Heat Map

RSS

Solution Elements

Rationale

Comments

Problem Summary

Heat Maps represent the relative frequency or density of entities or records as aggregate patterns of “intensity” on a map or diagram (such as a geospatial map or structural product diagram).

They help users rapidly perceive spatial patterns in record distribution, identify areas with greater or lesser density, examine degrees of variation based on spatial factors, and explore relationships between particular facets and aggregate patterns of distribution density across a given spatial area. Heat Maps enable users to understand how those density patterns and distributions change when the navigational context is updated (such as when the user selects a specific facet value or invokes a keyword search), and make actionable decisions based on aggregate patterns of relative density and other spatial considerations. Heat Maps help users to understand patterns of distribution that transcend regional boundaries. Such patterns can be difficult to discern from the analysis of individual records alone. Some typical use cases might include:

Where are the main areas of traffic congestion within the city? What are the hot spots?
What is the global pattern of activity for terrorist incidents over the last 3 months? What about the last 3 years?
Based on eye tracking data, which areas of the web page received the most attention from the user? Which received the least? etc.)

Usages

A heat map is useful when:

A user needs to understand, explore and make decisions based on aggregate patterns of density across a given spatial area, e.g. when:
- Their goals and scenarios involve understanding how certain facets or attributes relate to patterns of relative density that transcend predefined regional boundaries
- Their modes of discovery involve understanding patterns of relative frequency or density within and across a given spatial area (e.g. crime incidents, road traffic accidents, etc.) to guide further investigation or action
- These modes of discovery can be facilitated by summarizing and representing records as zones of relative intensity across a given spatial area, while enabling flexible “drill down” on specific areas irrespective of predefined boundaries. For example:
  - “Why are there so many severe accidents in this area?”
  - “What products are we selling in this “hot” sales zone, and to what type of customer?”
- Useful information scent about spatial distributions relative “intensity” can be meaningfully summarized via discernable variations in visual rendering (e.g. hue, saturation, brightness, etc.)
The records include spatial attributes that allow them to be plotted as aggregate patterns of density that transcend pre-defined regional boundaries within a two-dimensional space
The analysis of individual records in isolation would either be impractical or would not effectively communicate the aggregate values or patterns
The analysis and presentation of density constrained to predefined spatial regions would be misleading or irrelevant to the users’ questions or discovery tasks

Constraints and Challenges

Heat Maps represent aggregate patterns of distribution or density rather than individual records
Heat Maps are not an appropriate means to communicate the precise location of individual entities or records
Heat Maps do not provide a simple means of comparing predefined regions (e.g. geographic states, product sub-assemblies, etc.)
Heat Maps are best suited to holistically communicating relative variations in “intensity” along a gradient, rather than precise values

Solution Elements

Expand All

Define a function by which record density values can be mapped to “intensity” values that align with the users’ discovery scenarios, mental models, and expected level of analysis (e.g., will the user need to analyze geographic areas at the level of states, counties, cities, etc.). For example, the greater the number of traffic incidents at a given location, the darker the colour
- Ensure the mapping implies a consistent polarity (e.g. from dark red to pale yellow, etc.)
Show More...

Define a readily distinguishable and meaningful visual coding scheme -- e.g., hue, saturation, brightness, etc. – to represent the range of density values that will be represented

Whenever possible select visual codes that are contextually relevant and meaningfully associated with the context (e.g., shades of red, green, yellow for varying magnitudes of “risk” or performance relating to key metrics; or varying shades of “green” for regional performance relating to environmental metrics, etc.)

Show More...

Apply the display value (hue, saturation, brightness, etc.) corresponding to the density of the records at any given point
- Provide a readily accessible and visible legend to communicate the mapping between visual representation and meaning
Show More...

Ensure the colors are sufficiently transparent for the user to discern important map elements beneath (city names, geographical boundaries, etc)
- Consider using a slider to allow the end user to control the opacity of the heat map
"Spatial Key - Features Tour - Map" SpatialKey.com, 15 December 2010
Show More...

Apply a suitable default value for the scale of the Map so that the key areas can be discriminated effectively
- For example, show a world map view if the user needs a global view of terrorist or health incidents, a regional view if the user needs to see crime statistics in the north eastern United States, etc.
Show More...

Consider displaying further details on demand (e.g. when the user selects an arbitrary point or region on the map).
- Provide a summary of the records in that section (e.g. total number of incidents reported, type frequencies, average severity level, etc.)
- Consider allowing the user to view individual records associated with a given area
"Wildlife Tracking - Big bird sightings" myHeatmap.com, 15 December 2010
Show More...

Consider providing support for users to compare change over time, e.g.
- An animated sequence showing how the number of cases of Lyme disease increase in certain areas and decrease in others. Include VCR-style controls to start, rewind, play, stop, pause, fast forward, etc.
- A tool that allows users to choose and compare specific temporal snapshots of the data
"Computational Epidemiology Group : Syphilis - Description browse" The University of Iowa, 15 December 2010
Show More...

Consider providing support for more complex measures such as recency (relative to a specific date), or “burstiness” (i.e. the presence of spikes in the distribution), etc.
- For example, the number of cases of Lyme disease might appear to be comparable in two areas, but in one of them the cases may be much more recent. Likewise, a set of yearly sales figures may appear to be comparable in two given areas, but in one of them the pattern may be highly seasonal (represented as spikes in the distribution)
- Ensure the legend accurately reflects the revised meaning
Show More...

