Bar chart or Line? Scatter plot or box plot? These are the questions we ask ourselves when we set out to make a chart. In this article let me tell you how to pick a chart type so you can make best charts for every situation.
Why is it important to pick right chart?
Because right charts lead to right decisions. We use charts to tell stories, evaluate alternatives, understand trends or find-out if everything is normal. So, an incorrect charting choice can lead to poor judgment of the messages where as a correct chart can lead to right and faster decisions.
How to pick a chart type:
Chart making process can be divided in to 4 steps
- Find-out what you want to say?
- (Re)arrange the data
- Prepare the chart
- Format the chart
1. What is the purpose of this chart?
This is the first and most important step in chart preparation. You must ask yourself, “what is the purpose of this chart?”. Once we know the clear reason why the chart should exist, we will naturally be able to select the correct chart type for that reason.
But I realize that finding the reason itself can be a bit tedious. So I have listed down 6 common reasons that we often have to make a chart:
- to Compare
- to show the Distribution
- to explain Parts of the Whole
- to tell the Trend over time
- to findout the Deviations
- to understand the Relationship
Let us understand these reasons along with the type of charts that go well with these.
1. To Compare:
What it means? You want to compare one set of value(s) with another.
Examples:
- Performance of Product A vs. Product B in 5 regions
- Interview performance of various candidates
Charts that can be used for this reason:
- Bar Charts,
- Column Charts
- Scatter Plots
- Pie Charts
- Line Charts
- Data Tables
2. To Show the Distribution
What it means? You want to show the distribution of a set of values (to understand the outliers, normal ranges etc.)
Examples:
- Distribution of Call waiting times in a call center
- Distribution of bugs found in 10 week software testing phase
Charts that can be used to show distribution:
- Column Charts
- Scatter Plots
- Line charts
- Box Plots
3. Parts of Whole
What it means? You want to show how various parts comprise the whole
Examples:
- Individual product sales as a percentage of whole revenue
- Browser types of customers visiting our website
Charts that can be used to show Parts of Whole:
- Column Charts
- Bar Charts
- Pie Charts
- Data Table
4. Trend over time
What it means? You want to understand the trend over time of some variable(s).
Examples:
- Customer footfalls on the last 365 days
- Share price of MSFT in the last 100 trading sessions
Charts that can be used to show Trend Over Time:
- Column Charts
- Line Charts
- Data Table
5. Deviations
What it means? You want to see which values deviate from the norm.
Examples:
- Failures (or bugs) in the context of Quality Control
- Sales in Various Stores
Charts that can be used to show Deviations:
- Column Charts
- Bar Charts
- Line Charts
- Data Table
6. Relationship
What it means? You want to establish (or show) relationship between 2 (or more) variables
Examples:
- Relationship between Search Phrases and Product Purchases in your website
- Relationship between in-store sales and holidays
Charts that can be used to show Relationship:
- Scatter Plot
- Line Chart
- Data Table
How to pick a chart type when you have more than one reason for the chart?
Simple, use common sense. If I were you, I would either cut down the messages to one or make 2 charts (each conveying one message). If that is not possible, I would consider using dynamic charts or combination charts.
2. (Re)arranging the Data
Even when we know the message and corresponding chart, sometimes, our data may not support us. We then have to rearrange the data. Using excel formulas, pivot tables, tables and data cleaning tools we can easily massage the data.
Once we have the data in required format, we proceed to step 3.
3. Prepare the chart
Since you have already picked the chart type in Step 1, this is very straight forward. Most of the regular charts are available in MS Excel as default charts. You can insert them with few clicks.
But for some special chart types, you may have to prepare the chart by helper series, formatting etc.
4. Format the chart
While most formatting is done as per individual taste, there are some ground rules that apply on almost all charts. Here they are,
- No non-zero axis scale on bar charts [reasons and discussions]
- Make subtle grid-lines (or remove them) [how to remove grid lines]
- Add labels to important points [labeling techniques]
- Add descriptive, bold titles
- Position axis, scales at the right places (for eg. y-axis to the right on a large time series chart)
- Use simple, easy colors
A final word:
The ideas in this post are meant to be guide lines, not final words in the world of visualization. While these rules can help you make a good chart, a great chart take so much more. Knowledge of your data, Passion for what you do and Genuine focus on your audience’ needs can make your chart truly outstanding. All the best.
