Long-term And Short-term Effects of Advertising: From Measurement to Improvement

In the digital realm, being an advertiser is no piece of cake. The more you venture into this rabbit hole, the more secrets you uncover. It’s like an ever-evolving landscape, one minute you know all the rules, and the next minute those rules lose their relevancy. Metrics, data, numbers, these terms have become the defining factor behind an ad’s success. However, the enormity associated with these terms often blindsides people to miss out on valuable insights. Until now the concept of a long and short-term impact on advertising techniques was quite unheard of. However, with the deployment of state-of-the-art Machine Learning and Artificial Intelligence concepts we have enhanced our ability to predict and adapt to newer situations.

Terms like Total ROI, short-term, and long-term effects have started appearing on the radar. But how does it all impact the digital landscape? What promises does it hold for the future? And does this newly found knowledge look good? All of these questions can be answered if we look beneath the surface and know the game.

The new players       

Meta in conjunction with Neilson, Nepa, and GfK came out with an interesting research article on the long-term and short-term impact on individuals or a community of individuals through advertising.

The study was conducted in Europe and it encompassed a number of media channels scattered across a plethora of devices. The key aspect here was the ingestion of data from various advertising campaigns across many domains.

The research

The advertising industry has been plagued with the issue of determining the long-term effect of advertising for some time now. Even to his day, it is highly debated among the advertising industry players. A few methods have been tried before but they happen to have one or the other shortcoming with respect to the variables involved. It is indeed a complex procedure to determine such an impact.

The research in light here encompassed 5 Meta studies with 3.5K+ campaigns on Instagram and Facebook. As mentioned above already, the key participants were from Europe, namely, Germany, The UK, Spain, Italy, Poland, etc.

Determination of the long-term effect was carried out by the deployment of Marketing Mix Modeling. Since there wasn’t a particular method to determine the required outcome, sophisticated concepts of advanced econometric modeling, coupled with the deployment of dynamic time series modeling were done. This move guaranteed the ability to separate the short-term and long-term effects and variables associated with them.

It was also here that the calculation of Total ROI came into the picture.

Short term ROI + Long term ROI = Total ROI

Other attempts made previously

The determination of long-term effects has been a tricky matter from the very beginning. Previous attempts (discounting the present study from Meta) haven’t been fruitful. While promising at first, there have been some miscalculations in their placement and subsequent deployment for the needed measurement. This is in part due to the inclusion of ill-considered variables and their placements.

1. Measuring the decay in advertising: In simpler words, the method here relied on measuring the impact of advertising on the sales figures stretched over a long period of time. The impressions gained by the advertisement were correlated with the KPIs observed over a brief period of time to measure an immediate or long-term impact on the ROI. While the approach looked good on paper, the reliance on different factors for representing the short-term and long-term effects called for the creation of a unique system. Something that was even more complex to comprehend, let alone create and use.
2. Looking at brand equity: Measuring the movement of choice considerations over time is what this approach hoped to achieve. And it did achieve the result to a certain extent. When mixed with the Marketing Mix Modeling technique and its components, the determination of long-term effects became a bit easier. You could now foresee the trends that lay ahead and plan accordingly. However, the lack of accurate data on brand equity became one major hurdle. Couple that with the fact that you could smoothen out the edges and see the long-term effect but the weekly movement in sales would still correlate. This approach was nearly there, but it couldn’t address the grass root concern.
3. Deploying methods without a floating base: The changes in how a system is measured, plotted against the key changes made to the system over time gives you a floating base. Every method that you need to deploy works either on a static base, or a floating base (dynamic base). Until now, the methods relied on the usage of a static base. The problem here was the fact that a static base works for short-term effects but it fails to give accurate results in the long run, as the variables change with the passage of time. The measurement of brand equity suffered from the lack of a floating base in order to calculate the long-term effect.

The findings

1. One of the crucial findings of the study pointed towards the fact that the Total ROI was 2.5x the size of the short-term ROI (Please note that the total ROI here includes the long-term effects as well) The study also found that close to 60% of the Total ROI was made by the long-term effects compared to a 40% share of the short-term effects.
2. The share of long-term effects in Total ROI in the Retail and Telecommunications, Technology and Durables, and CPG industries were 59%, 76%, and 42% respectively.
3. It was observed that Instagram and Facebook drove the most significant long-term ROI over the 5 studies conducted. A whopping 79% of the total sales effect under Technology and durables from Instagram and Facebook was generated from the long-term effect compared to 69% from the other long-term effect. The latter was generated through TV.
4. Now, coming to the channels, it was found that TV was the greatest contributor and driver of long-term ROI. A significant portion of this Total ROI was generated in the long term. Out-of-Home and Digital Out-of-Home Advertising also seem to contribute to the long-term effects of advertising.
5. Usage of the right strategies for long-term growth on platforms like Meta can result in high returns. Brand building is the key here, advertisers should focus on increasing the frequency, reaching a broader audience, and longer campaigns.

So, what really works for measuring the long-term effect accurately?

Meta in its study applied the concepts of Marketing Mix Modeling for the most part. While this approach has been tweaked with the addition of a few more sophisticated methodologies, there’s one more way with which the long-term effect can be measured. Unobserved Component Modeling can be a good way to deduce the pointers that paint the long-term effect.

A UCM (Unobserved Component Model) is a subset of time series in which it decomposes the same into components. The catch however is that some of the newly found components can be observed while others cannot. The observable components are the ones that as the name suggests can be directly observed and measured. The unobserved components on the other hand are those that can influence the time series but cannot be observed.

The modeling base shifts over time, a viable case that seemed to be missing earlier (refer to the “Deploying methods without a floating base” point). This shift allows you to:

1. Identify whether the long-term effect exists or not.
2. Quantify the long-term effect and boost the ROI.
3. Identify and deduce the factors determining the long-term effect.

And now that the dust has settled, we’ve witnessed complexities rising and falling throughout this piece. Trends and perhaps as understood so far, the base shall undergo changes with the passage of time, giving rise to new technologies and methodologies to decipher the labyrinth of digital ads.

Striking a similar note in thought leadership, here’s a sneak peek into our philosophy on the long-term effects:

“UCM models are well-suited for modeling complex patterns and trends in the data, but there are many other methods available that may be more appropriate depending on the context. We at Cubera are agile in adapting or doing a mix-match of models suitable for predicting the impact of advertisement and work with publishers and advertisers alike to maximize their individual interests.”  

-CTO Office, Cubera

And that’s about it!

Your data is valuable. Bring your data to the right place to receive the value share of the data you generate. We at Cubera make it possible.