If the user’s goal is supported by more than one data set, allow the user to toggle between them.
- Maintain a consistent color scheme reduce the user’s cognition time when changing and evaluating data sets
- The data set selection remains clearly visible on screen and may even form all or part of the title to the map
Show More...

Consider displaying static reference points (such as geographic borders, city names, etc.) on the Map if this aids in the interpretation of the density values
- Minimize the number of reference points displayed on the map to those necessary for users to accomplish the anticipated tasks. For example, a typical geospatial map might include borders, state names, etc. but not streets, rivers, and so on unless they were essential to the anticipated usage of the application
- If the usage is potentially broad, consider including a control that allows the end user to select which types of reference point will be displayed on the map.
"Spatial Key - Features Tour - Map" SpatialKey.com, 15 December 2010
Show More...

Update the content of the map as and when the navigational context is updated (e.g. via the selection of navigational refinements or invocation of keyword search, etc.)

Show More...

Allow the user to adjust the scale of the map to focus on specific sub-areas. (Note that this action would update only the user’s current view, not the navigational context.)
- Allow the user to select arbitrary points along that scale, i.e. zooming in or out should not be constrained to incremental changes.
- Include a control for panning the map to allow users to access areas of the map that extend beyond the limits of the current view.
- Note that the value of a heat map may be lost as the user zooms in beyond a certain level of detail. One possible consequence is that the entire visible area is rendered with the same visual treatment.
Show More...

Consider allowing the user to toggle to other faceted map rendering formats when needed for more specific detail, or automatically switch to a Point Location Map when the zoom level exceeds a given threshold.
- Since heat maps show aggregate values for relative density, the ability to toggle to a Point Location Map gives the user the opportunity to explore specific records.
- Consider allowing the user to toggle to a table/list view which shows the area name in the first column and all metrics in numerical form in the subsequent columns. Each column is sort enabled for ascending and descending.
Show More...

Top

Rationale

Expand All

Heat Maps represent the relative frequency or density of records as aggregate patterns of “intensity” on a map or diagram (such as a geospatial map or structural product diagram).

They differ from Region Maps and Point Location Maps in that:

The metric they represent is density, i.e. a count of the records in the vicinity of any given point on the map
The mapping between the density and intensity values is intended to communicate a position on a gradient, e.g. differentiate between regions of low density and regions of high density. It is not meant to communicate precise locations of individual records or absolute density values

The consequence of (2) above is that heat maps are most effective when applied to data sets where the key relationship manifests itself as a spatial dependency, i.e. the value at an arbitrary point N is related to the values at points N-1 and N+1. Note that this relationship may be causal, i.e. the value at point N directly affects point N+1; or associative, i.e. the values at points N and N+1 are both affected by a common influence (possibly unknown). These relationships are precisely the sort that exploration via faceted analytic applications can uncover.

Examples of such phenomena would be sales figures across a territory; assuming that advertising and promotions take place within a given region at a given time, then the density of sales would be expected to cluster around locations where promotional activity was most effective. An alternative phenomenon might be eye tracking data from usability studies; assuming that users’ attention is drawn toward certain elements of a given web page then the fixations would be expected to cluster around those regions.

The corollary to the above is that for heat maps to be effective they need to provide context in the form of reference points, so that the density gradients can be understood in relation to other static features on the map. This in turn provides the cues for the selection of refinements to update the navigational state, and thus provides insight into the relationships underpinning the formation of density patterns and distributions.

Top

Comments (0)

Top

Related Patterns

Point Location Map

Region Map

Analytics Applications

Additional Examples

myHeatmap

Various heatmap examples at myHeatmap.com

Heat Map

Problem Summary

Usages

Constraints and Challenges

Solution Elements

Define a readily distinguishable and meaningful visual coding scheme -- e.g., hue, saturation, brightness, etc. – to represent the range of density values that will be represented

Apply the display value (hue, saturation, brightness, etc.) corresponding to the density of the records at any given point

Ensure the colors are sufficiently transparent for the user to discern important map elements beneath (city names, geographical boundaries, etc)

Apply a suitable default value for the scale of the Map so that the key areas can be discriminated effectively

Consider displaying further details on demand (e.g. when the user selects an arbitrary point or region on the map).

Consider providing support for users to compare change over time, e.g.

Consider providing support for more complex measures such as recency (relative to a specific date), or “burstiness” (i.e. the presence of spikes in the distribution), etc.

If the user’s goal is supported by more than one data set, allow the user to toggle between them.

Consider displaying static reference points (such as geographic borders, city names, etc.) on the Map if this aids in the interpretation of the density values

Update the content of the map as and when the navigational context is updated (e.g. via the selection of navigational refinements or invocation of keyword search, etc.)

Allow the user to adjust the scale of the map to focus on specific sub-areas. (Note that this action would update only the user’s current view, not the navigational context.)

Consider allowing the user to toggle to other faceted map rendering formats when needed for more specific detail, or automatically switch to a Point Location Map when the zoom level exceeds a given threshold.

Rationale

Heat Maps represent the relative frequency or density of records as aggregate patterns of “intensity” on a map or diagram (such as a geospatial map or structural product diagram).

Comments (0)

Leave a comment...

Your comment:

Related Patterns

Additional Examples

Further Reading