will have the wordy bullet points we started with, but the presentation will have the two-by-two grid. This is because the report is due in first and the presentation not due for another week – and no one works on the presentation until the report’s gone in. At which time the authors realise the bullet points lack impact for a presentation, so they dream up the two-by-two grid instead. Unfortunately, it’s too late for the report to benefit from the idea.
The moral is: don’t work on the presentation as an afterthought to the report, you won’t put your best foot forward. Of course, it’s easy for me to sit here and give this advice. In reality it’s far too beguiling to work on tomorrow’s deadline than next week’s. There are ways to help this problem, but there isn’t space to cover them here.
These grids are an excellent way to make comparisons if asking two or three questions. They also help you look like a consultant. Boston Consulting Group’s famous two-by-two grid looked at each business unit’s ‘Market Growth Rate’ and ‘Market Share’ to see if it was a ‘Star’, ‘Question Mark’, ‘Cash Cow’ or ‘Dog’. McKinsey took this a stage further and produced a (wait for it) three-by-three grid (you saw it coming, didn’t you?). McKinsey categorised each business unit’s ‘Market Attractiveness’ and ‘Business Unit Strength’ not as ‘High’ and ‘Low’ but as ‘High’, ‘Medium’ and ‘Low’. And you probably guessed that last bit too.
The tables of ticks and crosses and the grids work well if the answers to questions are discrete (e.g. growth is ‘good’ or ‘average’ or ‘bad’). Next we see how to make comparisons if the answers are a range.
A cluster chart
You are looking at which types of company might buy one of your subsidiaries. You categorise groups of potential purchasers according to the following questions:
- Their strategic fit: What’s the degree of overlap with your subsidiary’s line of business?
- Their ability to pay: How easily can potential purchasers get funds to do the deal?
There are two questions, so this is similar to the two-by-two grid example – except here there are ranges, e.g. quoted competitors range from those that have a ‘good’ fit to those where the fit is ‘average’. A strictly divided grid won’t work, so instead try a cluster chart, one that has ovals or circles to represent the ranges (Figure 4.10).
It shows that unquoted competitors have a good fit with the company up for sale, and they range between those that are rolling in it and those that are as poor as church mice. Quoted competitors are all rolling in it, but some only have a moderate fit with the company for sale. And so on.
The more items to compare, the more these charts effortlessly highlight clusters, patterns, ranges, detail. The chart in Figure 4.11 looks at expected movements in regulatory controls and price changes for 13 different commodities (OK, it’s not the most realistic of examples, but it is just illustrative).
They say a picture paints a thousand words, and this cluster chart does. If we tried showing this information in words, it would be awful. Readers would struggle with detail and overview. Comparability would be almost impossible.
We could try showing the information in a table – for tea’s expected price change, the table would say ‘(10%) to 0%’. It wouldn’t be as intuitive, though, because the table would put quite a strain on short-term memory if readers wish to compare across different commodities. Because the cluster chart uses its physical space to show the information, it’s far less of a strain. The cluster chart really is ideal when answers to questions are not discrete but in ranges.
Even though we’ve seen several ways of making comparisons, we haven’t yet covered all the bases. For instance, what if we added a third question and the answers are in a range? We do this in the next section – and as it’s one of the tougher bits of the book, you’ll probably need your thinking cap on a bit more.
Are these Venn diagrams?
