Charts in Dashboards: What’s coming

3737
9
05-29-2024 12:00 AM
Labels (2)
Soobin-Choi
Esri Contributor
5 9 3,737

Over the last few releases, ArcGIS Dashboards has made some changes to how data visualizations work behind-the-scenes. With the June 2024 release, that effort will be completed, meaning we will have set ArcGIS Dashboards up to bring far more chart types and features than ever before to users. 

 

As part of this effort, we will also bring in a couple of new improvements. 

Chart legend

You will have the options to place your pie chart or serial chart legend at the top, bottom, right, or left of your visualization. 

Chart actions in mobile view

You can now configure charts on a mobile dashboard to be the source of actions. Mobile dashboard viewers will be able to select one or more data points on the chart to trigger events such as filter, zoom, flash, and pan. 

Keyboard navigation 

Dashboard viewers can now navigate a chart using the Tab key on a keyboard. On a chart that is a source of actions, viewers can also use the Enter key to trigger actions. 

New chart defaults

To promote best practices, new grouped values pie chart slices will be sorted in descending order based on statistics. The slices will go from largest to smallest, in a clockwise direction. As a dashboard author, you can remove or modify this sort setting at any point in time. 

When "Parse dates" is enabled on a new serial chart, we will automatically switch the chart from a column chart to a line chart, as it is more effective for visualizing change in values over time. As a dashboard author, you can modify this chart option at any point in time. 

 

We tried to make the re-plumbing effort as smooth and non-disruptive as possible for existing dashboards. However, there are a few things to look out for, specifically, if you have serial charts that fall into either category: 

  1. Serial charts with "Periods labeling" customizations 
  2. Serial charts with more than 10,000 groups or features 

If neither sounds relevant to your dashboard, great! Keep an eye out on our What's new page around mid-June to check out all the latest enhancements and features. Otherwise, continue reading to see what will change and how your dashboard might be impacted. 

 

Serial charts with "Periods labeling" customizations 

If you customized the period labels on your serial chart using the pattern formatter, they will not be preserved with the update. Instead, they will be formatted using one of the pre-defined formats available from a drop-down menu. If you have charts with customized pattern strings, we recommend that you make note of those as they may require fine-tuning after the release.

Before June 2024 update

SoobinChoi_0-1716926645524.png

After June 2024 update 

SoobinChoi_1-1716926671902.png

 

Refer to the table below to see if your desired format will be supported. 

PeriodFormats
Year
  • 2020
  • 20
  • Feb 2020
  • Feb 20
Month
  • Feb
  • Feb 2020
  • Feb 9
  • February 2020
  • February
  • 02/09
Day
  • Feb 9
  • Feb 9, 7 AM
  • Sunday, February 9, 2020
  • Sunday, February 9, 2020 at 7 AM 
  • Feb 9, 2020
  • 2/9/2020
  • 2/9
  • 02/09
  • 9
Hour
  • 07:08
  • 07
  • 7:08 AM
  • 7 AM
Minute
  • 07:08
  • 7:08 AM
Second
  • 07:08:06
  • 7:08:06 AM

 

We have provided various date and time format options to enhance flexibility. These options offer consistency, efficiency, and awareness of local preferences when displaying temporal data on charts. Localized date and time formats will be available when dealing with temporal data from different regions and in different languages. This makes the process of creating charts smooth and more intuitive. 

Localized date and time formats

SoobinChoi_0-1716927192312.png

 

Serial charts with more than 10,000 groups or features 

The upcoming June release will bring changes that, while optimizing certain aspects, may impact performance, leading to slightly slower response times for serial charts with a high volume of data points. This adjustment is part of our ongoing efforts to enhance the overall functionality and user experience. We have tested and discovered that charts trying to process and display tens of thousands of data points will bog down the browser, if not crash it. To mitigate this issue, we have set a hard limit on serial charts to return up to 10,000 data points. This means a grouped values chart will display up to 10,000 aggregations or groups, and a feature-based chart will display up to 10,000 features. If you have charts that are, with or without your knowledge, fetching and rendering more than 10,000 data points, you will see a "Data limit exceeded" warning.  

Data limit exceeded warning

Screenshot 2024-05-28 at 1.33.46 PM.png

We recognize that the forthcoming adjustments might affect the speed and efficiency of dashboards with charts designed for large datasets. However, these changes are aimed at improving the broader system's performance and user experience, ensuring a more robust and effective platform in the long run. To ensures a smooth transition, we recommend doing the following to prepare:

  • Filter your data at the layer-level: Bring any data filter down to the layer-level, if possible, even if you have runtime filter actions configured for narrowing down data. 
  • Make sure that each chart has a purpose: Effective dashboards and visualizations communicate information at-a-glance and draw attention to what is important. If you have a dashboard that is focused on incidents over the last 30 days, verify that your charts represent the same period. If your dashboard provides a summary of requests, highlight the top 5 or 10 request types. Learn more about how to design effective dashboards. 

If you encounter any issues with your existing dashboards, you can reach out to Esri Technical Support for assistance. 

 

9 Comments
AlessioDiLorenzo
Regular Contributor

I'm struggling to understand the decision to limit serial charts to 10,000 points. In many situations, especially when representing annual data trends in epidemiology and environmental research, this limit is easily exceeded (aggregation or hard coded filters are not always possibile). Our charts remain performant even with double (or more) the data.

