COVID Recovery Analytics

Resilience; n.; defn: the capacity to recover quickly from difficulties (Oxford Dictionary)

About 4 months ago now (though it seems much, much longer) our worlds were turned upside down by first the pandemic and then the lockdown used to deal with it.  Only now, here in Canada, are we seeing what appears to be the light at end of the tunnel (or is that an oncoming train??).  But many regions around the world are not there yet, and may not be there for many months.  And even as our case volume falls here, there is the constant awareness that a second wave may be in our future.  It does appear that until an effective cure or vaccine is found, we must find a way to live with COVID-19.

The impact on business has been dramatically bad; I won’t repeat many of the stats I am sure you have heard all too often. Many brands have put advertising on hold; hunkered down and waiting for the storm to pass.  In the meantime, though, these companies are bleeding business and piling up losses.  Some will not make it to the other side.  Some will have to open up and learn to function in this COVID economy, because there is no alternative.

(That said, some companies are in sectors that have benefited from these unusual times.  Online entertainment, ecommerce, products related to cleanliness and personal protection; these have all seen healthy sales volumes, if not increases vs the Before Times.)

How will we navigate this new normal?  How can we safely guide sales and brand-building marketing, while controlling risk? 

In our corner of the marketing world…advanced analytics…the answer lies with technology that we have been developing and perfecting for many years before COVID-19 showed up. 

Let’s start with predictive models.  This technology has been used by most marketers, although in my experience, some use it in a very tactical way, to, for example, help them build lists of customers for a campaign.  Predictive modeling is capable of much more.  It is possible to build models that predict near term (up to 6 months) outcomes with a high degree of accuracy and granularity (deciles, markets, segments, channels, etc)….certainly high enough accuracy that we can make more confident marketing decisions with them.  Such models are generally holistic in data inputs…covering a range of factors that drive results, both those under a marketer’s control and those that are not.

We have built models that incorporate the nearly 4 months of datapoints we have seen since lockdown.  We were pleased to see those models are as accurate…or even more accurate…than similar models on the same business from the Before Times.  That is very encouraging. We are talking accuracy rates…the difference between predicted and actual sales, by day or week, in the range of 93%.  That is more than enough to enable us to design high performance, low risk marketing campaigns.

But what if the virus flares up again?  To answer this, we have incorporated COVID case volume data by day into our models.  The data is allowing us to see strong differences between regions across Canada (an effect we would expect to see in the US, or other countries).  It makes sense in looking at the case volumetrics why some areas are still struggling with re-opening, while others have a much healthier outlook.  More importantly, as we move through the summer and into the fall/winter, when the concern is for the virus to make a comeback, we will have the data coming in, daily, to act as an early warning signal for our clients.

The second technology that enables a safe restart is optimization.  At Custometrics, we build custom optimization algorithms that take full advantage of the data and predictive models, to manage complex allocation problems…how much to spend by channel, market, time, tactic, etc.  Further, these optimization models also produce accurate forecasts for different planned scenarios.  This forward looking approach, allowing us to test various campaign designs for both effect and risk, gives clients more confidence to spend.

Finally, a critical component to recovery analytics is testing.  Testing expands our knowledge of marketing effects and builds the accuracy of our predictive models.  It also gives feedback to our decision makers that has immediate effect on the next campaign cycle. 

Bottom line; the combination of communication-smart, COVID-aware predictive models, custom optimization and smart testing will help us cope with the most challenging market we have seen in our lifetimes.  There is much to be cautious about, but also developments here that we can take courage from.

DB

Istock photo credit:  frankpeters