The oval shapes in Figures 4.10 and 4.11 may remind you of the Venn diagrams you learnt at school, ones that showed who was wearing a blazer, who was wearing a tie, and who was wearing both (Figure 4.12). However, as the table below explains, cluster charts and Venn diagrams are very different:
| Function | Venn diagram | Cluster chart |
| What it does | The Venn diagram is to categorise. | The cluster chart is to show relative position. |
| The overlap of the ovals | Some ovals must overlap. The Venn diagram exists so people can see which items are in the overlap area and which aren’t. |
It doesn’t matter if there is no overlap. |
| The size, shape and position of the ovals | So long as their content and overlaps are correctly stated, the size, shape or position of the ovals isn’t significant. | The size, shape and position of the ovals is critical. |
| The number of ovals | The Venn diagram is confusing if there are too many ovals – see below for a real-life horror. | The cluster chart works even better if many ovals – it visually summarises something that would be awful in words. |
If these differences aren’t crystal clear, you might end up doing a real turkey – you might unwittingly use oval shapes to categorise instead of show relative position, like the one in Figure 4.13 which categorises companies by activity. For instance (see bottom left of the chart), Forsyths Ltd is in Transportation and Financial Services but not in Utilities, Mining, and so on. And, in case you are wondering, not only did this really exist, but the original looked exactly like this one. It really was this bad.
It would be better as a table of ticks and crosses.
Covering the bases
Right, take a breath, here goes. Let’s revisit a previous example (shown again in Figure 4.14), the one about potential purchasers for your subsidiary, and add a third question: how keen are we to sell to these groupings of companies?
What happens next depends on how this third question affects existing groupings. If we are willing to sell to one group, less willing to another, and not willing at all to the other two, the third question doesn’t affect existing groupings at all. We had four different groupings before we added the third question, and we still have the same four different groupings after.
To show the answer to the third question, we can then simply signal it typographically. That is, we show each grouping’s name (‘quoted competitor’) in a darker or lighter font – dark means ‘OK to sell to’, light means ‘less willing to sell to’. We explain this distinction in a note to the chart, and all is fairly clear (Figure 4.15).
But what if our ‘willingness to sell to’ changes within a grouping? What if we don’t mind selling to some quoted competitors but do mind selling to others? We can’t simply signal answers to the third question typographically because there are both types of ‘willingness’ within the ‘quoted competitors’ category.
It’s probably best to revisit the groupings so as to simplify. We could create more ovals and, for instance, have two ovals for quoted competitors: one for those companies we are willing to sell to and one for those we aren’t. We could then typographically signal them with light and dark fonts, but it’s a bit useless to have an oval that says ‘quoted competitors we want to sell to’. It’s as useless as saying ‘I’ve done an analysis of people we want to promote, and there are two main groupings: people we want to promote and people we don’t’ – but then not giving names. (Also, when I tried doing it with this chart, it didn’t work because the two ovals lay on top of each other. It wasn’t visually effective.)
So instead make the analysis more granular. Give names. Look not at groupings of companies but at individual companies – look at competitor ABC, supplier DEF and client GHI, and so on. If you know each company’s ability to compete, strategic fit and your willingness to sell to them, you no longer have ranges. You can do a table of ticks and crosses, or a scatter-gram of names (Figure 4.16). This chart has only a few names on it to illustrate. Dark font means ‘OK to sell to’, light font means ‘less willing to sell to’.
This assumes we can get rid of ranges by drilling down from groups of items and looking at individual items. But what if there’s a range because we just don’t know, because we are predicting the future and all we can say is ‘between good and excellent’ (such as in the example about commodities’ expected regulatory controls and expected price changes). If that’s the case, we’re back to doing a table. Not ideal but probably as good as it gets.
To summarise, if you’ve three questions whose answers are ranges, the best way to show the information depends on why the answers are in ranges. If it’s because we’re assessing a grouping of items, then split the grouping up until you no longer have ranges for one or all of the answers – at which point you can clearly show the information. If, however, it’s a range because we just can’t quantify the future precisely, then we are more limited in what we can do – it’s pretty much a table.
The next section is the last before we summarise the chapter and it’s a bit different. It’s about decision trees. These help readers make choices by leading them through successive stages until they reach their preferred choice. We look at decision trees because, if people are making comparisons, they probably want to reach a decision. And they might think that comparing answers to questions means they are asking the right questions to reach that decision. As we will see, they are not necessarily. But they would be with a decision tree, which is why trees deserve a mention.
