by Matt Manning
Every project manager has conducted a “post-mortem” after their big product has launched or their information service redesign has rolled out to their customers. Nits are picked, fingers may be pointed and, on occasion, heads can roll. In these days of pervasive predictive analytics, though, can we realistically prevent major process errors from occurring so post-mortems are not needed in the first place?
The idea is a simple one: Assign a small team to break what you are building before someone else finds the vulnerability. It’s similar to what ‘white-hat’ hackers do. The team anticipates and mitigates negative process outcomes in-house before enterprising customers discover the problems themselves.
The types of problems discoverable via a pre-mortem include:
- Back-door ways to circumvent access permissions so the 1-week free trial subscriber can’t download the entire database on day one.
- Search engine optimization fails that hide the fact that high-quality content is available on your service.
- Advanced searches that yield no results and don’t offer a way out of the dead-end.
Once discovered, problems like these can be tested and resolved before they are made public. For instance, adding a “Do you mean X?” mechanism to refine a search, giving the users the chance to make a custom data request, or suggesting coverage of a particular subject area can steer users to useful content and improve customer satisfaction even when a search result is unsuccessful.
So when it comes to the performance of your new or improved information service, maybe conducting your own ‘pre-mortem’ before you launch is a better alternative than the traditional post-mortem. The job you save may be your own!
posted by Shyamali Ghosh on February 24, 2019
by Matt Manning
The term “crowdsourcing” is not aging well. Maybe it was a bit too folksy to begin with or, more probably, its slow evolution caused it to seem passé before it ever even hit its stride.
Whatever the reason, the meaning of “the crowd” has morphed into two distinct things at this point:
- an anonymous online suggestion box, donation site, or survey; or
- a ton of individuals working on small data tasks designed to improve the performance of computer vision- and location-based applications (i.e., artificial intelligence training).
The former “crowd” gets a lot of press and is a legitimately useful tool for harnessing the power of citizen scientists and other civic-minded people. The latter massive pool of freelance labor (that can start and stop virtually on demand) is where the smart money is still being spent to drive innovation. How and why, you ask?
Well, the “why” is this: the value proposition of on-demand labor is still compelling because of the staggering cost-savings it brings major industries, including
- Driverless vehicles (taxis, buses, trucks)
- Retail without check-out clerks
- Remote monitoring (vs. on-site security)
These three labor-intensive areas become much more profitable when the formal labor force required for them declines by 90%.
The “how” is essentially that once enough human data is provided (via annotated audio, image, and video files) then the data can be programmatically analyzed for patterns. These patterns can then inform the algorithm’s code (i.e., “a shape that looks like this is X% likely to be Y”).
So while we bid a fond farewell to “the crowd” as a business-to-business term, we should remember that there are now literally hundreds of thousands of human beings training next-gen tools combining sensors and software that will define the reality of our near-term future. Call these people what you will, but never forget that they are the ones fueling a fast-paced future where labor is portable and everything is cheaper and more efficient.
posted by Shyamali Ghosh on October 22, 2018
