WELCOME!

Welcome to the Six Simple Steps of Data Reduction.

This course will help you to extract meaningful summaries from data. Your reports and presentations will communicate far better by following a few simple rules.

We will teach you how to find patterns and relationships in numerical data and how to reduce these to summaries that can be easily understood and communicated to others. In short – to show what story your data is telling.

The course is divided into two sections:

1  Section 1 discusses data reduction guidelines and table layout.

2 Section 2 discusses when and how to use graphs effectively.

If you have any questions about data reduction please contact info@marketingscience.info

Let’s get started!

SECTION 1 - Tips To Organise Your Data

Given how busy we all are, and how little time we have, learning to organise your data so that everyone can quickly see the key points, is immensely useful. And your colleagues will thank you!

Turning Data Into Information Into Action

We all know that marketers don’t suffer from a lack of data. The challenge is to understand and communicate what it says and means.

Data can be dangerous! Even simple data can confuse us and mislead our audience.

As an example, consider the following three statements:

• A: Of the patients taking this drug 3% had heart attacks compared to 5% amongst those not taking the drug
• B: Patients treated with drug B had 40% fewer heart attacks
• C: For every 50 people taking drug C one heart attack will be prevented

What We Know:

• Data consists of haphazard numbers.
• Information means storylines… with patterns, associations, context and meaning.

Let’s look at an example that shows market share data split into its components of penetration and frequency of purchase.

Now lets have a look at what the table would look like if it followed the principles of data reduction. What can we see from it now?

• Brand G is the market leader and has the highest penetration and average purchase frequency.
• Penetration and average purchase frequency decrease in line with market share. This is a pattern otherwise known as Double Jeopardy. This pattern was completely lost in the previous table.
• Brand C has a higher average purchase frequency and lower penetration than what we would expect. This is known as a deviation from Double Jeopardy. This brand may require further analysis.

The updated table makes it much easier to write a few brief sentences that summarise the findings in the table and even explain the important information.

Finding The Story In Data

We usually start with a huge mass of undigested data.

The first challenge is to put it into a form where we have the best chance of seeing a story (pattern). The storyline summarizes the key take-out that a reader should obtain.

The main aim of analysis is to:

• Organize and clarify data, and
• Develop a storyline that:

– Can be easily seen in the data
– Is simple so that it is memorable
– Can be verbalized so that it remains memorable

To extract the storyline we need to ask ourselves:

What is the table or graph saying?

If a table or chart doesn’t have a storyline there is no point showing it to others.

Rule: always write a storyline under your table.

What to look for?

• Look for patterns in the data
• Check for any patterns that you may expect to see
• Are there discrepancies from expected patterns?

There are six simple rules that help us do this which we will now show you.

Six Simple Rules For Effective Data Presentation

1. Round to meaningful digits
2. Order by size
3. Use averages to provide a visual focus
4. Show the main pattern in columns
5. Use table layout to guide the eye
6. Give a brief verbal summary

Rule 1: Round To Meaningful Digits

Rounding to meaningful digits simply means taking a number such as 23.43% and rounding it to 23%. We do this to make the numbers easier to read and to remember.

Meaningful digits are the digits that vary in the data we are looking at. Usually they will be the first two digits in any number. But if we have a column of numbers where all the numbers are say, one hundred thousand and something, then the meaningful will be the next two. The brain effectively filters out (but notes) all the similar ones, and then is able to compare the next two that vary.

Like this:

• 119,000
• 123,000
• 168,000

Meaningful information is not lost by rounding and this results in a much clearer table that allows you to see patterns and communicate findings.

With meaningful digits in mind, what would you round these numbers to? (no decimal points)

Here is a table showing different socio-economic groups for TV viewers who are either light, medium or heavy viewers. It is hard to see any patterns or differences in social-class between viewing segments. Are light TV viewers the same as heavy TV viewers?

Table 1: Socio-economic groups by commercial TV viewing segments

The table above needs several corrections. But let’s concentrate on the rounding first.

See below the same table with the rounded numbers.

Table 2: Socio-economic groups by commercial TV viewing segments

You can see right away by rounding to two meaningful digits, the table becomes a lot easier to read.

If we just look at the top row for the AB social class (which is the highest socio-economic group) we can see immediately that heavy viewers are the least likely to belong to this social group. Only 9% of heavy viewers belong in the AB socio-economic group.

Let’s look at the next step for data reduction using a different example.