A decision tree
If your slides or notes have bullet points that start ‘if this, then that, if that, then the other’, a decision tree will probably be better. Bullet points such as these often arise when people explain procedures. The slide in Figure 4.17 explains how much discount to give. It depends on the value of the sale, the day of the week (there are bigger discounts on Monday), whether its cash or credit, and whether the manager has signalled a ‘Special Week’.
At first glance, the bullet points seem fine for explaining the discounts. But hidden within them are 12 different discount rates, so when you use the bullets to work out a discount, they suddenly seem less innocuous, more impenetrable. They set readers a puzzle to decode and require a fair amount of concentration. Try using the bullets to work out the discount for a £150 credit card sale on Thursday that’s also a Special Week. And when you’ve eventually worked it out, there’s that nagging feeling you might not have got it quite right. Also, there’s the kicker at the end that can be easily missed: the doubling of discounts if a Special Week. And there’s the kicker in the kicker – the doubling is subject to a maximum discount of 20%.
Figure 4.18 shows the discount rules as a decision tree. At first glance, it may look a bit more intimidating than its bullet point equivalent, but it is much easier to follow. People follow the path that is relevant to them and read only what they have to read until the tree gives the discount rate to apply. It is an easy-to-refer-to guide.
I once did a decision tree to help a client explain industry changes to its staff. It was only then that the client realised they hadn’t properly understood the changes – they had drafted bullet point slides to explain them, but the slides were wrong. And because the bullet points were a bit impenetrable, no one had noticed. Thankfully, the decision tree was so clear, it highlighted the error.
Decision trees aren’t just for procedures, though, they are for big strategic decisions too. They allow managers to spend less time struggling with confusing wordy descriptions of the issues and more time discussing them intelligently from a position of knowledge. A new CEO is wondering what the board should do with a non-core subsidiary that is a marginal player in its industry: (1) flog it, (2) hang on to it but let it carry on unchanged, or (3) give it money to make a big acquisition so it becomes market leader? The decision tree in Figure 4.19 shows the issues and arranges them in order.1 The CEO can quickly see how different assumptions lead to different outcomes, because it shows the tipping points. Also, it’s easy for the board to refer back to it during a conversation. (To simplify this example, I’ve assumed the CEO doesn’t care if there’s a big loss on disposal, it can be blamed on the previous CEO.)
To see how much the tree simplifies, here’s the tree in words (Figure 4.20 – hold on to your hats, it’s quite a ride):
It’s awful, it would send a glass eye to sleep. It’s dense, repetitive, confusing and a huge strain on short-term memory. Try referring back to it quickly during a conversation about the subsidiary.
Of course, if you can establish someone’s beliefs, you can dispense with this long list of alternatives and simply state the one outcome that best fits those beliefs. I can almost hear the boss saying ‘Stop all that intellectualising, just give me the answer – what do you recommend?’ But with issues like this, there are many decision-makers and you don’t know all their beliefs – and often there isn’t any great consensus amongst them. Some think diversification is bad, some think it good. Some think there is synergy, some don’t.
And this is the decision tree’s strength – it works for all these people. It doesn’t just give an answer, it gives the answers. Everyone can quickly establish the course of action that best fits their own assumptions.
But often, the board doesn’t even get the turgid bullet points, it gets something far less evolved – a list of issues (Figure 4.21). This is lazy thinking. It doesn’t identify the hierarchy of these issues nor relate them to possible outcomes. For instance, if the board thinks a diversified group is unacceptable and believes there aren’t synergies, it should sell the subsidiary regardless of industry attractiveness and whether companies can be bought at reasonable prices. But the list of issues makes no attempt to work this out. Also, the person doing the list hasn’t shown how the outcome sometimes depends on the trade-off between two issues – do synergies more than compensate for the unattractiveness of the industry (assuming there are synergies and assuming the industry is unattractive)?
Decision trees are a huge help for complex issues because they identify these hierarchies and relate them to outcomes. They ask questions in the right order.
