Route increases accuracy with a new traffic intensity model

Route has updated the traffic intensity model (TIM) that is central to the calculation of audience measurement for out of home advertising.

Why is this important for advertisers?
It means that Route’s audience information is based on more precise data, making the estimations more reflective of real-world behaviour. It helps to pull apart the data and ensure that the “best” inventory receives the greatest audience. The change will directly affect the audiences for roadside ads (bus shelters / billboards / phone boxes), taxi advertising and bus advertising.

What is a traffic intensity model and what does it do?
The traffic intensity model generates the volume of travel around GB. It is a major building block in the calculation of out of home audiences.

The model comprises four key elements:

1. A map
2. Population counts
3. Travel survey (MST) data
4. A statistical gravity model

Like all good adventures, the Traffic Intensity Model begins with a map.  We use HERE maps as the common mapping geometry for both indoor and outdoor locations. Then, we feed in a variety of approved external count data to inform the volume of people using various roads, stations, shopping centres, supermarkets and airports. These are assigned to our map. Next, we pull data from our travel survey’s MST meters that tells us about traffic flows and patterns. Finally, we use a statistical gravity model to extrapolate the data across the country. It means that every road or pathway along which people can travel has a volume and average speed assigned to it.

What has changed?
There are a number of changes to the model in this data release. These include:

– Updating the geometry on which the model is based
– Refreshing the counts that serve as data inputs
– Introducing new, more granular, data to inform the speed of travel
– Increasing the resolution of the model to make it more precise at more discrete intervals

Update to geometry
The latest iteration of the model is based on HERE Q1 2017 maps. This represents a significant change for us, as we have moved to a common geometry for all environments, inside and out. This ensures greater parity in the measurement of all frames no matter their location.

Refreshing the counts
The data being fed into the model have been updated to reflect more contemporary population volumes.

Introducing new speed data
The new traffic intensity model features fresh data inputs from HERE which provide greater detail about the speed of travel for vehicles.  Previously the traffic speed data was derived entirely from the Route travel survey. It’s worth noting that pedestrian speeds are still generated from the travel survey.

Overall, the pedestrian data captured from the MST devices show a drop in the average speed at which people walk. In terms of vehicular travel, HERE speed data demonstrates faster movement on motorways and dual carriageways, but a reduction in average speed for all other road types across the network.

How have we increased the resolution of the model?
A “link” is a unique identifier for the strip between two junctions on a road or path. Some junctions might be 100m apart. In other cases, the distance may be 1km. This means that links were of varied length. Route had around 3.8 million links making up the entire road network of Great Britain.

In the new version we have split the 3.8 million links into smaller, more uniform “sub-links” of up to 20m in length. Now we have 26.5 million sub-links to play with, each of which has an average speed and volume assigned.

Fig 1: Example of a geometry before and after transitioning from links to sub-links. The crosses represent the start/end point of links.

It means that what was previously a link of say, 100m, will now be split into five sub-links of 20m each. And whereas before, there would be one average speed and one average volume assigned to the 100m length of road, there will now be five of each. This gives much more detail about what is happening and enables greater precision in representing the speed and flow of travel.

What does this mean in practice?
Think about a stretch of road. It may have a set of traffic lights at each end and a bus stop in the middle. Previously the traffic intensity model would assign a single average speed and volume to the full length of the link. Now, we are able to better reflect the behaviour of vehicles as they speed up and slow in response to the features on the road.

How does this affect audience calculations?
Route’s visibility research demonstrates that people are more likely to see out of home advertising when they move at slower speeds.

Therefore, it is of interest to brands to target ads in locations where:

1. There are lots of people travelling
2. Those people are moving slowly

The new Traffic Intensity Model better enables this. With an increase in the number of sub-links, we introduce greater variance in both volumes and speed of travel at very local levels, and at different times of the day.

Where advertising is located on sub-links with lots of people travelling slowly, the audiences will be most positively affected. Take the example below as an illustration of how this works in practice…

Fig 2. Shows a map of streets in central London. On this we have highlighted a single ‘link’ and identified the average volume of pedestrian traffic which will travel on it for a given period of time. We’ve also appended an average speed of travel to this link.

Fig 2: A central London link, with average weekly volume and speed

In the example we see 35,841 pedestrians on this road with an average speed of travel of 1.38 meters per second.

Fig. 3: Example of the same geometry after sub-links were created with volume of use and average speeds.

Figure 3 illustrates how this same road looks in the new Traffic Intensity Model. The link has been split into four sub-links. Each of these have their own volumes and associated speed of travel. Two of these sub-links are less well used than the average for the link as a whole, whereas two have higher use. If an ad is situated on a more heavily used sub-link, then there are going to be more people with a realistic opportunity to see the ad, which is likely to increase the audiences.

At the same time, there are two links (one on either end) which have a lower average speed of travel. This is likely to mean that those people who have travelled on the sub-link are going to have a higher probability of seeing any ads there, than the average for the link as a whole, because they are travelling at a slower speed. This again means that the audience is likely to go up.

Advertisers benefit from the new model by targeting advertising located in high density areas, where people travel at low speeds – or ideally, both. The model means that Route now better accounts for this in its audience measurement.

In the example given you will note that the highest volume is found in the middle of the link. Intuitively this may seem odd. It may beg the question – what happens to the people at this point? Where do these people go? This represents one of the changes which has been applied within the pedestrian model and reflects ‘residual’ audiences – those who do not flow all the way along a link, but rather wait or leave the link entirely. In this example, there is a bus stop in the middle of the link. At this point, we will have some people who leave the link and get on the bus meaning their pedestrian journey ends, equally we’ll have some others who enter at this point as they get off the bus, which is why it may be higher. Alternatively there is also an entry and exit point to the St Mary’s Hospital complex from which people may flow.

To what extent are the audience data affected?
The result is that weekly impacts for roadside frames are up by 12% from the previous quarter. Most formats benefit from this increase, with lamppost banners, six-sheets and phone-boxes most likely to gain audience.

Elsewhere we see that audiences for supermarkets and motorways service areas are up from the last release of data whereas shopping centre exteriors are down and taxis remain relatively stable. It’s worth noting that rail stations are unaffected by this change. Bus audiences also remain unaffected as the model undergoes its own development. This is due to take effect in R30 (in March 2019).

While the upgrade of the traffic intensity model represents a significant step in Route’s methodological evolution, the progress does not stop there. Future releases will feature the introduction of more granular analysis, more MST data and a brand-new digital measurement.

Leave a Reply

Your email address will not be published. Required fields are marked *