Territory generation techniques

A question we often get from our customers is whether or not we support the concept of assigning patient demand within clinician service territories and do we have any tools for building and maintaining these territories?  The answer is yes to both questions, but depending on your requirements and specific situation, there are a few ways of implementing this technique.

While we have other ways of solving the problem that territories attempts to address, we have implemented several ways of building territories with varying degrees of effectiveness.  We will describe each of these approaches as well as discuss the benefits and drawbacks in the sections below.

Postal code based generation

Postal code defined territories are where most agencies start.  While easy to describe and relate to your employees, postal code based territories are arguably one of the worst ways to divide your service area.  Postal or ZIP (an acronym for Zone Improvement Plan) codes were created to help speed the routing and delivery of packages by the United States Postal Service (USPS). Postal code areas shift (or don’t shift) over time as residential and commercial activity changes. As such, postal codes do not necessarily represent a helpful division of your workforce with respect to creating beneficial service areas.  A quick Google search on problems with ZIP/postal codes will net you plenty of information about the shortcomings, but here is one such article:


Zipcode territory

Image 1: Postal codes can sometimes create odd geographic boundaries.

In addition to the problem of postal codes not confirming to natural geographic areas, many of them are also not contiguous.  They may be represented as separated, non-adjacent areas, or can have “donut-holes” cut out of the center of them.  If you choose to build your workforce territories this way, based on your operating region, you may have to consider  these postal code oddities.

The main benefit for using postal codes is the ease of communication to your employees and general schedule maintenance.  At any point you can usually just look at the postal code of a given service appointment and know who within your staff is likely to be assigned.  Most people are comfortable “thinking” that postal codes represent a certain area. Generally speaking they do, but the actual postal code of a given location may not match a person’s mental map.  The example shown here in Image 1 is not particularly nasty, but you can see it still has problems (the blue dotted line represents the postal code boundaries).

Free form generation

While they take much longer to produce, custom formed territories are generally superior to postal codes.  Free-form territories allow you to take into account any number of factors when designing territory boundaries.  Factors such as population differences within your overall service area, geographical nuances (e.g. a bridge where traffic typically backs up) and even individual preferences can easily be accommodated.

The free-form approach produce the best territories, all things considered, with the main drawback being the time it takes to generate.  This usually leads to an unwillingness to re-evaluate the territories on a regular basis or when conditions change.  Over time this causes the established territories to more and more inefficiently manage  patient demand.  The effects can be disastrous to patient and clinician satisfaction, not to mention your operational bottom line.

The Homecare Intelligence product incorporates the ability to “draw” free-form territories, but most clients choose to use this for ad hoc adjustment after the lines have been drawn in an automated fashion.  Optionally, once the territory is drawn, we can display calculated statistics for the area (average number of patients, standard deviation by month, min, max, mean, etc.) if the historical data is available.

Algorithm based generation

Our algorithm based territory generation technique came from our clients growing tired of manually drawing out territories.  To address this, we created an algorithm for generating the territories automatically based on the average territory population.  In this way the algorithm can generate equal sized (in terms of average number of patients) and  adjacent partitions of the service area, though not equally sized in terms of square miles covered.

Patient population territory

Image 2: An example territory generation output showing equal census weighting for each area. The color codes represent different provider branches.

Working with our clients, it quickly became apparent that rather than creating larger, overlapping territories, they preferred to have smaller territories that could be assigned to staff as a primary territory, yet allowing them to be assigned as the backup or alternate choice in the adjacent territories.

From the example algorithm generated territories depicted in Image 2, you can see the more rural areas are larger than the ones downtown.  This is due to the territories being generated based on patient density, not the geographic area.

The algorithm we use is called the “shortest-split line” algorithm.  It was specifically developed as a response to gerrymandering of congressional voting districts in the US.  Voting districts are political dividing lines, and depending on how the district is drawn, they can significantly affect the outcome of an election.  The split-line method  divides a geographic region into equal (by population) parts based on the density or number of parts you wish to achieve. Here is a link to a study comparing current congressional districts with a split-line version: http://rangevoting.org/SplitLR.html.

Splitline territory with postal code overlay

Image 3: Split-line generated territories overlaid with postal codes.

A big benefit to the algorithm generated territories is that they all contain roughly the same number of patients…usually with less than 5% deviation from the mean.  Territories can be generated from actual patient address data, population data provided by the US census bureau, or a variety of other sources.

The territories can also be fed geographic “seed” areas.  For one client, we performed the following territory generations:

  1.  The population as a whole, regardless of the branch, then assigned the resulting territory to whatever branch contained the largest geographic region.  This happens a lot where two branches meet.  The upside is you can evenly divide branches, by dividing the territories evenly.  The downside is the branch boundaries will likely change slightly from where they are today.
  2. Split each branch’s historical patient population separately.  The upside is the branch boundaries don’t change.  The downside is one branch may end up servicing a larger number of patients than another.  

Hybrid approach

Freehand territory with no overlap

Image 4: An example of post algorithm-based generation free-form editing without territory overlap.

We’ve found that using an algorithm generated territory is merely a good starting point.  Even though they might be technically correct, there are still instances where they needed to be tweaked to follow a road or natural obstacle (like a river or lake).  Combining our free-form territory tool with the results from the algorithm-based generation activities to allow for adjustments has proven to net repeatable results while remaining cost effective in terms of time spent.  Image 4 is an example of the territory in “edit” mode that allows a user to change the boundaries and save them.  This particular example respects adjacent boundaries and does not allow for overlap.  

The concept of territory overlap opens up some unique opportunities, as well as additional considerations that need to be taken into account.  This will be the subject of a future follow-on post.


Technique Benefits Drawbacks
Postal code
  • Widely used and understood
  • Reasonably fast territory generation and clinician assignment
  • Does not produce even divisions based on patient demand
  • Does not necessarily conform to geographic areas
Free form
  • Excellent results
  • Takes a prohibitive amount of time to produce
  • Produces even division of demand
  • Quick territory generation
  • May not take important factors into consideration
  • Produces even division of demand
  • Allows for other factors to be managed
  • Takes a bit longer to generate territories versus a completely algorithm-based approach