A couple of weeks ago, the Oklahoma City Thunder roared to a shocking three games to one lead in their best-of-seven NBA Playoff series against the Golden State Warriors. While a few loyalists held out hope that their team could come back and sweep the final three games, the rest of the sports world wrote the Warriors off. Media pundits cited the highly unlikely statistical probability that the Warriors would be able to come back.
A few of the more aggressive sports analysts even got specific. “About 4%” was the likelihood that was repeated convincingly by many sources. They referenced actual data – and seemingly lots of it. Even non-mathematically inclined people can think back to junior high school math and recall that the probability of flipping a coin three times and coming up with three consecutive heads (or three consecutive tails) is 1/8 or 12.5%. If the teams were roughly evenly matched, then we might expect somewhere closer to the neighborhood of 12.5% likelihood that this series should result in the victory for a team that is down 3-1. To the casual observer, a success projection so far removed from our own snap judgment might imply that there is a deep mathematical analysis.
But when one cuts through all the noise surrounding these pundits’ arguments in favor of Golden State’s pending demise, we quickly see that they were all attaching to one only statistic. In the last 232 NBA playoff series that started 3 – 1, only 9 teams who were down came back to take the series. The 4% was simply the quotient of winners (9) divided by all previous 3 – 1 series (232).
So what is so wrong about this approach? After all, none of these pundits said that a Golden State comeback was an impossibility. Golden State was clearly in a suboptimal situation. But was it really as dire as the prognosticators suggested? Probably not.
The approach taken by these armchair data analysts was incomplete because they were relying exclusively on one statistic in a vacuum. A more complete approach would have included a detailed analysis of the history of those earlier 232 series. Were there identifiable characteristics of those nine winning teams that mirrored Golden State? Even if we assume that the 9-of-232 statistic was an important factor, was there anything about Golden State’s team or each of the next three games that would make it a potential outlier? Was there some anomalous circumstance that guided Golden State to a 3 – 1 deficit, but might not be present in the next three games?
This flawed data approach is not just reserved for professional sports media. As companies are retooling to become more data driven, we see these same mistakes play out every day in sales organizations around the world in which sales reps and their managers attach to one clearly understood metric or another and ignore all other data points that could suggest a very different outcome. They then make seemingly reasonable conclusions on all manner of sales challenges based on those singular data points:
Purchase History: “This prospect purchased at his previous company and loves us! This deal will definitely close!”
Forecast category: “My team always closes exactly 90% of our commit forecast.”
Perceived rep tendencies: “Diane is a sandbagger, but she always comes through in the end, so I will just put her at quota.”
Lead Score or Opportunity Source: “We have one week left in the quarter. Please stop wasting your time on opportunities with a low lead score.”
Deal Age: “Karen is typically spot-on with her forecast, but this deal has been in the pipeline for too long, so I don’t believe it.”
It is understandable why sales leaders are left to judgment calls based on these isolated variables. Every organization has to contend with finite sales resources and a constantly ticking clock. Sales managers need to focus their reps on deals that appear likely to win. Likewise, they have to submit forecasts that at least acknowledge observed patterns of rep behaviors.
However, in the absence of an advanced analytical examination of all impactful signals of deals winning and losing – and the interconnected nature of those signals – negative consequences are to be expected.
In the short term, we have to assume that certain deals that should win will get less attention than they merit. Other deals that face the longest of odds will be subject to overinvestment of time. Over the longer term, a lack of awareness of the true drivers of wins and losses will make us less equipped to architect our sales organizations for success.
Whether we are discussing a sales rep who has a well-established pattern of overpromising, or a basketball series that has well-established patterns of winning and losing, there are almost always other data points that will influence the true probability. Understanding that data will minimize the surprises that leadership teams (and Thunder fans) hate.