TV audience measurement – why big isn’t always beautiful
In this guest post, Chris Walton argues that the media industry needs to take a new approach to TV trading
There has been a significant amount of coverage recently about how successful The Voice has been. Indeed, audience figures of 2.6m+ people are very impressive these days. Based on reports, this is apparently double the size of audience that Nine was hoping for in the lead up to the programme launching.
There must be some very happy people at Nine. Whilst a successful TV station is not built on a single programme, you could be forgiven for thinking The Voice is almost single-handedly saving the network at a time when it most needs it. This is largely because of the much increased rates the network is now charging (one hears six figure sums bandied about) for spots that air during the programme. And who can blame the sales team for doing this? Make hay whilst the sun shines.
However, the success of The Voice in audience size terms has highlighted how woefully inadequate the entire approach to media trading is. ‘The Voice won in all demographics’. So what?
I do want to make it clear that this isn’t a beat up on Nine at all – all they are doing is capitalising on the opportunity the current way of doing things presents them with. They are now able to proudly boast that more people who are aged between 25 and 54 years of age watch this show and their channel than any other. They can claim it for The Voice with any demographic group they choose.
So, we know the demographic make-up of who is watching the show, and let’s put aside for a minute the debate about the size of the OzTam panel and its ability to accurately tell us even this on anything other than a very broad level. But if this is what we know about them, what do we not know about them? Casting an eye over the main sponsors, as well as those brands that regularly advertise in and around the show, the following questions occur:
- What proportion of this audience currently drives a small car? How many of these are actually in the market at the moment for a new small car in the right price range?
- Which audience members have demonstrated switching behaviour when it comes to choosing a telco provider? For existing customers to what extent is mass marketing greater at up-selling than direct communication?
- What proportion of the audience travel on a regular basis? And when they do travel do they pay premium or budget fares?
- How many of those viewing demonstrate loyal supermarket purchasing behaviour?
We don’t know. At least not accurately and in almost real-time.
Huge amounts of money are being spent based on woolly hypotheses that are linked to fluffy data. The amount of wastage this creates is staggering.
The vast majority of marketing dollars are spent trying to get people to buy stuff. What would a small car manufacturer prefer – paying to reach 2.6 million and selling say 30 cars or paying to reach a million people and selling 50 cars? What marketers are looking to buy and what media owners such as Nine are currently selling are two different things. Indeed, whilst we have all heard a thousand times or more that the media landscape has changed and continues to change, the approach to trading has not…its been the same for decades – a very broad shotgun approach that has no link to actual buying behaviour. It is stuck in a time warp that values two pairs of eyeballs more than one pair, regardless of how much more (or less) likely the larger number are to actually buy something the advertiser is selling. In other words, size of audience is secondary to relevance of audience and the resulting likelihood to buy as a result.
The current trading system, in rewarding quantity over quality is doing us all a disservice, including media owners. False peaks are being created due to an over-demand for high rating shows. False troughs are also created by the evaluation and pricing of airtime being based on broad and woolly demographic definitions, rather than the relevance, response and sales effect which different environments provide.
The time is nigh for a new approach to trading that leverages a deeper and more accurate understanding so that advertisers can spend their money and place their brand in places where their paying customers actually are.
- Chris Walton is principal of Quantium Media Solutions. He is a former CEO of media agency MindShare
You only have to look at the ads running on Hobart WIN during The Voice. There seems to be no logic at all. It really is a scatter gun approach, and I doubt the many local advertises are seeing value for money.
It certainly seems that WIN has sold and programmed ads with only one thing in mind – does the ad set fill the time required for the ad break. No consideration seems to have been given to the audience demographic at all. An example –
Cars
Funeral insurance (no medicals if you are under 70!)
Computers
Computers again – I wonder how these two businesses feel about competing ads running back-to-back
Sewing machines
Cars – a recap of the first ad
User ID not verified.
Fine words Chris.
Ratings are somewhat antiquated now, surely there must be a better system to guide where marketing dollars are best spent.
