Big Data: Avoiding Illusions to Accurately Predict Sales

Big Data: Avoiding Illusions to Accurately Predict Sales:from Business 2 Community by Greg Leibman
I recently watched an intriguing TV program:
“BBC Horizon: Is Seeing Believing” (You can download the episode from this link)
I was especially intrigued by an ‘illusion’ that highlighted the way in which the brain processes multiple inputs of information, underscoring how mixing data from your eyes and ears can produce a powerful and surprising illusion in sound.

• The researcher says ‘BAAA’ clearly mouthing a ‘B’ as he speaks.
• He then says the same word but clearly mouthing an ‘F’ as he speaks.
• When you watch his mouth say ‘FAAA’ that is what you hear. However, in every case, the sound is actually ‘BAAA’ (e.g. if you close your eyes), but the added visual of the ‘F’ sound leaves your brain no choice but to ‘hear’ FAAA.

The illusion above occurs because of tape delay, allowing the mix of visual and aural data recorded at different times, to appear to the brain as simultaneous input. The brain is fooled by the mismatch in time to ‘hear’ a different sound than is actually produced based on the visual input of the shape of the mouth.
With so much data available from each sense, our brain filters and combines information to create a model of our surroundings that we can store, comprehend, and use to take action. This experiment alters the timing of the sensory inputs and leads the brain to make incorrect assumptions about the interaction of the eyes and ears, and hence a false perception of reality.
This had me thinking about collecting and using big data to drive sales. Similar to the brain, to get an accurate perception of your prospects and clients, you need to have your big data sources properly linked in time.
It’s clearly important that combining data from various sources requires good linking keys to make sure all the data is about the same entity (e.g. customers and prospects). However, if you are not careful about making sure the data is linked in time, you run the risk of producing illusions in the data! Here are a few examples of data sources especially prone to this problem:

• D&B data. This data tends to be collected, processed, and distributed such that the ‘latest’ file time-stamped ‘March 2012’ actually represents a snapshot of data from at best ‘February 2012’ or even further back in time. You may want to lag your ‘connection’ to the rest of your data set by one month, or simply assume the D&B data can at best be anchored to a quarter, rather than a specific month.
• 5500 Form Data. These publicly available company filings (representing a high level description of retirement and welfare benefits data) are provided by the government. This data is typically at least 1 year old, due to the inherent process to collect, store, and wait the prescribed legal timeframe to release to the public. Be sure to pay attention to the Filing Dates to properly link to your other sources.
• Billing v. Transactional data. Even internal data can provide temporal mismatches. Often billing systems will capture the same transaction (e.g. buying a product) up to 1 full month after the actual event. This can produce havoc in a predictive model, if you perceive a billing event as happening a month after the actual sale.

Just like the brain takes the myriad of inputs from the senses and synthesizes a comprehensible perception of the world around us, historical data built on properly aligned temporal observations of company behaviors can expose key buying patterns and events that enable models to predict, for example, the propensity for a specific sales action.
While our brain does an amazing job providing a model for the world we perceive – it can be fooled. We need to be careful in our methods of connecting big data to account for when the observations occur, so we too are not fooled as we use big data to analyze, predict and inform sales teams to enable big sales!