And they ask the right questions. Cast your mind back to cameras – even after answering all the questions, we can’t say which camera would be best for someone because we don’t know what matters to them. If price is most important, then the cheap compact is the best, and so on. So even after comparing cameras in detail, we still can’t choose a camera.
And that’s because we didn’t ask the right questions – which means we can’t use the list of answers to work out outcomes. (I’ve seen various Internet sites try and fail to do a tree for cameras. It’s obvious they’ve failed, they ask questions like ‘Do you take sports photos?’ and have no line to follow if the answer is ‘no’.)
Decision trees do such a lot. They ask the right questions in the right order, and for each set of questions they identify the outcome. They give the answers, not an answer. They present the information clearly, concisely and in a way that can easily be referred back to and which doesn’t put a strain on short-term memory. That’s quite a few good reasons to use them when you can.
Finally, if you’ve seen When Harry Met Sally, you may remember Meg Ryan’s speech that’s reproduced at the start of this chapter. She bangs on about ‘if this, then that, if that, then the other’. It needs a decision tree. So here it is – and because Meg had such well-defined priorities, Figure 4.22 was easy to do.
There are tips on constructing trees, on checking them, on making them look neat (e.g. tricks to avoid criss-crossing lines). I drafted a 25-page chapter on the topic but don’t have room for it here, other than to give a final thought: remember your audience. Some people love decision trees, others hate them. In particular, don’t use decision trees on the young or old – research has shown they struggle with them.
Final thoughts and recap
I did think of summarising with a bullet point list of the pros and cons of bullet point lists of pros and cons. It would have had a certain irony to it – but I couldn’t. When I dummied it up, it was lopsided – no pros, all cons.
Which shows just how bad bullet points are at making comparisons. They are unnecessarily wordy and repetitive. They suffer from inconsistencies and differences in terminology that confuse. They have no underlying structure to impose a discipline on the analysis, so the analysis is often incomplete – and often neither author nor reader spot this. Also, the physical separation between the pros and cons makes it difficult to compare.
We’ve looked at several better ways to make comparisons. I don’t doubt there are other ways too, but this chapter covers enough for most situations. And it’s not just business situations either. All of us constantly compare alternative ways of doing something, be it different ways of renewing a passport or different courses we might do at college. No matter what you are comparing, the ideas in this chapter will help.
Also, to see how to make numerical comparisons (e.g. productivity or staff utilisation by department), see Chapter 3.
Recap
The three-by-two grid in Figure 4.23 shows which is the best way to make comparisons.
To help readers reach decisions rather than make comparisons, try a decision tree like the one on the left. Decision trees ask the right questions in the right order, and for each set of questions they identify the outcome. They give the answers, not just an answer.
Whichever way you choose, think about the strain on readers’ short-term memory. Think about how their eyes move around to compare items. Are you giving puzzles to decode? Ask yourself how effectively you show both detail and overview.
And at last you can now make more informed decisions about how to compare alternatives so your readers can make more informed decisions.
Applying Figure 4.23 to itself
Let’s apply the grid to itself (Figure 4.23). We want to know the best way to show different ways of making comparisons and we are going to use the grid to help us decide.
See the question at the top left of the grid: ‘How many questions are you asking?’ Well, to decide how best to make comparisons, we know there are two questions that help you decide: one of the questions is down the left, the other is along the bottom.
Now see the question along the bottom: ‘Are the answers discrete or a range?’ We can see they are discrete – in Figure 4.23, nothing sprawls over two or more adjacent parts of the grid.
So we have two questions, and discrete answers. Now look on the grid, and see the best way to show the information – and lo, according to the grid, it’s a grid. Which is what we’ve done. Which is reassuring.
1 Some of this tree is based on Michael Porter’s three tests for diversification. These are reproduced in Chapter 12 of Managing the Multibusiness Company, a collection of essays edited by Michael Goold and Kathleen Sommers Luchs. Chapter 12 is one of the most fascinating reads I have ever seen in business.