I suppose no analytic system is perfect, however isn’t it that roughly about 3000 people decide the nations viewing destiny?
Dan
Chris if you really want to talk to an in market Small Car buyer you’d look online where you can really target them with accuracy. Truth be told the modelling even your company uses still relies on legacy data. There is also a huge disconnect between a TV ad playing and someone walking into a car dealership regardless of the targeting, and relevance of the offer.
TV is a mass, broad channel, and targeting is really only one consideration. Using properties like the Voice is more about brand association and getting into popular culture.
User ID not verified.
Many marketers have a huge over-reliance on the objectives TV can help to deliver. It is a mass reach platform and a blunt instrument.
Other channels cost-effectively provide tight targeting and accountability to a media plan (targeted digital media for example). We shouldn’t expect TV to necessarily do the same when it does other jobs better – building brand profile, emotional connections, content links etc etc. Why try and force fit (and measure) an objective in an innapropriate channel? By this standard, simply seeing how many prospects in your audience potentially saw your message is completely appropriate to the objective, and from this point your other channels pick up and narrow the focus.
The issue is that CPMs across media are back to front in this sense, where TV CPMs are overvalued and digital (even the most tightly targeted options) is heavily commoditised. This is because of the historical value of TV on a plan as the champion chanel, but not reflective of consumption and technology today.
User ID not verified.
Chris – good food for thought. Thanks!
Brett’s comment about brand association and popular culture makes sense to me. (But is that the ‘woolly hypotheses that are linked to fluffy data’ you refer to?)
Your comments about lack of (ability for) targeting via broadcast ring true too.
In today’s world that’s where brands should try and take advantage of the handheld screens on in tandem with the big screen. ie Until there is the “new approach to trading” – and with it more accountability – that your article craves, brands will continue to use broadcast media for “fluffy” measures and start to (more intrinsically) link great social media activation with their broadcast media buys. This can provide more engagement and more accountability.
User ID not verified.
Spot on Chris!
User ID not verified.
Chris, a thought-provoking and well-argued post.
I’m not going to comment on the relative technical faults and merits of the current legacy trading system. But I will comment on how to overcome some of these issues in relation to the balance of audience vs. campaign knowledge.
On this discussion, I’d definitely draw attention to the several points John Grono made underneath this previous Mumbrella post:
https://mumbrella.com.au/the-voice-rates-2-6m-for-nine-ten-gets-less-than-10-share-88744
As John says, we shouldn’t criticise an audience measurement system for failing to be a campaign measurement system as it wasn’t set up to do this. For specific campaign measurement I think something more bespoke is required e.g. media agencies can tap into any one of a number of online panels ‘omnibus’ service.
Clients and media owners have another option, as owners of consumer-facing brands and thus brand equity – they can set up and maintain a branded online Community Panel comprised of customers/viewers etc which they can then use as the “Voice of the Customer.” Or for non-customers, they could tap into an omnibus service as already mentioned.
Whether it’s a community panel or an omnibus panel, participants are profiled by demographics upon joining. From then on, a series of regular surveys could ask about psychographic and buyergraphic factors relevant to specific ad campaigns – but obviously this would be highly selective and couldn’t cover all ad campaigns.
In this way, a mass audience can be segmented into niche audiences as required, and then the target consumers can be asked about all the things we don’t know from audience measurement data i.e. campaign-specific things like the ones you listed Chris (whether in the market or not, consideration sets etc).
As an example, during my time at HWT (aka the major Victorian division of News Limited/News Australia/News Corporation) on the Herald Sun brand, audience measurement data was used extensively but was also supplemented with custom insights from a branded online community of regular readers (we called it the “Herald Sun online reader panel”).
We worked with numerous clients and/or their media agencies to help them answer many of the questions they didn’t know about their campaigns in our media outlets – the questions were regularly “did our advertising work?’ and “was our campaign effective?”. The answer was often “yes” and the community panel provided *some* evidence for marketers, as an added-value service to them.
