Stitch Fix tests UI variations to isolate complementary human stylist inputs
Eric Colson · Combining Machine Learning With Expert Human Judgement // Eric Colson, Stitch Fix · Mar 18, 2016 · at 12:48
Eric Colson, Chief Algorithms Officer at Stitch Fix, outlines how the company isolates and measures human cognitive contributions within its algorithm-driven recommendation platform.
0:00 / 0:16exact quote · 16.4s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“We can reveal it by subtly varying the information we show our humans in this custom UI we built, and we can figure out what pieces of information help them with decisions and which pieces don't, and that way we can ensure that they only make positive contributions and that they're complementary to that of the machines.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Eric Colson
Stitch Fix employs 65 PhDs on its data science team
“We have a, one of the largest data science teams in Silicon Valley. You have 65 PhDs working on this.”
Netflix relies completely on implicit data rather than explicit user feedback
“At Netflix it was like pulling teeth to get the customer to tell you anything about themselves, right? They rely completely on implicit data.”
No single demographic segment makes up 1% of Stitch Fix's customer base
“We do not have a single segment that we can claim is, ah, representative of our customer. There's not even a one percent segment. Like our largest, we tried cutting them in every way. There's no rhyme or reason. It's everybody.”
Stitch Fix has no browse pages, product pages, or search boxes
“Nowhere on the site or on the app are you gonna find a browse page or a product page or a search box.”