Rule 2: Order By Size

Ordering by size is another very important step in data reduction. The table below is rounded to two meaningful digits but it is still hard to spot which country has the highest growth in GDP.

If we ordered the table in descending order from the highest GDP growth to the lowest for 2005, it becomes much easier to see a pattern.

So here we made a decision to order according to the 2005 growth rate, which was the most recent data. In general, if you have data that repeats over multiple years, sort by the most recent year – or create an overall average and sort by that.

See the below table after we have ordered the data using the 2005 growth rate.

We can see now that China has consistently achieved the highest growth in GDP for the past four years. We also see that Germany has achieved the lowest overall.

This table is a lot more meaningful and easier to read than the original one. The only thing that changed was to order it by size and move the per cent sign (%) into the column header.

It’s usually helpful to be consistent in repetitive tables that may appear in the report you are writing. For example, if you sort this table in descending order of GDP, it would make sense to keep the countries in the same order for the rest of the report for similar tables. For example you may produce GDP growth for a prior period (the 1990s), or a forecast for the next 5 years. By keeping the order of the primary table, you can quickly see if a country is changing its position.

The other thing that would help this table be even more meaningful would be to use averages as a visual focus, which is discussed next in Rule 3.

Rule 3: Use Averages To Provide A Visual Focus

The average column allows us to have a focus in the table. It gives a single point of reference for comparison, rather than comparing everything with everything. We are able to see what the average GDP was for each year as well as for each country over the past five years, and quickly observe that GDP varied only slightly by year, but much more by country. We also see that two countries, China and India, consistently grew much faster than the rest and that 2004 was the highest growth year for these countries. By using averages, the table conveys a lot more meaningful information.

Rule 4: Show The Main Pattern In Columns

Numbers are easier to read downwards in a column than across the page in a row. Remember this first ‘bad’ table shown to you earlier?

It was hard to tell what the table was trying to communicate. If you wanted to know which brand had the highest penetration (%), you had to scan the entire row and this is tricky for the eye.

But when the data was presented in columns (as well as being sorted), it was much easier for the eye to work out what the table was trying to communicate by scanning down the columns.

Something so simple like portraying the data in columns as opposed to rows can make a big difference.

Rule 5: Use Table Layout To Guide The Eye

The use of columns, spacing, font size and lines can make a table much easier to scan.

Let’s look at an example:

Below is a table showing brand performance measures for detergent. There is both observed (“O”) and predicted (“P”) data. The data has been rounded, sorted, averages are used and the main pattern in the data appears in columns. This table may be neat but it isn’t necessarily effective.

There is a lot of white space in the table – too much! There is also a lack of horizontal and vertical lines. You don’t know where to focus your eyes.

Let’s look at the same table but now with a helpful layout design. Go to the next slide to see the transformed table.

The data appears in the same order as the previous table (it is reduced and sorted) but the use of lines and spacing helps us see the information a lot easier. By reducing the spacing, the rows are easier to read (double spacing rows is never a good idea in a table). The spacing between “O” and “P” is also reduced so these numbers are easier to compare.

Column spacing should be customised to the data that is being displayed. Headings should be made to fit the data, not the other way round. It is also helpful that the average row has been separated with a horizontal line. The eye can now scan down the columns and see what the data is saying, i.e. that penetration varies widely with market share: purchase frequency is more similar, but also declines with market share.

The headers in the columns are shown in bold. The numbers in the ‘P’ columns have also been shown in bold to help differentiate from the ‘O’ columns.  Bold is useful for identifying exceptions but be mindful of using colours in a table. Many people print reports out in black and white and depending on what colour you use to highlight exceptions, they may not show up. As many as 10% of men have trouble distinguishing colour codes due to colour blindness.

Here are some tips to help you layout your table:

• Use a smaller font than main text (2 points lower tends to work best)
• Rows should be single spaced (with an extra half space every 5 rows or so)
• All columns need not be equally spaced
• Line numbers up on a decimal tab (right align and centre don’t always work)
• Selective horizontal and vertical lines (grouping things that “go together”)
• Some use of Bold can be helpful, but beware of colours (to highlight correlations or exceptions)

Now let’s look at the final rule in data reduction.

Rule 6: Give A Brief Verbal Summary

Articulating explicitly in words what the figures in a table are saying helps the reader greatly. If you look at the table below “China continues to have the highest growth in GDP” helps the reader understand right away what the table is about and also helps the reader remember the table.