Was our media owner community panel the perfect cross-media, media-neutral campaign measurement tool? Of course not. But it was a lot better than having nothing in place or assuming the client/agency would have all the campaign measurement ducks in a row to know how effective each channel (and each brand within a channel) was at driving claimed action or behaviour change.
So in summary, Chris I think you’re right: the media industry should take a new approach to trading, by continuing to supplement (syndicated) audience measurement with (ad hoc) campaign measurement.
Whether that is through econometrics or community panels or something else is a debate for another day.
User ID not verified.
Agree, big doesn’t necessarily mean better.
It’s a topic I’ve also been thinking about. My point of view is available here:
http://simonislawson.tumblr.co.....ice-for-tv
User ID not verified.
An interesting post Chris. As you know I am a big fan of econometric modelling as I think it is at THIS end of the marketing spectrum that advertisers learn what works for THEIR brand.
However, to extrapolate that approach to a trading system – which as you know is underwritten as a ‘currency’ – is a vanity of mathematics. My reasoning is that every advertiser would want “their target” as their trading currency. Therefore, if you look at cars we would end up with around 100 car demos at least. Any such system would collapse under its own weight.
While this approach has definite appeal youalso have to look at the validity of such ‘classification’ data. As an example I bought (online) a trekking watch. I researched it online and I bought it online. Being a niche product online is the perfect medium for it. The problem is that three years later I am STILL getting ‘targetted ads” trying to flog me the latest Suunto or Protek. No thanks – purchase decision was made ages ago. Same goes for cars. I’m very happy with my Audi A4 I bought two years ago. No more online ads please. How many Audis and trekking watches do you think I need! So much for the way this type of targetting is being done – a complete waste of time and money. However, if the same data was analysed temporally then the ads would have stopped long ago, the gross volume of ads would decline, and the advertiser would have saved money.
So, why DO mass programmes like The Voice work?
It’s because if you use a probablistic approach you will see why. Take The Voice (and using a single recent episode) as an example. It TARPd 17.9. That is, around every fifth Australian was watching it. If you look at the distribution of that audience it has some ‘hot spots’ (40-54, females, GBs, kids in the home – around one-in- four) and some ‘cold spots’ (P18-24 and Under 18s – around one-in-eight)
The thing is, the users or target-users or all brands must fit into (at least one) of the age/gender demos we use. The above data shows that between one-in-eight and one-in-four had the opportunity to be exposed to the ad. And all in a single two-hour period.
So by matching the profile of the brand to the probabilties of the audience, a brand can maximise the probablity of exposure of their ad to their most likely customers. Of course, the target may still do nothing in the short-term after seeing the ad, and the ad may be ‘filed away for later’ and become part of the brand equity.
Further, not all advertising is meant to have immediate effects. A clear demonstartion of this would be advertising things like refrigerators. There are around 9 million homes in Australia. If we assume each home has a refrigerator that has to be replaced (or freshly installed) every 10 years we’re talking 900,000 fridges per annum or around 18,000 fridges a week that are consumer purchase decisions (fridge sales are higher but not all are consumer choice).
So, the objective of advertising a Kelvinator fridge in The Voice is NOT to get people off their couches and down to Retravision to buy a fridge that week. The purpose is twofold (a) to influence people who are about to buy a fridge to buy a Kelvinator – the propinquity model (b) to ensure that when they next buy a fridge in a few years time that they at least have Kelvinator in their consideration set – the awareness model. The fact is that the majority of people fall into the latter and judging the success of the advertising just on the former undersells the effect of advertising. The ‘awareness’ people are filling the purchase funnel – even if the eventual sale happens online when there is no advertising activity.
Econometric models can tease this out by looking for sales spikes in proximity to advertising activity (quantifying the propinquity effect) as well as looking for an upward trend in the base-line sales, or even better base-line sales share, (quantifying the awareness effect).
