Configuring Charts
Last updated
Last updated
This guide will show you how to create charts, which make it easier for you to understand your performance data and quickly diagnose faults.
Different chart types may display some different features owing to the way they graphically represent data, but many chart configuration features work the same across all chart types. The following instructions apply across chart types, using a line chart as example. Charts with other or differing feature configurations are discussed at the end.
Navigate to Connected Devices > Data Views > Analytics from the side navigation bar.
You'll see the chart builder on the left and any pre-saved charts (called presets) on the right.
Note: the chart builder defaults to showing all standard and advanced chart features. You can Hide advanced options at the bottom of the chart builder for simpler chart creation. This section explains how to use the standard features. See to understand how to use advanced features.
Select the Metric you wish to visualize from the dropdown.
You’ll see all the metrics for all the tests that are assigned to test schedules on your account.
With some metrics, you can further refine the metric parameters and/or add another metric below the initial metric selector (viewable with advanced options).
In the Chart dropdown, select the type of chart you wish to create.
The following steps use a line chart to explain each feature.
Select a Date range: choose one of the pre-defined options or customize the date range.
The predefined date ranges include:
Last 2 days
Last 7 days
Last 30 days
Last 6 months
Select the relative date range (Last n days) to show the last few days, weeks, months, or even years. You can save it as a preset (see to find out how), and it will always show the most recent data for the set time period. Presets work the same way for the predefined date ranges listed above.
Specify a Fixed date range to look at results during a specific window of time, not necessarily starting from today. You can narrow down the range even further to specific times. Saving as a preset means you’ll save that specific window; the preset won’t update to today’s date.
Select your aggregation filter, located next to the Date range filter.
This is used to work out how much data is aggregated into each point on your chart (time is on the x-axis).
For example, an “Hourly” aggregation over the “Last 2 days” would display 48 points on your chart’s line, an average speed for each hour in your time range, whereas a “Daily” aggregation over the same period would display two points, one for the average of each day.
Note: While most aggregations (“Hourly”, “Daily”, “Weekly”) will display time in order on the x-axis, others such as “Hour of Day” will show you the hour (e.g., 4 pm) on the x-axis and take all results between 4-5 pm in the date range specified. This allows you to analyze occurrences such as “the average hour” or “the average day”.
Click Apply changes to load your chart on the right.
In most charts, graph inspection will show you all the raw data points forming an aggregate point, both as a scatter plot and as a table. The table includes a shortcut to view the Device Agent information by clicking the Unit ID.
To start graph inspection:
Click any point on your graph.
This reveals a scatter plot and a table.
Use the ‘Show / hide columns’ option to add or remove fields.
To see a table of all the points on your graph, select the Data Tables tab. Here, you can also view aggregate functions outside of what you're plotting on the chart, along with their average.
Under the chart title, select the Data Tables tab.
Click the Show / Hide Columns button to add or remove statistical functions on the same aggregate point.
When viewing a chart, you may want to export the data it’s showing you to put it into other systems or share it with others. You can export data as a CSV file to do just that.
Note: Almost every chart type can be exported as a CSV. The exceptions are box plots and bar charts; for these, you can export a CDF plot or summary table instead, as they contain the same information as box plots and bar charts.
Below the chart and legend, you’ll see a series of action options and icons. To export the data as a CSV file, click Download CSV.
The CSV file will export every statistically significant function available for every aggregate point on your chart, similar to viewing the data table with most options checked.
To export the chart as an image, click Download image. Note that Summary Tables cannot be exerted as images.
If not already done, click Show advanced options at the bottom of the chart builder to reveal more chart features that can help you better visualize, and make sense of, your data.
Under the Date range feature, select any Filters to restrict the data shown on your chart.
For example, you might restrict your chart to only show three packages: BT 160, BT 300, and BT 500.
Use the Include/Exclude radio buttons to add or exclude as many items as you wish from the dropdown list. If there are tests/agents which have no value for your selected filter, they won’t ever be shown if you’re using an inclusion filter, but they cannot be excluded with an exclusion filter.
Depending on the filter chosen, such as “Technology”, the field may provide auto-completion of values once you start typing, in which case it will not allow invalid values to be used. Other filters, such as “Target”, allows you to enter free text. Just hit Enter after each entry. All entries must match exactly, and the filter is case sensitive.
There are also a variety of range filters. Range filters typically accept two inputs, usually an upper and lower bound. Some examples are the metric value filters and agent count and sample count filters.
Note: Often, entering one value is enough. For example, if you enter 20 in the “range min” (left-hand) box, the chart filters down to results that are greater than, or equal, to 20. All values specified are included in ranges.
To view the prefilters applied to your chart, click the Show prefilters button under the Filter field, and hover over the “info” icon.
A pop-up appears listing the prefilters applied to your chart.
To remove the prefilters, uncheck the Enable prefiltering box.
Optionally, Split your data into groups. This means that data will be grouped into different chart features (such as lines, bars or scatter points) on your chart so you can compare them.
For example, if you split by “Package/Product/Tier”, you will see one line for each of the three BT products you filtered by, above.
Confidence intervals are vital for drawing statistically valid conclusions and statements from data when sampling.
