The worst customer experience I ever had went something like:Cust: Hey – we just bought 40k worth of software – help us make some “analysis”. Me: I…eh…uh Cust: Here’s some data – make a map Me: OK – <magic> Here! Cust: OH HOLY CRAP THAT CAN’T BE RIGHT. OH SH*T IT IS RIGHT….. Me: Hey – it’s your data – I’m just putting it on a map and making some “analysis”. I AM HERE TO REPORT THE TRUTH
That was early on in the consulting career. Since then I’ve tried to be a bit smarter about how data looks and how the customer is going to react. Mainly the customer. Controlling expectations is an art I have yet to master but I’m getting better at it. There is absolutely nothing worse than making a map and having the unexpected truth appear in front of them.
So one of my clients gave me (and anything worthwhile has been removed to maintain some confidentiality) this which is equal to about 30,000 geocoded points.
There was the general talk of “make some analysis”. This all started as a spreadsheet but that’s another story. Questions is how do you make sense of that many points? The map would have to be pretty large scale and the customer isn’t going to want to dig through that many points to make a decision about their data. I ran into this article on “hex binning” a few weeks ago on GISLounge which I found on Google Plus (because for all its goofiness and lack of users – I enjoy it more than Facebook – try it out). I could do a point density map – and while pretty it’s not going to help matters much. Heat Map? Hot Spot? So I decided on binning.
The Entire point to binning is attempting to make order out of disorder. Taking point data and creating meaning out of scattered data. Why a hexagon? Why not a rectangle? I could have done a rectangle – but hexagons look (to me) more natural than a rectangle. Former Geologist here.
QGIS has a plugin that does hexagon grids (MMQGIS Plugin). ArcGIS has a plugin/script that does that also. So out of that pile of 30,000 points:
Anything under 5 points I discarded based on what I know about the customer. Yes – the legend is a bit odd number wise – that will be fixed. The brighter red actually forms a pattern (which you can’t see because I’ve pulled any other data). This is easy to explain to the client. It makes sense. I ran the area of the Hexagons to be comparable to a 7.5 minute quad for this exercise. Anything smaller and there was too much (and I’m losing the empty hexagons for the final map) detail. Anything bigger and detail was lost. Overall – not bad for my fourth attempt today.
Make a decent map….or at least try to.