Agenda item

Thomas Rossington and Kelly Navon from Visitor Insights provided a presentation in order to explain the data to the Committee.

 

Visitor Insights had been working with the Council for the last 2-3 years to assist in analysing the footfall data through their platform.

 

The presentation provided:

 

·       Data source and how data is collected.

·       Data processing and extrapolation.

·       Terain platform.

·       Key features (Historic Trends, GDPR Compliant, Granular Outputs and Varying Scales).

·       How the Terain platform worked on location-based analytics.

·       Data pipeline refreshes.

·       Terain’s product enhancement cycles.

·       Terain’s three geofences.

·       Additional visitor insights products.

·       Customer access strategy.

 

One Member queried how data would account for visitors without a mobile phone.

 

Visitor Insights clarified that representative samples by extrapolating the data to provide a real-world context. The modelling and attribution of visits based on the devices picked up and switched were based on visitors that do not have a mobile phone or do not have their device switched on.

 

Visit attribution algorithm which would take a representative sample of the UK, ensure from data aggregations that provided enough of a cross section of the market and factor in elements such as, the population size of the area in which the device was coming from and the weight applied to that visit attribution when looking at allocating visits to a certain area. Taking into consideration the weighting according to a population size, to allocate for a certain area. There were instances where the sample size may be relatively low in population where the device usage was low. A weight was still applied to that, however, it was difficult to allocate visits from a low population size.

 

Clarification was sought around how figures may be inflated or deflated as a result of trains travelling through Grantham.

 

Visitor Insights clarified that a device or visitor would need to spend a minimum of five minutes within a geofence, to ensure they were a legitimate visitor. Any trains passing through a geofence at 50/60mph would not be counted as a visit, which would ensure data set was not corrupt. 

 

A consideration in terms of allocating locations to geofences was the velocity and time devices may ping. The system tracked pedestrians rather than vehicular data, which was calculated by velocity of the device.

 

A query was raised on what percentage of the figures provided was made up of non-device footfall.

 

Visitor Insights would look into a sample representation of the market to extrapolate the data cross. In some instances, 25% of the population count in a specific area to extrapolate against which would provide enough context for a sample size. They would not require 100% to 80% view of the entire population from a device count perspective to be able to extrapolate and get the real-world context.

 

The footfall figures being produced were not 100% certain of being accurate. It was ensured that real-world context was being used to ensure trends largely align with alternative data sets available within the market.

 

It was highlighted that a percentage of the figures of non-device footfall was unknown.

 

The data set from August-September 2023 was broken down by day and then into four sections (morning, afternoon, evening and night). It was queried why within the 2023 data, there were 17 sections (equated to 5 days) which collected 0 data.

 

The specific time period of August-September 2023, Visitor Insights saw a drop in the original data supply. The algorithm and extrapolation factors depended on original device coverage and the quality of geolocation data received from their supplier, in the timeframe from August-September 2023, there had been a drop-in service with their supplier which was rare.

 

If the August-September 2023 data was filled in with estimations, this would equate to around 20,000 visitors.

 

It was stated that data peaked in evenings when national sporting events and local events were taking place, it was queried whether this had been taken into consideration.

 

It was confirmed that a number of functions within the algorithm that take into consideration the average number for specific geofences. The system did not take into consideration specific events within geofences, however, it would flag up if an abnormal dramatic increase in the data.

 

One Member queried if a visitor sat in the same location for 3 hours, would their data be included hourly or as one visit.

 

Each classification of a smaller geofence such as, shops and pubs. As part of the algorithm, devices that were in a location for a longer period of time were factored out and equated normally. A device would be allocated as one visitor for 3 hours in the same place, however, if the period of time was longer, it may be a staff member.

 

It was questioned as to why outliers were discounted alongside the percentage of total data that anomalies and outliers represent.

 

The GPS data received for anomalies and outliers needed to be refined due to duplicate signals stacked upon each other. Duplicates were removed via necessary measures by looking at reference data. Another form of excluding anomalies and outliers would be to remove vehicular data.