Even more incomprehensible is forcing the line chart for time series:
sometimes it's fine, other times it's not. For example, what happens to a stacked bar chart with dates on the x-axis? Is the line chart a modifiable default or a mandatory choice?

We were eagerly awaiting this update, but in our case it will be a downgrade and force us to abandon Operations Dashboards for other products or return to developing web applications with Python or JS, with a great loss of time. I hope I'm wrong, but I fear it will be so.

Soobin-Choi
Esri Contributor

@AlessioDiLorenzo 

I hope that I'm understanding your concern correctly and hope to provide some clarification for any confusion.

New chart limits

The 10,000 limit will apply to serial charts in 2 scenarios: 

1. You have a "Grouped values" serial chart that is trying to display more than 10,000 individual categories or grouped values. 

2. You have a "Features" serial chart, where each column or data point symbol on the chart represents one row or feature in your dataset, that is trying to visualize more than 10,000 rows or features.

In both scenarios, your chart will display up to 10,000 columns or data point symbols and display a "Data limit exceeded" warning message.

For example, in my screenshot below, I am aggregating service request entries from my layer based on a date field and have set the "Minimum period" to "Day". This means that each data point on the chart represents an aggregation of the total number of requests submitted per day

Notice in the data table preview that the underlying layer, in fact, has 64,167 rows. That's OK because the chart itself has aggregated those entries for each calendar day and the chart isn't trying to show 10,000+ unique calendar days. The chart should continue to aggregate and display all grouped values correctly and you will not see any warning messages regarding data limits. 

Screenshot 2024-06-20 at 1.59.16 PM.png

Using the same service request layer, I have configured another serial chart, setting "Minimum period" to "Minute". This means that each data point on the chart represents an aggregation of the total number of requests submitted per minute

As you can see in the screenshot below, now the chart has exceeded the 10,000 limit in an effort to aggregate entries spanning close to a year's time into individual calendar minutes. In this case, the chart will return and display up to 10,000 data points (or minutes) and display a "Data limit exceeded" warning message. 

Screenshot 2024-06-20 at 3.53.05 PM.png

The same applies for "Features" charts. Now, rather than aggregating your data, a feature-based serial chart will display one data point per entry or row in your layer. In the screenshot below, I have a serial chart where I am visualizing the trip duration for each "trip_id", which is a unique ID assigned to every recorded bike trip. The layer contains 56,107 trips, but I've set the chart to only return 10 categories using "Maximum categories". As you can see, the chart will render according to configuration. 

Screenshot 2024-06-20 at 4.26.17 PM.png

Now, say you want to show the trip duration on more than 10,000 bike trips all at once. Notice that I have raised the "Maximum categories" to 10,001 to trigger the chart limit. The chart also displays a "Data limit exceeded" warning message as a result. 

Screenshot 2024-06-20 at 4.27.51 PM.png

 

Time series defaults to line series 

Serial charts will switch from column to line series when "Parse dates" is enabled on new charts. This update should not impact any existing time series charts. Additionally, the line chart default is not mandatory and will be modifiable at any point in time. I appreciate your pointing this out, I updated the blog to address this.

LouisaM
Emerging Contributor

LouisaM_0-1719927751574.png

what is the solution to ensure that all labels appear?

 

LouisaM
Emerging Contributor

@Soobin-Choi 

what is the solution to having all labels appear?

LouisaM_0-1719927802698.png

 

Soobin-Choi
Esri Contributor

Hi @LouisaM! I would suggest trying the "Wrapped" category label placement option -

SoobinChoi_0-1719974906918.png

 

 

AlessioDiLorenzo
Regular Contributor

Hi @Soobin-Choi,

Thank you for the clarifications. Now everything is clearer and I would say that I was worried more than necessary.

KR

RebeccaLeigh
New Explorer

Hi,
I'm finding the 'Pan', 'Zoom' and 'Flash' features are not working with the new serial charts. Items on previous dashboards are fine but the new ones do not seem to function.

Soobin-Choi
Esri Contributor

Hi @RebeccaLeigh , please reach out to Esri Technical Support for assistance and investigation on this issue. 

mpboyle
Frequent Contributor

@Soobin-Choi 

After the June 2024 update, some of our existing dashboards are pretty much unusable. 

We have dashboards that display CPU and RAM performance of our servers.  We have one feature layer that stores this information and contains around 60k rows.  Inside the dashboard, we use Arcade feature set expressions to filter the data down to specific machines, resulting in about 8,500 rows per machine.  Each row represents the server performance at 5 minute intervals, giving us about 30 days worth of information. Using feature sets also allows us to set a refresh interval on the data, which we have at 1 minute.

After this update, the dashboards with serial charts take forever to draw.  Comparing the same charts using the same feature layers in our ArcGIS Enterprise deployment (11.2, which does not yet have this update) is like night and day.  The ones in Enterprise load almost right away, while it takes several minutes for the ones in AGO to draw or render the browser unresponsive.

I have tried creating hosted feature layer views to filter the rows within the parent feature layer, and while this seems to make some improvement, we lose the refresh ability by consuming a direct feature layer.

Below are example dashboards using the same AGO hosted feature layers as the source:

ArcGIS Online

ArcGIS Enterprise

Contributors