The short version

Yes. But with qualifications. Which may be very important.

The long version

Here goes…

An econometric model, for the benefit of any laymen who may have wandered in here by accident, is an approach to marketing attribution that takes your historic sales and decomposes them by the things that caused them (we call them ‘drivers’). Take the last 12 months, for example. Suppose your sales were £20m. Some of them came from market demand, some from seasonality. Others came from promotions. Still others were due to customers seeing one of your adverts and deciding to buy your product. Econometrics takes the £20m and says, of this, 10% was due to advertising, 20% to promotions, 5% to seasonal events and the rest (65%) were down to your distribution and ‘brand’ and would have happened anyway. An example, you understand.

The next step in any econometric study is to calculate the ROI of the marketing activities. In the example above 10% of £20m sales was down to the advertising you did. I make that £2m. The cost of the advertising was £750k. Your ROI (in revenue terms – ignoring costs of goods for simplicity) was 2.67.

If you think about it, what we have done is work out the relationship between media and incremental sales. £1 invested in media gets you £2.67 extra in sales. Now, looking to the future, suppose you tell me you plan to spend not £750k on media in this coming year, but £1.2m. Couldn’t we use the relationship we just calculated to sort of predict the sales we will get from our £1.2m investment…? I think you see where we’re going with this.

And, in essence, the answer is ‘yes’. Take the £1.2m planned investment, multiply it by 2.67 and you get £3.2m of sales driven by your media. It seems very easy, right? And, in a sense, it is. But if, and only if, the relationship between media spend and incremental sales remains unchanged from last year. There are at least 3 issues with what we just did. Not issues that necessarily make it incorrect, more assumptions we are making, perhaps without realising it.

1. We assumed there were no diminishing returns in our media investment

We took our measurement from our econometric model during a period when the media investment was £750k. Now we’re proposing to spend £1.2m, an increase of 60% in the media budget. Will the additional £450k work just as well as the previous £750k? It might do. But it might not if the extra spend goes into bombarding the same people with the same message which they’ve already rejected.

Diminishing returns is something you should definitely not ignore, particularly when you are considering significant changes to the overall amount you are proposing to invest. The good news is that you don’t need to ignore them, and your econometric model should be able to give you a read on them and allow you to adjust your forecast accordingly. As long as you know to do so. Which you do now.

2. We assumed that media will cost the same in next year as in last year

Media inflation (in fact, inflation generally) is rampant at the time of writing (November 2022). You might be planning a 60% increase in your media budget, but if the cost per rating (or click, or thousand impacts) is set to increase by 20%, you will only have a real increase of 40%. It sounds obvious but, perhaps because media prices are not always that transparent, it is often overlooked.

3. We assumed that the relationship between media spend and sales remains unchanged

Of the three issues, this is the most subtle one and the one with the greatest potential to undo our efforts at soothsaying.

Look at it this way – why might the relationship between media spend and sales change? One reason could be that you’re planning to use a new ad. Now, if it’s similar to the last one, on which our historic measurements were based, you might be ok. But what if it’s radically different? Consumers may respond to it differently from how they responded to the old ad. In fact, you’re probably hoping so, and that they respond even better than before. But it may be different.

Or, another reason, suppose that your market has become more contested with 2 or 3 new entrants all competing for the same sale. If they’re shouting as loudly as you, it could be your voice is harder to hear, meaning you need to shout even louder just to be heard as before.

Or, there has been a global pandemic which has decimated the market for what you sell. That happened recently.

Or… anything else, really. The important point is that when we do any sort of marketing measurement or attribution, what we’re doing is measuring a human response to stimuli. Whilst there may be some stability and repeatability in that response (if there isn’t, we may as well pack up and go home now), we are not measuring anything as deterministic as a physical law. It’s an ROI, not Planck’s constant.

So, should we use an Econometric Model to tell the future?

Yes. The point here is that businesses are going to plan and predict with or without a model as the starting point. And, we would venture, doing so is likely to end better with some sort of attempt at objective measurement – which is what a model gives you – at its heart, than mere wishful thinking or waving of fingers in the air.

But. As we also say about econometrics, it is not a panacea, nor a silver bullet, and using it with knowledge of its hidden assumptions or limitations will leave you in a far better place than blind subservience to a model.

If that’s whetted your appetite to know more about econometrics and how it might help your organisation plan better, then download our Guide to Econometrics for the Modern Marketer or if you already have some thoughts on how an econometric model could help, get in touch.


Guide to Econometrics for the Modern Marketer