References:
Communicating Numbers – White Paper by Stephen Few [PDF]
Resources & Further Help:
- Chart chooser – Juice Analytics
- Data vis 101 – How to choose a chart – Hubspot
- Selecting right char type – KD Nuggets
- Charting Principles
- Charting Tutorials from Peltier Tech
- More articles from Stephen Few
- Charting Wisdom from Jorge Camoes
What is the process you use for Chart Selection?
I would love to know the process you use when selecting a chart type. Please share using comments.
21 Responses to “Distinct count in Excel pivot tables”
The distinct count option works well but I have found that if I have a date field and want to group by year, month, etc. that option seems to be disabled. I need to do both, distinct count and group by year/month.
Example data; sales orders with item quantities with dates.
Challenge; sum the item quantities, count the distinct orders and group by month. How do I do this?
Perhaps that's not possible due to the grouping?
@Al... When you use data model based pivots, you cannot group values manually anymore. Why not use Excel 2016's default date grouping option? In this case we have just a few dates, so Excel is not grouping them, but if you have an year's worth of data, when you make the pivot with date in the row label area, Excel automatically groups them. If you have fewer dates or want to use your own grouping, just create a table with all dates, add columns with month, week, year etc. Then connect this table (these types of tables are usually called as calendar tables) to your data on date field as a relationship. Now you can create reports by month, quarter etc easily.
Is this the only way to do it in 2013? I find it rather cumbersome to have to create another data table listing dates with the another column for MONTH() and YEAR() to be able to summarise data for senior level...
I know people find adding calendar tables cumbersome, but it is a best practice and let's you add more layers of analysis quite easily. For example, adding analysis by weekday vs. weekend or by financial quarter or YTD calculations (you would need either Power Pivot DAX or some very carefully setup pivot table value field settings)
I had absolutely no idea this was possible. Very useful, nice work!
Doesn't work for 2010 version though (or at least not my works version)
Hi ,
The post has the following in it :
These instructions work only in Excel 2016, Office 365 and Excel 2013.
when i have 2 different Pivot tables, one without the enabled “Add this data to data model” option, and the other one with it enabled.. is there anyway i can link slicers between them?
if the answer is NO,, what to do ?
Quick note, the “Add this data to data model” option is not available for the Mac version.
perhaps outside scope of this article but I have found when I attempt to create a pivot table from an external data source (connection to a sql view) the "Add this data to data model" becomes greyed out. Anybody experienced and found a solution so I can start getting distinct count in my pivot tables?
Is there a way to still add a calculated field when using distinct count?
I found I can't change the date source after tick the " add this data to the data model", can you help to adv how to change the date source in such case?
Is there a way to update the source once you have added to the data model? I receive a new spreadsheet weekly and would like to update the connection so my tables pull from the new source.
Hi Crhis, I like how you have hulk (superhero) as your avatar. Do you know that there is a superhero in Excel too? It's Power Query. You can use it to solve your problem in a simple click. Here an intro if you need some guidance.
Powerful Introduction to Power Query
A big Thank you. It worked.
Hi, have survey data that I need to analyze but the challenge is that my key fields are showing horizontally. I tried to transpose the fields using Power Query, but unfortunately the new fields are returning same values on a pivot table despite using distinct values
How I can a do a pivot table with discount conts in some columns and then generate shor report filter pages. pls it drives crazy
Hi. Why grand total pivot of distinct count is 13? shouldn't it be 67?
Great Answer! Saved me lots of time!
Thank you!!!
Worked awesome! Thanks!!
Hi Chandoo,
I am using pivot tables for distinct count and now I need to update them with new set of data. But when I update the source data, all the columns and formatting of Pivot table disappears and I need to build it from Scratch.
Is there a possibility that I can update the source data with new rows added and also retain my pivot tables?