UK Radio Player anomaliesFollow @mediaukdiscuss
Do shoot me if I’m being Mr Pedantic here, but I’ve just been browsing through the radioplayer.co.uk list of stations.
I wonder if ‘Heart Angelsey’ gets a over the average number of listeners based on the fact that if you click on ‘H’ on the list of stations you get a long list of Heart stations, unless you hover over each one you have to identify which is “your station” by the frequency listed.
I wonder how many bemused listeners are surprised if they click on the first one in the list without noticing the ‘Angelsey’ bit and are greeted with Welsh language content, if they just so happen to click at the right time of day.
This does make the assumption that most people use Radio Player via the radioplayer.co.uk landing pages. It’s intended to be discovered via a radio stations existing website – so via heart.co.uk, in this example – where care is taken in guiding a listener to the correct station website. I don’t believe Radio Player was ever intended to become a single page listeners were directed to, to then find a station to listen to.
When typing my London SE postcode into the search engine, it brings up the usual stations and for some reason, Town and Country Broadcasting stations Scarlet FM, Radio Ceredigion, Radio Pembrokeshire, Bay Radio, Radio Carmarthenshire and Bridge FM.
THANK GOD! I’ve spent the last 18 months obsessing about the detail of how stations are presented to users, so it’s great to find people who are equally interested.
Running Radioplayer is often an art rather than a science, and one of the things we’ve learnt is that there’s no right way to do anything.
Take search, for example. The comment above is actually music to my ears. Martin says ‘When typing my London SE postcode into the search engine, it brings up all the usual stations…’.
Phew! It’s taken a fair bit of effort to ensure that those ‘Route One’ searches (stations, places etc) return the obvious results.
Our ‘raw material’ is the metadata from each of the 330 stations in Radioplayer – there’s no database intervention from us at all.
So in designing our search algorithm from scratch, we learned a lot about how to ensure ‘the usual stations’ appear as expected. Especially given the variable nature of our source data.
Beyond that, we want to encourage some serendipity too. So once you’ve scrolled past the most relevant results, you’ll often see some vaguely surprising stuff.
In this case, it’s probably recognising your search characters as a postcode, and, once it’s given them a strictly location-based interpretation, it’s adding other results at the bottom which match in other ways. Maybe the ‘se’ part of the code?
Of course, we could filter those ‘extra hits’ off the list, but we want to make sure a wide range of radio is offered to the user….. always with the most strongly relevant results first.
We’re in a constant state of development, which is why feedback like this is so useful.
Remember – if you’d searched for anything on the 1995 version of Google, you’d probably have got some pretty strange results!
Many thanks for taking the time to comment. Mike.
Interesting – how about doing something cleverer on proximity for postcode searches? You’ve got centroids for the postcodes, presumably, and each station’s metadata includes a bounding box (two points of lat/long). I’m guessing that currently the approach taken is to return results where the centroid falls within the box. Wouldn’t a better approach for getting relevant/useful results be to calculate the centroid of the bounding box (and a radius) and figure out, the closest stations, perhaps with larger stations being weighted lower than smaller (radius) stations? This would promote niche/local/community stations while still getting the big stations accurately listed for searches on postcodes and locations…
As for composite searches – postcodes are very easy to identify with regular expressions. And at first glance, this sort of search (say, ‘tw20 0ex rock’) gets me BBC Surrey plus some rock stations. But the location portion is still somewhat lacking – ‘tw20 0ex’ doesn’t pick up Eagle 96.4 FM, for instance, not even under ‘More results’, despite it being closer and more relevant than all the Absolute stations listed in that search.
It’s got a way to go and certainly has some bugs left to squash, for sure.
Suggestion for the RPUK/UBC team: Do you log how often the ‘more results’ button is clicked and have you considered adding a button for “I couldn’t find what I was looking for”?
It’s worth pointing out that if a postcode doesn’t return the result you’d expect, it’s more than likely because the station hasn’t entered that in their postcode metadata, not necessarily because the search algorithm has failed.
It’s up to each station how much or how little data they choose to contribute.
‘more results’ clicks are not tracked, but that might be an interesting statistic.
I’ve put in a Wimbledon postcode and Radio Jackie is the first entry shown, so well done to Jackie for that. The Town and Country stations from Wales also show in this list!
We want stations to have as much control over how they appear as possible. Wherever we can, we avoid editorialisation and rules. So a station can describe its geographical relevance however it likes.
But do bear in mind that the fewer geo terms you include in your metadata, the more ‘search power’ each one has, if it’s hit. So it pays to define your localness properly.
In the above example, I imagine Jackie has done just that, and the T and C stations have chosen to define themselves much more widely. That’s why they appear much lower down the list.
If any station wants advice or guidance on how to set up their geo metadata, drop me a line via the contact page at www.radioplayer.co.uk.