 

The percentage of total data that were anomalies or outliers were unknown, due to the output seen on Terain at the end was post refinement and processing of the data.

 

It was highlighted that an individual may have more than one mobile device, it was queried as to how this would be accounted for within the data.

 

Concern was raised on an individual parking their car in a train station car park (within the geofence) and walked to the train station (which was outside the geofence), it was queried whether this person would count as a visitor to the town within the data, when in fact they would be leaving the town. 

 

The algorithm was able to identify when one individual had two mobile devices as they would be in the same ‘home-zone’ where the devices were stationary between 7pm and 7am. At this stage, one device would be deleted within the data to avoid duplicates. Within the raw mobility data, a hashed advertise ID was unique per device, but not to an individual.

 

In terms of the train station car park, a person within the geofence of a car park for 5 at least 5 minutes would count as a visitor.

 

One Member outlined an indisputable strong correlation between days where train strikes took place and the visitor figures. It was highlighted that the entrance to the train station was within the geofence, meaning the data would be impacted by people that commute to work.

 

It was confirmed there may be an impact on train station data on particular days due to train strikes, in conjunction with the use of other modes of transport being used as an alternative.

 

A query was raised on what the margin of error was on the boundary of a geofence.

 

There were several types of geofence. Cluster geofences were defined by a geofence line. For location geofences, as they were smaller, a slight buffer would be implemented by a few meters radius for a GPS signal to ping an estimated area.

 

Clarification was sought on the fact that the data was designed to reveal trends, rather than be exact figures. A wider list of assumptions was requested, it was felt that 5 minutes was too little time to assume somebody was visiting Grantham, a 20-minute timeframe would be preferred alongside moving the geofence away from the trainline.

 

Further information was requested on certain modules that the Council currently did not have access to.

 

Visitor Insights confirmed that data allowed to unveil trends and analysis. It would not be 100% accurate and was based on assumptions.

 

In terms of a delayed train, the velocity of speed seen on the GPS data would examine whether the train was passing or had stopped or delayed.

 

Visitor Insights provided a visual to the Committee on the area that had been defined since the beginning of the contract. The geofence did not cover the train station tracks, therefore, any data in terms of actual trains would not be included. The geofences could be amended at any time.

 

The resident analysis module allowed the user to create postcode grid zones up to 200 postcodes in total. Using current cluster and location geofences, data could be produced on people living in that postcode and where they travel to and from, which can be filtered for a period of time over days/nights/weekdays and weekends.

 

A query was raised on how far back data could be tracked to.

 

Historic data could be viewed from January 2018 onwards.

 

Clarification was sought around the train station and whether these were in the same geofence as Grantham Town Centre, Westgate and the Marketplace.

 

Grantham Train Station slightly overlapped the geofence for Grantham Town Centre.

 

During debate, Members commented on:

 

·       Concern was raised on the lack of information provided on days where no data had been provided.

·       That GPS signals on devices could be tracked back to home addresses.

·       Whether other competitive providers who could provide more informative data.

 

The HSHAZ Project Manager confirmed the data was commissioned in 2020 as a requirement of the Future High Street Fund reporting. The data provider went through a competitive process, however, following liaison with other users of the product, Terain seemed to be the most reliable. This contract would come to an end in March 2025, whereby alternative providers could be explored.

 

·       Whether the best indicator for footfall would be to engage with local businesses to see whether they had seen an increased turnover.

·       Concern was raised on the reliance on the data provided and it being misinterpreted. It was felt the data undermined work being undertaken for the growth of Grantham Town Centre.

·       It was suggested whether Members could go into businesses in Grantham and ask anonymously whether the road works within the Market Place had affected their business turnover.

 

The Grantham Engagement Manager continuously liaised with Grantham businesses throughout the process of the road works in the Market Place.

 

·       Whether Officers felt that the same contract should be renewed in March 2025,a new supplier be sought or whether the whole scheme be abolished.

 

The Assistant Director of Planning & Growth highlighted that procurement processes would need to be followed lawfully and via financial regulations.

 

Although the data may or may not be used in the future, the team would explore other measures in terms of the health, vitality and viability of Grantham Town Centre.