亚马逊增长30倍后工程生产力出现了拐点.pdf

1、Inflection Points in Engineering Productivity as Amazon Grew 30 xCarlos ArguellesSenior Principal EngineerAmazon World Wide StoresWho am I?Microsoft(1997-2009)Amazon(2009-2020)Google(2020-2024)Amazon(again!)Who am I?Microsoft(1997-2009)Amazon(2009-2020)Google(2020-2024)Amazon(again!)Engineers:3k 80k

2、 30 xEngineers:3k 80k 30 xStock:$3$210 60 xEngineers:3k 80k 30 xStock:$3$210 60 xMe:SDE-II Senior Principal Senior Principalkey lessonslearnedTheres inflection pointsMost obvious is number of engineersTheres inflection pointsI waste 10 seconds doing somethinghappens 5 times per weekit affects 100,00

3、0 engineersToil aggregates at scaleI waste 10 seconds doing somethinghappens 5 times per weekit affects 100,000 engineersaverage engineer costs$400kToil aggregates at scaleI waste 10 seconds doing somethinghappens 5 times per weekit affects 100,000 engineersaverage engineer costs$400kToil aggregates

4、 at scale$14MM/yr10 seconds$14MM/yrWhat is my ROI?Investment:4 months dev work2 months operations longer-term$200k Return:$14MM 70 xToil aggregates at scaleit affects 100,000 engineers$14MM/yrit affects 3,000 engineers$400k/yrOver the course of 5 yearsInvestment:$200kReturn:No growth=$70MM($14M/yr f

5、or 5 years)Hyper-grow=$14M yr1+$28M yr2+$56M yr3Toil aggregates at scaleand compounds over timeTheres inflection pointsSometimes the inflection point is a crisisTheres inflection pointsSometimes the inflection point is maturityTheres inflection pointsSome duplication makes senseIn periods of hyper-g

6、rowth each organization will optimize for its needs Some duplication makes senseIn periods of hyper-growth each organization will optimize for its needsWhen facing ambiguity multiple exploratory efforts make senseSome duplication makes senseBut theres a time and a placefor converging tooSome duplica