Using APIs and spatial analytics to optimise and plan your "new normal" social life
Back in 2019, I built an app that quickly gave me and my friends a list of the ‘best‘ places to meet up. This factored in the type of place, each person’s travel time, reviews, price etc. Whilst at the time I used it as ‘a bit of fun’ and a useful foray deeper into using APIs in spatial analytics, I’m hoping it might be more useful than ever as we enter a ‘new normal’, as a means to minimize travel for everyone and find new and less crowded places to meet.
This was written alongside my development in Alteryx as a guide to how I built the app, which was in a pre-lockdown world… So, flash back to 2019 and a nicer time, when we could go out and feel safe going to meet friends.
(Although re-reading some of this back, it seems I was still just as negative back then. You really don’t appreciate what you’ve got until it’s gone…)
November 2019
I was annoyed; I’d already been waiting on my own in a run-down London pub for 15 minutes when I received a text saying, "sorry mate I’m just leaving… but get me a drink, I’ll only be 1 min". I was still annoyed, but no longer because he was late; nor that the bar was terrible; not even that I had to buy him a drink. I was annoyed because the place we’d chosen to meet clearly happened to be right by his house, despite being miles away from mine. This probably makes you think I’m an awful person to go for a drink with. And you’d be absolutely right. Anyway, here’s how I think this kind of problem could be solved in future….
Like any rational analyst, I left the dingy bar after finishing my pint with my (equally dingy) mate and set about thinking how this problem could be solved. Before tackling the challenge at hand, I wanted to understand what the specific problem was, and the possible constraints defining it. I summarized them as being:
Problem: two (or more) people need to decide where they want to meet
Constraints:
- It must be possible for both to reach this place within each person’s defined travel time (clearly my mate’s max travel was 1 minute, I should take a hint)
- Each person may choose to drive, walk, cycle, or take public transport to the location
- The place should be a balance between even travel times and quality of venue
- The place type can be pre-defined (café, bar, restaurant etc.)
- You should be able to see the place and reviews before you go
This was a location problem, though I knew I’d need place information to solve it – and rather than pre-select thousands of venues, I’d want a way to dynamically access a huge volume of place data in any location, whenever the app was run.
So – with my obvious personality problems aside (there are some things even complex algorithms can’t solve) – it was clear that this was a spatial and API dilemma that I could solve using Alteryx.
The rest of this blog is dedicated to the app’s user interface; how it’s built from a technical perspective; and my next steps with it. I’ve included a link to the app on the Alteryx Community to the bottom of this post. Please feel free to download and try for yourself, or just comment feedback – the link is at the bottom of this article. All feedback is welcome.
How to Use the App
Like many APIs, you’ll need to generate an access token to use the Google API. You can do that by signing up here.
Be sure to check both routes and places:
I have chosen to associate a card with my API (as doing so gives you free credit to do lots of calls for free), but be sure to read through any documentation here.
The app’s UI sits across three tabs:
Results
Results are provided in a report, showing: where your travel polygons overlap, the top-rated places within that spot, and some other interesting points and images…
How Does It Work?
There are a few key tools that are integral to the app. Running the app in debug mode allows you to see exactly what each of these tools are doing.
- You dropped two (or more) pins onto a map at the start of the app; these pins will need to be turned into spatial objects, so we can run spatial analysis on each person’s location. This is done using the ‘Create Points’ tool, by taking the XY coordinates and appending a spatial object.
- The points dropped are fed into a Route 360 API macro; this API leverages the spatial point created in Step 1, to create a polygon(s) around each person showing the areas that can be reached by them within their chosen transport type and time. (note the blobs decreasing in size as you get further away; this often the case with public transport, where each blob is (in this case) a tube stop and the rings around them is how far the person could walk from each station in the allotted time they have left to travel)
- Step 2 above is run separately for each person, but really what we want is to combine these two polygons to see if the two journeys can overlap (you will go to the ball) or not. The ‘Spatial Process’ tool allows you to create a polygon showing the overlapping regions, as is shown below:
- What if – like above, we have multiple areas that overlap? To stand the best chance of finding a good place (and to minimise the number of API calls we’ll have to make further downstream), we can use the ‘Spatial Info’ tool to calculate the area of these four polygons and just take the one with the greatest area for our place analysis.
- Now we have our ‘perfect meeting point’ (or polygon with a centre point), but we still need to search for places within it that fit our criteria (in this case, a bar). The search radius you defined at the start of the app comes into play here, acting as a ‘distance from’ the centre point in our Google Maps API call.
- This information, along with the X and Y of the centre point and place type, allows us to build a search URL, which will look something like this:
https://maps.googleapis.com/maps/api/place/nearbysearch/json?location=518746104,-0.110362243&radius=500&keyword=Bar&key=yourGoogleAPIkey- This URL is effectively the search you might have done in Google Maps. We need Alteryx to do the work beforehand to find the perfect meeting lat long, but also, doing it via Alteryx allows us to return many more results than we’d see directly in Google, not to mention, see a lot more information about each place.
- We can input our newly built URL into a ‘Download’ tool, which, along with ‘JSON parse’ allows us to see all the data in a structured way:
- Additional reporting and blob convert tools allow us to build out tabs for the user, to click through a list of places ordered by Google rating, see a map showing their locations, and view some example pictures:
- Reporting tools are then used to generate the maps and tables you see in the final output:
Next Steps
Recent refusal to meet anyone anywhere other than ‘the middle’, as well as lockdown, means I’ve got a lot of free time on my hands to test, improve, and add features. A couple of features I’ve added during lockdown include, but not limited to:
- User keyword search. E.g. you might want a restaurant, but a free text field which is added to the google URL would also allow you to state that the restaurant serves ‘fish and chips’
- Actual travel times for each person; the google maps API allows you to return travel time in minutes and miles, and can also return step by step directions
- Multiple (>2) person functionality. Apparently, some people have more than one friend, so I’ve added the ability to state where multiple people are in the UI
- Ranking by review is good in theory, but poor in practice. If I’m meeting friends from work (assuming everyone works central), the app will often suggest meeting for a pint at The Ritz… so I’ve created an index that factors in all of these variables, including price
- General tidying of the UI: e.g. initially, in the dropdown selection you can see a range of business types that Google will search for you; naturally this lends itself well to bars and restaurants:
… I assume (and hope) that nobody would ever want to use this to find the best rated dentist or zoo between them. But who am I to judge.
The new updated app UI looks like this, and can be run through your browser on any device:
…and the output includes much more detail, including:
- How long it takes each person to travel to each destination (see Andy, Chris, and Dave columns below)
- The mean and standard deviation of travel times (to ensure fairness!)
- Price, rating, volume of ratings
- A ranked index, based on a scoring system I’ve created that weights and combines all the metrics outlined above
Last Orders
I’ve found that though the use case was initially used as a means to learn the Google APIs and continue to use spatial Analytics tools in Alteryx, in 2020 onwards, the places we meet, and how we get there, is a problem more relevant than ever.
As you can see, even now this is a work in progress; I’d welcome any feedback on the app, or to hear from you if you want to work together on this or something similar. I’d be happy to meet at a nearby zoo, assuming it fits my criteria of a rating above 4.7 and a travel time of less than 27 minutes.
I welcome any feedback or thoughts!
Cheers
