“Prediction is very difficult – especially if it is about the future.” Niels Bohr, Danish physicist

Today Adobe Digital Index released its prediction for daily online sales this holiday season. Everybody’s got a prediction.  Why is Adobe’s any better?  3 reasons:

  1. Purely data driven: Based on statistical model, not informed conjecture
  2. Relevant: Marketers can dig into the specific segment that matters to them
  3. Actionable: Daily predictions explain key sales drivers (visits, conversion, average price paid)

Purely data driven

We’ve analyzed more than 150 billion visits to 500+ retail websites and found shopping patterns that are startlingly consistent and surprisingly robust. The regression model used to predict total US online sales looks only at normalized online sales over the last six years and is able to predict this year’s sales with an R2 value higher than .95. Roughly translated, 95% of the variation in day-to-day online sales can be explained by factors in the model. It’s truly impressive that the model can have such a high degree of accuracy without relying on imprecise external data and economic indicators. People ask whether hurricane Sandy or the recent election will dramatically affect online shopping. The only response I can offer is that there was an election in 2008, an economic recession in 2009, stagnant high unemployment in 2010, a debt ceiling crisis in 2011 and the data shows that through it all online shopping patterns have been persistent. If some event in 2012 were to change those patterns it would truly be “statistically significant.”


We’ve created an interactive website that will allow marketers to slice and dice the data in 168 different ways and dig out the most relevant insights. Marketers always have to wonder whether their actions are having the desired effects. Did our Cyber Monday deals cause people to buy or would they have anyway? Did we start our holiday promotions too soon or are people just not ready to shop yet. Historical data only tells the marketer what happened, not what would or could have happened. By analyzing the industry as a whole, Adobe is able cut out the noise to reveal the underlying consumer behavior to help marketers know what to expect. For example, while typical retailers should expect that conversion rates will jump to 60% above normal on Cyber Monday, apparel retailers can expect a much higher jump up to 215% of normal just because shoppers will be in the mood to buy clothes. Retailers who see different results than their segment will know it’s because of their atypical actions.


Big data also gives us the ability to provide a daily prediction.  Why is a daily prediction important?  Because it enables marketers to overlay their own patterns against the average and discover how their promotional strategies are creating separation between their website and the rest of the category. Besides knowing that sales will be up again marketers need to know what will drive sales day-to-day. For example, this Friday online sales are going to jump to 200% of normal for the typical retailer. Why? Not because people will be shopping that much more. Website visits will be up only 10% above today. Online sales will spike because conversion rates will jump up to 40% between today and four days from now. By understanding these shopping patterns retailers can tailor their promotions to the mood and purchase propensity of their customers.

You’d never finish reading this blog post if I tried to tell you the rest this data says about Black Friday. And that’s just one day. So go ahead and find the answers yourself here. Let me know what you learn and what you’re interested in. And follow me @tyrwhite if you’re dying to hear more.