Successful businesses continually evolve their practices through identifying areas they can improve in and gain a competitive advantage. One such emerging area that businesses are now turning to is the use of location data.
So how can you use location data to better effect in your business? First, build a location data strategy and start this by documenting the challenges and any improvement targets. Then try expressing how location data can solve these. Take the case of a home delivery company. The challenge might be:
“To increase on-time deliveries, I need to understand how long it takes a van and driver to find and deliver to each address on a route.”
Expressed as a location challenge, it might look like this:
|Challenge:||Location data needed:|
|Identify the exact location of each address||Accurate geocoded addresses|
|Find most efficient route between each address||Routable road and path network|
|Find best place to park the van||On-road parking restrictions|
|Calculate the time to drop-off parcel||Distance from roadside to property entrance|
Sometimes it seems harder to relate the business activity to location. For instance, an online fashion retailer looking to grow sales might not think that location matters if it sells via the internet. However, its online customers live somewhere in the real world and their location may well influence their behaviour and spending patterns. If the online retailer has competitors with a bricks-and-mortar presence, this may reduce their customers’ spending in the area around a store. Analysing where your customers live compared to where your competitors are based may provide insights that help the online retailer to devise targeted sales campaigns.
There is a lot of location data out there. So, where do you start? Where do you source it from? How do you compare and evaluate different location data sets?
When I joined Ordnance Survey, I quickly found I had to learn a whole new language and set of concepts. On my first day, a colleague said: “just remember, it’s all about pixels and polygons”. This baffled me at the time, but it stuck in my mind. Eventually I saw the wisdom of his words. To get started, you need to know the different types of location data.
- Raster data: this is a ‘dumb’ picture made up of lots of pixels, much like a digital photo. Raster maps are great to look at and for absorbing a lot of contextual information. They’re often used as a backdrop map to overlay information on top. A classic example of a raster map is the Ordnance Survey 1:25,000 and 1:50,000 map series. Raster maps cannot be queried. But you can view a raster map on a screen without specialist software or expertise.
- Vector data: this is a series of points, lines and polygons (shapes), which collectively make up features, such as a map or a flood-risk zone. It’s possible to assign each of these features attribution, such as its function. Vector data can be analysed and queried to output answers. You will need a Geographic Information System (GIS) and a certain amount of expertise to use vector data. OS MasterMap is a good example of a vector mapping product. Another example is Listed Building data from Historic England, which uses a point to identify listed features and attributes each point with a description and a Listing Grade.
- Textual data: sometimes location data is stored as text, with an implicit location embedded in the text. Addresses or postcodes are great examples of this. Land Registry Price Paid Data typifies textual data. You do not need any specialist software or expertise to look at this data, but you may need a GIS for spatial analysis. Maptive is a great example of GIS software, they have many different features, including demographic overlay mapping.
Determine your constraints
Think about and find answers to these questions. This will help you later when you evaluate different datasets;
- What is the geographic scope of my business challenge? Your organisation could be global, national, or it could be limited to a single region in England. A lot of location data is not consistently available for all geographies. This means it is not always possible to create solutions that work consistently across geographic boundaries. Does this matter? If you have a wide geographic scope, consider if you could segment your solution into regions. If you can, this provides the opportunity to use best of breed data from each region, but this obviously adds complexity to your data management and solution build.
- What level of granularity is needed?
- Property level data is unique to the exact property you are interested in. Typically, this means the property has been individually inspected and its unique characteristics captured. Data with this type of location precision is less common and is generally more expensive, but allows for very precise decision making. So, if valuing a property for lending, you may decide you need this level of precision.
- Area level data is the same for a whole area. It involves clustering properties within an invisible boundary (e.g. Postcodes, Wards, Census Output Areas, etc.). Good examples of area-level data are socio-demographic or crime statistics published at a postcode level. Data with this type of location precision is more common and much of it is open data. But, it may be too imprecise to support decision making. For example, a postcode-level flood-model will rate the whole postcode as having the same risk, when in fact it will have varying levels of risk.
- How authoritative does it need to be? Not all location data is created equally. Its trustworthiness can vary. This is often driven by its method of collection/creation:
- Surveyed data – the ‘gold’ standard. It involves manual inspection by a trained expert. Usually highly accurate and reliable, but can be limited in scope and geographic coverage.
- Crowd sourced – this may be very accurate and up to date, but equally it can contain gaps and omissions that are difficult to predict. Timeliness of updates and quality control can be other issues.
- Modelled or derived – this is data that is created by an algorithmic model. An example might be a flood model or a model predicting the number of storeys in a building or the age of construction. This data will be predictive and not accurate all the time. But if this is accepted, it can be hugely beneficial as it may yield data that is not otherwise available.
- What expertise is in the organisation? As noted above, some location data will require specific software and skills to use. Is this already in the organisation? If not, is it prepared to invest in it? Can it be outsourced to a specialist?
- Is open data OK? There are many open datasets, particularly from UK Government, with a location element. These are fantastic resources and often contain data not found elsewhere. As with paid-for data, ensure you do proper due diligence before using open data. Look at the legal T&Cs, check the update frequency, completeness and quality.
Conduct a location data inventory
Chances are, your organisation already collects location data. However, many organisations do not take good care of their location data assets, let alone exploit them. Find and assess each of your data stores. If your location data is well managed, ask whether it can be used for other purposes. For example, a bank that lends money against a house will obtain a lot of information about the house during the application process. If this data is retained, well managed and made accessible to other departments, it could be used to issue a competitive home insurance quotation to the new owner.
Work out how to join the location data back to your addresses
Not all organisations hold their business data in geospatial formats. It is possible you will need to find another way to link your data to the location data. Addresses are commonly used to geo-reference their key business data, such as customer or asset locations.
Addresses are codified descriptions of a location. To make them usable with other location data you may need to convert them to co-ordinates. This called ‘geocoding’. The first way to do this is to use the postcode. Postcodes are defined geographic areas created to help deliver mail. They usually refer to a number of clustered addresses. They can be used to approximate the location of an address. For better precision, you can geocode the address to a specific point, such as an exact building. To do this you need to match you addresses to an address register geocoded to building level. Once you have the address co-ordinates, you can overlay this on other location data in the GIS, (such as the extent of a flood-risk zone) and extract your answers.
There are other methods of joining data together that do not involve a GIS, such as using the Unique Property Reference Number (URPN). Every address in the UK has a UPRN, and more and more datasets being published are using them, meaning datasets can be joined together in a database using simple key matches.
Source new data to evaluate
Next, search for and evaluate location datasets against your objectives and constraints. Returning to the home delivery example from above, your search criteria could look like this:
|Location data needed||Geographic Reach?||Level of Granularity?||How Authoritative?|
|Accurate geocoded addresses||UK wide||Property level essential||Authoritative – must be accurate and maintained with support|
|Routable road and path network||UK wide||Routing to the exact Property essential||Authoritative – must be accurate and maintained with support|
|On-road parking restrictions||UK wide||Area level OK||Crowd sourced, modelled or derived is OK|
|Distance from roadside to property entrance||UK wide||Area level OK||Crowd sourced, modelled or derived is OK|
Once you have a shortlist, evaluate the data by creating a prototype to test if it solves the business challenge.
With a little thought and care, creating a location data strategy can be simple and straightforward. However, there may be times when it is better to have expert help. The good news is that there is plenty out there.
Richard Crump is a managing consultant for Ordnance Survey, advising companies and government on gaining advantage from location data.