So, my take-out is:
* don’t walk away from large advertising vehicles just because the data reported is age/gender based
* instead leverage the audience probabilties and match them to your brands purchasing propensities and reap the rewards. (This works for all media – not just TV by the way).
* and as an aside make sure your targeted advertising has a recency cap on it (such as … has searched for the brand, competitor or category within the past x days).
Sorry to bang on for so long!
User ID not verified.
You will be happy to hear that the new trading model you seek is not only nigh Chris it’s up and running. Give our Business science department a call…
Nick Keenan – Mediacom
User ID not verified.
Does anyone know how the SMH’s experiment with “choose your ad” went? (On the pre-roll for videos you had to choose which of two ads you viewed). I thought it was interesting and potentially a great idea on a number of levels, I don’t know why more companies don’t do something similar.
As TV moves onto the web, I can see similar moves happening. So in The Voice (for instance) at ad break time you can choose from ads a or b, gradually becoming more personalised as they collect data about you. Then option c could be for X “credits” you can skip the ad break completely. Credits can be paid for, or you can “pay” for them by performing actions, filling out a survey etc.
I also always wonder why they don’t have special streams when sporting events are shown in pubs – surely a massive swathe of viewing figures are in pubs, and the ads are totally non-optimised for that kind of environment (eg ads that work without sound, more beer ads etc). It’s a fairly easy way of targeting the likely watching demographic, anyway.
User ID not verified.
Interesting article, however I have to agree with John Grono’s comments. Ratings are about trading. Identifying the targets (planning) your advertising may have to take place using different tools and data. We are always victims of sample size to some extent. No one is happy having to move from precise targeting to a demographic surrogate for that target group when trading but that is the way of the world.
Census level RPD data sets allowing for addressable advertising may well come close to solving these issues, particularly when the RPD data includes online activity conducted on a connected TV. Some of these solutions may be closer than we think. The role of RPD in (virtually) expanding panel sizes is likely to be the bigger breakthrough in this area than econometric modelling.
User ID not verified.
What would a small car manufacturer prefer – paying to reach 2.6 million and selling say 30 cars or paying to reach a million people and selling 50 cars?
What an amazing point! Now we just need to find out the media that people use when they’re in the market for a car, easy! Or is your point that advertising in general is a waste of money? Really not sure, but I liked your hypotheticals that have no basis in reality whatsoever.
2.6m is massive ratings and the cost per person is relatively cheap. Even if only a small percentage of those in the market see the ad, that would be an extremely large number and would still end up cheaper than any targeted niche media or direct selling method.
User ID not verified.
I agree Researcher – if and when it happens. There are a few issues with RPD, but all are surmountable are can be taken into account.
The first is that at this stage RPD would be contained to those with a Foxtel, Austar or similar box – around one-in-three. Having said that household targeting with precision a third of the market is a win in anyone’s game (I’m just not sure I would call it ‘census’ RPD).
The second is that not RPD data is (generally) tuning data. Some substantive edit rules would need to be in place to ensure that the RPD data represents valide tuning (i.e. not tuning for hours on end to the same channel, indicating that while the STB is on that the TV is likely to be off). Again, a win in anyone’s game. It’s not quite targeting the person but it is targeting the household that person lives in and WOM can then do its thing. The UK Superpanel data of the 1990s clearly demonstrated the efficacy of such a strategy.
And Researcher … drop me a line … I’m dying to know who you are!
User ID not verified.
Yes John, it is imperative that a high quality and validated capping algorithm is applied to the data to clean up end of day tuning issues.
It will also be possible by modelling OZTAM data to fairly accurately determine who is watching within the home, in a recruited RPD panel scenario.
It is also then possible to project this across the whole subscriber base as long as the panel is representative of the whole subscriber base and all subcribers return RPD data.
It won’t be long before virtually all homes receive TV through an STB and as long as these are RPD enabled, sample size issues may be a thing of the past.
User ID not verified.
Evironment and association extremely important – just dont get our global auditor that 🙂
User ID not verified.