02. arjen de ruiter, bol.com

33
The value of data Performance & scalability challenges @bol.com jenderuiter des with input from Niels Basjes & Edith Kanters

Upload: big-fellows

Post on 09-May-2015

399 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: 02. Arjen de Ruiter, Bol.com

The value of dataPerformance & scalability

challenges @bol.com

@arjenderuiterSlides with input from Niels Basjes & Edith Kanters

Page 2: 02. Arjen de Ruiter, Bol.com

Topics• Value of data• Scalability &

performance• Tips

Page 3: 02. Arjen de Ruiter, Bol.com

Value: customers find what they are

looking for with search rank

Page 4: 02. Arjen de Ruiter, Bol.com

Value: inspire customers with

product recommendations

Page 5: 02. Arjen de Ruiter, Bol.com

Value: a personal shop full of

inspiration for every customer

Page 6: 02. Arjen de Ruiter, Bol.com

Value: provide us with explicit

feedback on your personal shop

Page 7: 02. Arjen de Ruiter, Bol.com

7

Everything data

CUSTOMER

Right frequency

Right Event

Right content

Right place

Right promotion

Right timing

Page 8: 02. Arjen de Ruiter, Bol.com
Page 9: 02. Arjen de Ruiter, Bol.com
Page 10: 02. Arjen de Ruiter, Bol.com

Sprints

Stor

ies

V

alue

ValueStory points/scopeRELEASE

As stories burn down, value increases

Page 11: 02. Arjen de Ruiter, Bol.com

15 million products: not Big Data

Page 12: 02. Arjen de Ruiter, Bol.com

15 million clicks a day: not Big Data

Page 13: 02. Arjen de Ruiter, Bol.com

>3.5TB + >20TB> 10.000.000.000 product combinations: getting bigger!

Page 14: 02. Arjen de Ruiter, Bol.com

It does not fit anymore

Page 15: 02. Arjen de Ruiter, Bol.com

It takes too long

Page 16: 02. Arjen de Ruiter, Bol.com

16

Scalable processing: it al starts

small

Page 17: 02. Arjen de Ruiter, Bol.com

17

Page 18: 02. Arjen de Ruiter, Bol.com

18

Page 19: 02. Arjen de Ruiter, Bol.com

19

Page 20: 02. Arjen de Ruiter, Bol.com

20

Big enough and cost effective?

Page 21: 02. Arjen de Ruiter, Bol.com

21

<schaalbaarheid>

Page 22: 02. Arjen de Ruiter, Bol.com
Page 23: 02. Arjen de Ruiter, Bol.com

Economic size

Different models

Page 24: 02. Arjen de Ruiter, Bol.com
Page 25: 02. Arjen de Ruiter, Bol.com
Page 26: 02. Arjen de Ruiter, Bol.com

26

Linear horizontal scalability:

start small, end big

Page 27: 02. Arjen de Ruiter, Bol.com
Page 28: 02. Arjen de Ruiter, Bol.com

MTBHIAGSD

Value

MTB

HIA

GSD

@andrianco, Netflix

Mean Time Between Having

An Idea And Getting Stuff Done

So deliver fast!

Page 29: 02. Arjen de Ruiter, Bol.com

Launch early & iterate

One of the company rules of thumb at Google is: if you are not embarrassed by your first launch, you have not launched early enough

Page 30: 02. Arjen de Ruiter, Bol.com

Trim the tail, don’t spend time on low value features

Value

Tim

e

Epic 3

Epic 2

Epic 1

Tail … does not add too much additional value

Page 31: 02. Arjen de Ruiter, Bol.com

Measure & act. Fail fast, fail cheap.

A/B testing to find the best variant of a solution

Page 32: 02. Arjen de Ruiter, Bol.com

Our main Big Data conclusion

Page 33: 02. Arjen de Ruiter, Bol.com