The verbal summary at the bottom of the table helps us spot and remember the pattern better as well as notice the exceptions. It is important to keep the storyline brief and simple – ten words at the most.

You now know the six steps required to make data user-friendly! Just to recap, let’s look at these rules again:

1.  Round to meaningful digits
2.  Order by size
3.  Use averages to provide a visual focus
4.  Show the main patterns in columns (scanning a column is easier so have the numerical data you are analyzing in columns)
5.  Use table layout to guide the eye

• smaller font than text
• rows in single spacing
• columns spaced to fit data
• numbers aligned on a decimal tab
• selective horizontal and vertical lines
• some use of bold

6.  Give a brief verbal summary (ten words at most)

Now that you know the steps required, let’s start practicing!

Please download this Excel spreadsheet: Telephone Exercise.

The table you have downloaded was produced by researchers looking for indicators to forecast the penetration of telephones in various countries.

It’s poorly laid out.

Your task is to rearrange the table using the techniques you have been shown, so that it is clear to see the patterns in the data.

Let’s see how you can improve it!

Using the six rules you have learned about on the previous slides, make as many changes as you deem necessary to improve the table. Once you have finished, continue to the next screen.

Existing Table:

The layout makes it difficult to see the storyline or pattern contained in the table. Let’s get rid of the internal gridlines to make the numbers clearer to see. Give it a go! Scroll down to see the table without the gridlines.

Revised Table:

Now let’s start working with the numbers. Let’s round the numbers and sort the list. Try sorting it by population in descending order.

There is also a lot of white space so try reducing the rows to single spacing.

Notice that we have used different levels of rounding for the population data, because of the very large variation in the figures (from 800,000 to 3,500). We have used “drastic” rounding, to just one digit, for some of the telephone incidence figures. In this case, using a decimal place would add nothing to our understanding, but distracts from the flow of the numbers.

There is so much data that it is still hard to see any patterns. Let’s try spacing the rows out. After every third country, put in a space to help create some white-space.

Please see the below table to see the impact.

The table is starting to get much easier to read but it is still hard to see if there are any patterns. As this is a table about telephone penetration, let’s move the Telephone column to be the first one and let’s sort the Telephone column in descending order.

Try reorganizing/moving the other columns so that they appear in a logical sequence.

Don’t forget to put in the averages for the columns. After you have done it with the file that you downloaded earlier, please advance to the next page.

With the data being sorted by Telephone penetration and putting the Railway column next, we can see that a relationship is beginning to emerge between the two. The total Population column has been put last, as it seems to have no relationship at all.

The table now shows that there is a strong relationship between the penetration of telephones in a country and the per capita passenger-railway km travelled. We can see that as telephone penetration decreases, so too does the railway passenger kilometres.

We can also spot some exceptions. For example, Greece has higher telephone penetration than we would expect on the basis of railway travel alone. On the other hand, Kenya and particularly India have much lower telephone penetration than we would expect given the amount of railway travel.

Look back at the original table, and compare how much clearer the data is now. The relationship we have identified cannot be seen in the original layout.

Can you think of a verbal summary to put at the bottom of the table?

Let’s try another example! The data below is extracted from The European Marketing Pocket Book 2006. It shows selected measures of economic activity across Europe.

Download the file here: EU Pocketbook Exercise

Now rearrange and present the table using the techniques you have learned.

According to Rule 4, we should display the main pattern in columns. Given that we want to look at the economic activity in Europe, it would make more sense to put this data in the columns and the different countries in rows. So let’s turn the table sideways or “transpose” the table!

Don’t forget to apply the other techniques:

• Round to meaningful digits
• Re-order rows in descending order of population
• Group columns to absolute and per capita variables
• Add column averages
• Use layout and space appropriately

Your column headings should look something like this:

Selected Measures of Economic Activity in Europe

Selected Measures of Economic Activity in Europe

By putting the data in columns, we can easily scan the columns and compare the different countries. We can see right away that Poland is the most different. For example, Poland has the most food stores, the lowest GDP, low internet usage and has the least amount of cars per person while all the other countries have quite similar levels for the various economic indicators, once country size is taken into account. For example, there is a strong association between ad spend and population.

All of this information was lost in the original layout of the table.

