You Read, You Watch, You Listen, You Buy – Don’t You?

Sue Moseley, strategic planning director, TSMS

Time after time, we know TV advertising works. The fact that it has grown into a 2.5 billion pound business is proof of that. However, there are still a great number of unknowns such as whether in large, stable, fmcg, markets advertising does actually produce incremental sales in the short term – or is its role purely one of longer term branding? The best way to gain this understanding is to measure purchasing behaviour and advertising exposure from the same source. This is precisely what we have done with TVspan.

TVspan is a joint venture between Meridian, who have made a significant investment in the project, TNAGB and TSMS.

Since March 1996, 750 Superpanel homes in the Meridian region have been equipped with BARB TV set meters in addition to the scanning equipment they already had to record the details of their purchases of fmcg products. The TV set meters cover all sets in the home, and record minute by minute viewing to all channels including the one or two who can receive Channel 5. Because they are set meters the measurement is entirely passive. Panel members are not asked to do anything more than what they already do – which avoids any respondent overload. TVspan provides a continuous measure of both product purchasing and TV viewing from the same households. The information we have across a wide range of fmcg fields, is for each purchase occasion, the brand and quantity bought, price paid, source of purchase, day of purchase, special offer details. For these households we also have minute by minute viewing which enables us to measure their exposure to specific advertising spots.

TVspan has two uses:

1. To improve media planning through more precise targeting.
2. To evaluate the sales effects of TV advertising.

Back to the main attraction: does TV advertising work short term and if so can we create a bench mark average against which we can compare different media plans?

The way we approached this was to take these heavily advertised and frequently purchased mass market fmcgs: Butter etc. Across these fields for each TVspan household we analysed all their purchasing and advertising exposure for the 6 month period of March – August 1996. This database comprises 193,000 purchase records with 69 advertised brands utilising 15,365 TV spots, providing a wealth of data.

The approach we took was to calculate an average brand share for each household. Irrespective of whether a TV ad had been seen across the period this provided us with expected purchasing behaviour by household.

We then divided purchase occasions by whether advertising had been seen prior to the purchase or not. This enables us to contrast the actual purchasing with the expected. If advertising doesn’t work then there would be no difference between the two.

We have taken an open minded, questioning approach.

The data has been controlled for the potentially biasing effects of the purchasing/viewing relationship where heavy viewers tend to buy branded products and where TV advertising is undertaken in association with other marketing activity, most notably price-cutting and multibuys. Previous single-source analyses have not always taken account of these factors and we have found if you don’t you can end up with misleading results.

So in answer to my original question of ‘What is the average short term effect on purchasing of TV advertising being seen within the week prior to a purchase’? Well revealed here for the first time. The answer is an extra 3.9% That is the incremental short term effect of having seen advertising in the week prior to purchase. This is an average across all the 69 advertised brands in our analysis and we have found some interesting differences between product category and brands (that is when sample sizes allow).

Now you have had time to digest this figure you’re probably wondering well actually ‘what did I expect it to be?’ If it had been massive – you’d have known it anyway from basic sales analysis. Just remember the markets we are looking at here are large, well established and heavily advertised. Arguably their key objective is one of brand maintenance. Our 3.9% is on top of the long term brand building effects of TV advertising.

To put it into context if you take a typical retention rate of 70%, then our 3.9% could grow to 3 times that size over 12 months. Just work it out for yourselves.

Right now we’ve got a benchmark – the challenge is to use the data so we can improve advertising effectiveness. So far we have looked at four areas which historically have been subject to a lot of theorising. Now we are able to put some hard figures behind these issues. These areas cover: The duration of advertising effect, in other words how long the advertising has an impact. Advertiser dominance – which effectively is clout. The ways in which different types of viewer respond to advertising. And what time of day is best for your product.

Our benchmark looked at advertising in the week prior to purchase. By extending this period to 2 weeks we see an increase in our advertising effectiveness measure which is now generating incremental sales of 4.5% indicating that advertising does act over a longer period. By extending the exposure window even further to 4 weeks the incremental value is pretty close to that for one week at 3.8% indicating that on average the short term effects are starting to dissipate after two weeks. The duration of effect is also determined by associated marketing activity. Where ads are standing on their own, with no associated promotions short term effects have been measured up to 4 weeks.

This is good to hear, and not really surprising to find that advertising works over a longer period of time than other work of this kind has indicated.

This finding has implications for coverage build and the pattern of that coverage delivery, but to understand that fully we do need to undertake further analysis to identify the effective frequency levels. Now we approached this issue of understanding the significance of share of voice and advertiser dominance by identifying all those purchases where the ads seen in the week prior to purchase were solus to the brand. That is, there was no exposure for the individual household to any other brands in their repertoire in that week. This produces a marginal effect, increasing the incremental short term value to 4.1% suggesting that share of voice will prove to have a relevance in media planning.

Research has demonstrated a variation in response to advertising amongst different types of individual. We split the TVspan panel into 5 viewing groups according to their weight of viewing of commercial television. The heaviest viewers showed the highest level of response to TV advertising. This group provided us with the most associated advertising and therefore solid data for analysis. The results showed a clear indication that, the heavy viewing group are major responders to television advertising with a short term incremental sales effect of 10.6% substantially higher than that of our benchmark. This does concur with other work done in this area and we know that heavy viewers are enthusiastic brand buyers who appear to like to reassurance of brand advertising. They are also above average purchasers overall which makes them a valuable group.

By having split the data into 5 groups it is getting a little thin across the other viewing groups. However, we are seeing above average responses from light viewers too particularly when measuring their exposure in terms of adstock. The indication here is that heavy viewers require regency compared with light viewers where ad effect has a longer duration. Another big TV issue is whether television is just a comodity best expressed in volume of impacts delivered? We have segmented our database to extract those purchases where the ads seen in the previous week included at least one peak time spot. This group have been contrasted with our base of no advertisements seen in the previous week . The results show a considerable increase over our benchmark producing incremental purchasing of 5.6%. In a way perhaps not surprising as this segment offers advertisers high impact, mass appeal where virtually everyone is available to view and they do. Other research has identified that this peak daypart has higher than average attention scores and that may well be a key factor in this. That summarises our findings to date.

The next step is to extend the analysis to a 12 month period to provide further depth and scope. Particular areas of further development are; linking the duration of ad effect to weight of advertising, the whole area of share of voice, and of course, responses by differing groups of people, not only in terms of their viewing, but also in terms of their purchasing behaviour, such as brand loyalty.

But, we have already reached some valuable conclusions. Short term ad effect is longer than 1 week – measurable up to 4 weeks. Heavy viewers respond more actively to TV advertising Peak viewing has an incremental sales effect

Finally perhaps most importantly Television advertising has a short term positive effect even in large established, stable, fmcg, product fields. TVspan proves that television works time after time.