All our charts, when showing averages, only show you the average of the sample of your customers (the population) tested. So, whilst we can accurately tell you the average of the sample, what we cannot be sure about is the average of the population. This is what confidence intervals are for. They provide a range, and a confidence to that range, that if you had an average of the entire population of consumers/agents, the data would sit in that range. On line charts, confidence level ranges are shown as a colored area. Bar chart confidence levels display as error bars.
To add confidence ranges to your charts:
Check the option below the chart type that says Show confidence intervals.
Click the dropdown to select the confidence level you want displayed: 90%, 95%, or 99%.
Note: The larger the sample (e.g., 1000 agents versus 100 agents), the narrower the confidence range will be. Also, the range will be wider the higher the confidence level you request (e.g., 99%).
It can be useful on charts to plot other statistics than just the mean average.
Above the Date range filter, select the Value to plot dropdown.
Select the statistic you wish to chart against.
For example, selecting “Median” instead of “Mean” results in the chart below, all other things being equal.
By default, our charts will always aggregate over all the test results but often it’s a requirement to aggregate by agents first.
When you aggregate by test, you see an overall average of all the test results in each data point (where a data point in a chart aggregated hourly would mean each hour) but if one agent reports more often than others, this could lead results to be skewed. In order to get around this, you can average all the results by one agent in each data point first, then take the average of those agents.
Check the option below the chart type that says Aggregate by agent.
Click Apply changes. The resulting chart appears below, all other things being equal.
You can enable a feature called normalization when using time series line charts. This is useful when plotting results by hour, but where the test schedule does not run test results every hour across every agent. Without normalization, you could see a saw-toothing effect (as in the image below), with some hours having measurement agents with much better results than other hours.
Normalization forces the chart to look at the underlying test schedule for the metric in question, and to normalize results across the blocks of hours that the tests should run over. This results in an even number of measurement agents and results being represented for all hours of the day, thus removing the spikes in the charts and making it easier to see overall trends.
What actually happens behind the scenes for normalization is, after all filters have been applied, the chart will look at the test schedule of each individual unit. If a unit runs a download test once every 6 hours between 6 am and 6 pm for example, and runs two tests at 8.37 am and 2.37 pm, then it would duplicate the 8.37 am result into every hourly period between 6 am and 11 am, and duplicate the 2.37 pm result into every hourly period between 12 pm and 6 pm.
You can enable or disable normalization in the chart builder under the Date range filter.
Notes:
Normalization can sometimes cause confusing results; for example, you may filter down to a single unit which tests once every six hours, but the graph shows a point every hour.
Normalization is only available on hourly, hour of day, two hourly, four hourly and six hourly aggregations.
Bar charts provide a quick way to compare the averages, or other values, of different data series. The y-axis shows the value being plotted, and the x-axis displays the data split that each bar represents.
With all other features configured as in the line chart example, the bar chart will display as below.
Scatter plots allow you to explore your data in more depth than aggregate data.
For scatter plots, an additional option appears to show a “regression line". This means the chart will generate the scatter plot, then if it notices a trend, it will draw a faint straight grey line to show you that trend (whether it be up, down, or straight).
Splitting is still available on scatter plots and will result in color coding your points according to the data series they belong to.
CDF (Cumulative Distributed Function) plots show the cumulative distribution of measurement results. The shape of this curve allows you to identify patterns in the data (e.g., step changes indicate a clustering of results around certain values) and also to immediately recognize whether performance is normally distributed among users or not.
For CDF plots, an additional option appears to draw a “tail distribution”. A tail distribution inverts the CDF plot and shows you the percentage of users that received a certain value or higher, instead of the percentage that received the x-axis value or lower.
Box plots are useful for seeing the range of results, across the median, upper and lower quartiles, and the minimum and maximum of each data series.
In addition to plotting your data as a visual chart, you can also create a summary table. Tables allow you to include additional aggregate functions. You can change the values of the columns inside a summary table (e.g. Mean to Median, or add more statistics (e.g., Mean and Median).
Add as many values to plot as you like by adding them to the Table columns selector and they will be shown as columns.
Multiple data series lets you generate multiple lines (or statistics) on the same chart.
Build your first data series using the chart builder, ensuring advanced options are on.
Under the Split function, click Add another series (adds a new blank chart builder) or clone from current series (adds a new chart builder populated with the current selections).
Note: If you want to use the same splits and filters as before but with additional metrics with the same units (e.g. both measured in ms), you can use the shortcut Add another metric button from within the same series instead.
View the new chart builder display that adds a column on the left indicating which chart builder (and hence which series) you are in.
Each number represents a separate data series.
Add more data series by clicking the “+” button under the numbered series.
Remove data series by clicking Remove this series at the bottom of the relevant chart builder.
Build your second data series (in our example, we’ll choose “Upload speed”).
Note: If you’re exporting larger amounts of data, it will be faster for you to use Connected Devices > Data Export. See .
Note: you can find a complete list of available filter types, such as those listed under Package/Product/Tier if you export the CSV from Connected Devices > Management > Management Suite > Agents. Click Download export, then select Agent metadata. You will receive a link to a downloadable file by email. See for more information.
In a standard chart, you can only choose one value at a time. To find out how to plot multiple statistics on the same chart, such as both the 20th and 80th percentiles, see .
For best results, enable Show Advanced Options to choose how you wish to split your data (see ). Additionally, under Value to plot, select an option in the Sort By dropdown and modify the order of the bars on the x-axis, either by the data series name or any statistical aggregation.
As with , you are able to sort your data by name or statistical function.
