Econometric Modelling in Advertising Explained


For the second article in our series explaining misunderstood industry terms, we’ll be looking at econometrics.

Econometrics has been used in advertising for decades, with Sir Martin Sorrell describing it as the ‘holy grail’ of marketing back in 2005. And as advertising has become more complex with the growth of digital channels, econometrics has become even more valuable.

But nonetheless, econometrics is still rarely directly discussed in our industry. And there’s still an element of mystery for some around what it actually is, and is useful for.

The Basics

Econometrics is a practice which uses statistical techniques to help describe economic relationships.

Economists use econometrics to build models, which they can use for forecasting. To use an example outside of advertising, an economist might want to understand how average wages are affected by economic growth. They could use econometrics to help describe that relationship, showing how much average wages change when the economy grows. And they could then use that to forecast changes in wages, based on how they expect the economy to grow.

These same techniques can be used in advertising. A marketer might, for example, want to show the relationship between how much they spend on TV advertising and growth in sales. Econometrics would help them do that.

These econometric models can be used for measurement (showing the return on investment from past campaigns), but also for forecasting (showing the expected impact of future campaigns).

The Technical Detail

Econometrics uses regressions (a statistical technique), to find and describe relationships within data. A regression analyses historical data to show how an outcome (for example, product sales) was affected by other factors (for example, TV ad spend).

Advertisers can then use econometric models to predict how changes to their marketing strategy might impact sales or other outcomes. For example, a brand might use econometric modelling to predict that if they increase their TV ad spend by £200,000, they should expect total sales to increase by £400,000.

An advertiser may want to run a regression to create models based on their own data. Or companies and organisations may use industry-wide data to create models which can be used by the whole industry. For example, econometrics has been used to answer questions around how long an ad campaign’s impact lasts, and even whether advertising is effective at all.

Regressions can be used to measure the impact of multiple channels at the same time, showing how each specific channel contributed to the final outcome. This is why econometrics in advertising is often called ‘media mix modelling’ or ‘marketing mix modelling’.

These marketing mix models let advertisers measure their past campaigns, by showing which channels are having the biggest impact on outcomes. And the models can also be used for optimisation, helping advertisers understand where they should spend money in the future.

The models can also draw out relationships between different media channels. For example, spending more on TV might increase the effectiveness of a brand’s social advertising, and econometrics can help identify these relationships.

One more thing worth bearing in mind is that econometric modelling is not the same as attribution modelling.

The two are similar, in how they are used to draw out the impact of different media channels. But the Institute of Practitioners in Advertising (IPA), an industry trade group, says the two are different in their application. Attribution modelling uses more granular data to draw out short term trends and relationships, whereas econometrics uses less granular data, and shows longer term and widely applicable relationships.

So for example, attribution modelling may be used to show the impact social media ads had on app downloads for one particular campaign. And it does this by looking at individual consumers’ behaviour – which ads they saw, and which ad they clicked on. Econometrics meanwhile looks at top level trends, like how effective social media ads are for driving app downloads generally, rather than for one specific campaign.

Pros and Cons

Econometrics helps take guesswork out of advertising, giving brands direction on where they should be allocating their ad spend.

And it also helps prove the value of emerging channels and formats, encouraging investment in those areas.

But econometric modelling is very difficult, and in some cases impossible.

Econometric modelling is a specialised skill, meaning a brand or agency can’t create a model themselves without hiring a specialist to do it.

There are a number of third parties and measurement companies which offer econometric modelling, including well known tech vendors like Nielsen and specialist consultancies like Entropy Consulting and Brightblue consulting. But an advertiser or agency still needs to understand econometrics well enough to know how to commission a model, and how to interpret its results. There are resources available to help with this – the IPA for example has a 60 page guide outlining the basics of econometrics.

Econometric modelling also takes a long time, and can be expensive, since experienced econometricians are in short supply. Econometrics works best with big data sets, and collecting, organising and processing this data can be a slow process. And as the advertising landscape evolves, these econometric models have a limited shelf life, meaning they have to be regularly tweaked.

The final major issue is data. Brands might not have the necessary data in the first place to be able to build the model they want to. Even when brands do have the right data, econometrics in some cases will struggle to pick out the impact of smaller channels.

Media consultancy Ebiquity outlined this problem earlier this year. Ebiquity wanted to use econometrics to prove the value of influencer marketing, as clients were increasingly keen for data on influencer marketing’s effectiveness. But the company was unable to run a national test using econometric modelling, because influencer marketing is too small a component of most brands’ campaigns.

“Consumer goods, telcos, and retailers spend so heavily on other media lines – particularly TV, often with hundreds of TV ratings (TVRs) per week – that even the most sensitive model would find it hard to detect any signal among the noise,” the consultancy said in a blog post.  “Influencer marketing simply cannot deliver the reach and frequency of TV.”

But this problem can sometimes be solved simply by restructuring the test. Ebiquity for example tested small, regional econometric models, and found that these were more effective.


Subscribe to Weekly VAN Newsletter

 
AgencyBrandsDataMeasurementTech

Leave a Reply