Let’s test your knowledge on the data reduction techniques that you have learned. Reveal the answer for each of the statements below:

Finally, effective data presentation requires practice. It also requires deliberate effort when trying to extract the storyline. When you open a dataset for the first time, it helps to reduce the numbers to two effective digits right away. This is not only easier to work with but it will be easier to spot any emerging patterns as you begin to sort the data. Always ask yourself what the numbers are saying before you set out to create your table.

• What is the purpose of the table?
• What is the storyline?

Asking yourself these two questions will make your task much easier.

SECTION 2: Making Graphs Work

Quite often, we come across graphs in a report and are stuck scratching our head trying to work out what the graph is showing us. It can be difficult to work out what graphs say, let alone to remember it later.

In fact, many presentation graphs communicate next to nothing.

Storylines

A graph has to have a storyline to help the reader or audience work out what the graph is saying, and then to remember it.

Let’s look at an example:

Figure 1: No Storyline

So what do you think the graph above says?

Now, if we look at the same graph but add a storyline, the graph starts to take on a new meaning.

Figure 2: “One line goes UP; All the others go DOWN.”

Having this storyline pointed out and also verbalized (“one line goes up”), we can perceive it better and also remember it better. If we look back to the previous figure, we can see right away that it’s obvious “one line goes up”

Did you think something similar about the previous slide? Simple verbalised concepts like “one line goes up” are more likely to get embedded into our long-term memory.

A picture with a verbalized storyline can be worth a lot of words!

A storyline therefore needs to be:

1.    Simple so it is memorable

2.   Qualitative not quantitative

3.  Verbalized so it can be remembered
“There was a graph. It said…”

We explain the reasons for these guidelines on the following pages…

Charts V's Tables

Graphs – unlike most tables – are great for communicating shapes (eg “it’s a curve”) but are very rarely able to communicate detailed quantities or numbers (eg. % buying are 10, 70, 2, 23 etc). If this is your purpose then use a table.

Lets look at an example:

The chart above shows how unemployment is increasing each year. But this would have been much easier to work out if a storyline had been used.

The numerical details in the chart are not easy to take in nor remember because it does not give the numbers explicitly in the first place. We can’t answer simple questions like: How much did unemployment go up? Did the proportions of female unemployment stay the same? What were these proportions?

Now if we look at the same chart but one that has a storyline and numbers embedded in it, it becomes much more meaningful.

Putting the numbers explicitly on any chart makes them easier to read off again and perhaps use. However, they are still hard to manipulate. It is still difficult to calculate differences, averages etc. For example, female unemployment is lower % in ’92 than ’86, but has it increased or decreased absolutely in this time? It isn’t easy to tell. This is simply because a chart is not designed to do this. When deciding to use a chart or table, ask yourself what information do you want people to recall and remember? If it is quantitative information, consider using a table. While the chart above has become more meaningful, in essence it is nothing more than a badly laid-out table.

If we look at the pie charts below, they are barely readable. It’s hard to work out what they are even displaying. There is too much small writing to take in, let alone remember. The information is lost and nothing is communicated well.

Figure 4: Segmented channel shares in different audience sectors (example of BAD chart!)

Now, if we were to apply the data reduction techniques that you have been taught, we could present the same information in a table.

The main message is now very clear: Viewing profiles are very similar across different demographics. Children are an exception, viewing a lot more child specific channels.

Channel shares in audience segments

Tables can communicate quantitative detail much better, if the tables are well structured.

Summary:

• Charts and graphs should make known results stand out clearly:

– They are not for puzzle solving

• A graph needs to have an explicit and simple visual storyline or message if it is to communicate anything at all.

– If there isn’t a simple story, don’t use a graph!

• The visual message needs to be expressed in words.

– People can recall a verbalized concept far better than a purely visual image

Remember:

• Graphs excel for qualitative information

– Eg. To communicate a shape (“it’s a curve”)

• To help the readers easily perceive and verify the main elements of the story

– Direct the reader to the patterns

• To convey any detailed quantities, the actual numbers need to be given explicitly.

– If this needs to be done, ask yourself if a table would be more suitable.

RECAP - What Have You Learnt?

• Reducing data for further analysis and for presentation requires work and is not necessarily intuitive, but we have some sipmle guidelines to help us.
• The important thing is to find the story – and present the story, not the raw data!
• Computer graphing packages (just like statistical packages) will not do this for us. We need to make a deliberate effort and…

PRACTICE!!!!

Now, try what you have learned on some of your own data. Look at a table from a report or presentation, and see if you can improve on it, so that the main message stands out clearly.

For a print-ready list of Data Reduction rules click here.