In the wake of COVID’s impact, the questions and uncertainty that surround measurement and analytics grows larger by the day.
With such an unprecedented and unusual event, the implications and effects remain largely unknown. When to model again, whether past models are relevant, and strength of impact are all top-of-mind questions for any data driven, analytics focused company. Outlined below are the top measurement questions that we at Marketscience have received since the onset of COVID-19. We’ve also shared our company POV on the best routes forward to keep your business focused on the future and analytically engaged to make the best, data driven business decisions possible.
How big is the impact of COVID-19?
This question was frequently asked after the immediate onset of COVID and the resulting furloughs and lockdowns. The big unknown was whether this would be a temporary pause on life that would recover as soon as restrictions lifted, or something that rooted itself more permanently into our ethos. While optimism loomed at the beginning, we have increasing evidence that what once looked to be temporary may soon be enduring, given the apparent inability to carry on with “business as usual” without a vaccine.
Since industries were, and still are, being hit disproportionately, COVID’s economic impact will vary accordingly. Some business may come back stronger, and some may not come back at all. But those who continue to weather this storm and innovate their offerings to mold with the times will be better placed to meet the demand of consumers as the economy makes a return to either the old or the new normal.
These drastic differences among industries will recover in an alphabet of shapes. Most economists are familiar with traditional V, L, and U-shaped recoveries, but COVID has created entirely new alphabetic recoveries like W, N, and K representing the birth of new products and the uncertain nature of lockdown and city re-openings. These letters may give some comfort that there are typical paths to recovery and thus an “end” in sight but what do they mean for modeling?
Can I use my existing model results to forecast budgets?
Yearly planning never seems too far away, and 2021 has had a major wrench thrown into its path as COVID continues to warp large chunks of 2020 data. With the majority of MMM results either not including COVID’s impact at all, or more likely, only a portion of its impact, how does one move forward with utilizing results for strategic action later down the line? Are my previous results usable given the current economic environment?
The good news is that if your organization has utilized a Marketing Mix Model previously, then you have a logical starting point from which you can tweak budgets accordingly. Although business performance is likely to have been impacted, it’s likely that relative channel effectiveness has not. While we wish there existed a magic model that will predict our COVID-influenced future, the next best thing is to try to forecast how those previous model results might now change and then use these revised results to strategically inform your future marketing plans.
In order to do this correctly, one needs to make reasonable assumptions about how COVID has impacted your businesses marketing opportunities. For example, consider the recent trends that have shaped the past few months: people are spending more time and money online (social media and display ad response curves should increase), people are not shopping in stores or malls at the same rates as before (promotions and distribution curves will lower), people are inside and watching more TV (TV curves increase), people are not traveling internationally, but are perhaps taking more domestic trips (OOH and Radio will become opportunities), and the list goes on. Of course, these trends will affect businesses and industries differently, but in combining past MMM results with rational media assumptions, your organization will be in a far superior place to meet the demands of today’s current environment.
When should I start modeling again?
A lot of companies have abstained from conducting an MMM during COVID because sales were too low or data too skewed to be deemed useful. While this is a common refrain, we urge companies to reflect that the reason models are useful in the first place is that they provide a rational starting point that you can strategically plan, deviate, and experiment with when things go askew. Having models in place isn’t about forecasting down to the 5th decimal point (a verified pointless and hopeless task), but about having a tool in place where you can change parameters, experiment with different tactics, and take action in a strategic, measurable way.
Utilizing a technique like Dynamic Linear Modeling (DLM) is one way to ensure that the fluctuations and changes brought on by COVID are sufficiently controlled for in the base of your model. At Marketscience, we promote a DLM integration with Hierarchical Bayesian approach which provides the best measurement and control environment to support your test and learn programs. As companies test their way into new learnings and better investments – based on the assumptions your organization made above in question number two – DLM and Hierarchical Bayes informs how marketing channel effectiveness has changed relative to historical numbers and predictions recently made.
Without analytics, creating a successful experimental test & learn culture would be significantly less fruitful. If businesses view “black swan” events such as COVID-19 from a longer-term lens, and work to ongoingly collect data and experiment with new strategies, we expect such organizations to rise from this much stronger than those who are patiently waiting it out. After all, an unstable world is exactly where analytics come into place.