Rejoignez 15,5 milliers d’autres abonnés

09 takeaways from this video:

00:00:36 Roy Price is a Senior Executive at Amazon Studios responsable for deciding which television shows will be produced by Amazon. IMDb is a popular website which rates thousands of television shows on a scale of 1-10. A 9-10 rated television is a successful, addictive, winning show such as Breaking Bad, Game of Thrones, and The Wire.

The problem with deciding which shows to air isn’t the risk of being ranked at the bottom: 1 or even 2. The risk is releasing one of the large majority of television programs that fall into the 7.4 average, mediocre ranking.

00:02:42 Amazon’s approach to identifying those 9s and 10s is to hold a competition whereby thousands of ideas are evaluated and sorted until the final eight ideas remain. A pilot eposide for each idea is then filmed and made available for the world to watch for free while Amazon monitors the very-detailed analytics (plays, pauses, scene skips, rewatches, shares, etc.) of those pilot eposides to identify which shows are the top performing and should be invested in.

Based on the above compilation of millions of data points, on April 19, 2013, Amazon released the show Alpha House, a political satire.

00:04:15 Meanwhile, Ted Sarandos of Netflix uses a slightly different data analysis approach to finding blockbuster television shows: all of the data they already have from their users’ past behavior: series they’ve liked, skipped, the actors/actresses who starred in them, etc. All this analysis lead Netflix to launch House of Cards, a television series similar to Alpha House, but about a single senator, which went on to become a 9.1 rated series.

Two very successful and intelligent data-savvy, data-driven companies collecting millions of data points to predict and then produce top rated television shows, and two completely different results. What makes the difference?

00:06:01 Multi-health systems is a software company specialising in data-analysis software which determines whether a prison parolee qualifies for parole or not. Basically, it’s the television series datamining software for humans.

Similar to the predictability of the next hit television show or the next hit music single, results may vary.

00:07:14 In Nature Volume 457, published 19 February 2009 Google made the bold claim that, using data analysis, they were capable of predicting influenza outbreaks based on it’s complex algorithms of Google searches. This article was highly regarded as a pinacle of scientific success…

00:07:43 …until a few years later it failed – for reasons unknown – and lead to a retraction of that originally published paper.

Despite its failures and unknown variables, data-driven decision making is rapidly moving into the modern workplace, law enforcement, and medicine industries.

00:08:35 For me (Sebastian Wernicke), the difference between successful and disastrous decision-making with data and seems to lie in our ability to:

  1. Deconstruct the problem into its bits and pieces so you can deeply analyze it
  2. Put all of those bits and pieces back together form an intelligent conclusion

[EDITOR’S NOTE: Recall in the TED talk titled the 3 phases of love, and why we love and cheat by Helen Fisher that “women tend to have better social skills, be able to collect and work with larger amounts of data, and be ‘web-thinkers’ with the ability to handle complex problems, whereas the average man removes ‘extraneous’ information to focus on the core issue solve problems using a step-by-step approach.”]

00:09:04 No matter how powerful data analysis is, it is (to date and prior to singularity) only useful for the first part: deconstructing the problem into its bits and pieces and deeply analyzing them.

Step 2: Putting all those pieces back together again, requires (again, to date and prior to singularity) the human brain, particularly an expert’s brain. So what made Netflix so successful was their using the right concoction of data-driving algorithms and the human expert brain.

Pure data-driven decision making lacks the most important element of success: the innate expertise of the human brain to conduct pattern-recognition and to reconstruct those data-driven pieces.

00:10:25 Data-driven decisions will always lead to safe, mediocre decisions because you can always fall back on the plethora of data that lead to your decision.

[EDITOR’S NOTE: Recall in Rory Sutherland’s talk the next revolution will be psyhological not technical that “

One thing all ‘the next big things’ have in common is that they come from a place you wouldn’t expect, and when people do try and predict the future they basically take the most visible form of progress they’ve seen in their own lifetime and extrapolate upon it.”]

Data should be used as a tool, not as a decision-maker. You should become an expert in what you’re doing, and take risks.

[EDITOR’S NOTE: For more on becoming an expert versus being a generalist, watch the talks Building a Generalist/Specialist Business by Dror Benshetrit and 10 proclamations to win new clients without pitching by Blair Enns.]


2 réponses à “172. How Netflix and Amazon Pleasure You through Data-Driven Algorithms”

  1. […] How Netflix And Amazon Pleasure You Through Data-Driven Algorithms […]

  2. […] [EDITOR’S NOTE: For more information on how Netflix and Amazon test movie and series releases, watch the TED talk How Netflix and Amazon Pleasure You through Data-Driven Algorithms.] […]