I’ve been recently working on an algorithm which sorts through a large list of possible advertising media and suggests a set of suitable candidates for inclusion in a campaign. I’m doing this in conjunction with a company in Hong Kong called AdvertisingPlaza.com who have created a media buying market place. They act as broker and middle-man between advertisers and the ad providers. The media they have on their books if of a wide range, from TV ads, to billboards to newspapers and magazines and online ads.
I’m quite excited…
The revolution is that the media buyers being targetted are not media buyers at all but regular business people buying ad space. This is something I have done a little of myself in the past and I struggled to know how and where to spend my money. I never had the budget or ongoing need to merit using an agency to plan a campaign for me, so I guessed. On reflection I probably got it partly right, but that means I also got it mostly wrong.
It turns out that this is not an uncommon scenario, so what we did (we being me and a brainy friend) is dream up a means of allowing a novice user to tell us about the following things:
- what the demographic of the target audience is
- what the intention of the campaign is
- how much money they have to spend
and we then look through the list of media that are available and come up with a suggestion for them which matches who they are trying to reach, what they are trying to do and how much they want to spend.
We have them tell us these things by filling in an interactive online form – with sliders! – in a web page on the AdvertisingPlaza.com site (oh, btw, if you go looking there right now you won’t find it because we’re still beta testing with selected customers). They press the magic button and in the background our algorithm chugs away and they are presented with a list that they can either just buy straight away, add to or change quantities, or contact the sales team to start haggling about price.
The clever bit is how we decide what we should include. It turns out this is actually quite a complex problem, not least because the advertising media tend to overlap – by which I mean that a billboard and a newspaper and a TV ad will have some audience in common. What’s more they individually have very different reaches and different impacts on the viewer.
Then there’s the question of cost. It’s possible to come up with the perfect media campaign, but how do you maximise the effect and still stay within budget?
These were the problems that we set about tackling and we have come up with a very novel way of doing it which doesn’t mean a huge computational cost as well. So how does it work? Well, that would be telling. Just as soon as we have some solid commercial success I’ll spill the beans, but until then it has to remain top secret. Suffice to say it has Flex and Python in it…