To Speed Through Failure

ToSpeedThroughFailure
&OtherRelatedConcepts
Marcus Frödin
@marcusf #sparkconf

What we will cover
‣ OrgPrimer
‣ TheAppsPlatformFailure
‣ Whatinnovationlookslike
‣ TheShiftToMobile
‣ Rituals&Culture
‣ AfterTheAppsPlatform

“Spotify is the only organisation I have ever seen that
treats simple problems as complex, instead ofthe other
way around.”
Simon Marcus,VP Strategy Operations.At Spotify.

With that a disclaimer
We think a lot about organisations, and that shows. If it sounds
complicated for the sake of complicated, it might be.
This stuff works for us, today. I make no claims we’ll use it tomorrow,
or that it’ll work for you.
Finally, this is my very personal take on our org. Many Spotifiers will
likely disagree.

Squad Squad Squad
Tribe
Squad Squad
Tribe
Squad

Tribe Tribe
Alliance
Squad Squad Squad SquadSquadSquad

Tribe Tribe
Alliance
Tribe Trib
Alliance
Squad Squad SquaSquadSquad

Tribe Tribe
Alliance
Tribe Trib
Alliance
Squad Squad SquaSquadSquad
TPD

We want to be
Optimized for growth, retention and experimentation.
Able to hold local leaders very much accountable.
Really good at getting it wrong quickly.
Embarrassingly parallel.
Faster, ratherthan more quality-minded.

Embarrassingly parallell and  
good at failing

Distance (Org/Time/Space)
Cost

Distance (Org/Time/Space)
Cost
Need

Squads are collocated, sat closest to
codependent missions and have all necessary
components to move forward.

Information Decisions
Obvious massive anti pattern

Hypothesis
Intent
Learning
Keytrait:
Asynchronous
Alignment
Autonomy
Feedback

Leaders spend time setting intent and helping
propagate knowledge.

SQUAD
TRIBETRIBETRIBE
TRIBETRIBETRIBE
TRIBETRIBETRIBE
SQUAD
SQUADSQUAD
SQUAD SQUAD
SQUADSQUAD

We’ll get to it.
For now it’s magic.

TAKE-AWAY #1
Autonomy & Alignment Interact

A model for talking about failure

Plan Action Outcome
Knowledge Gap
Alignment Gap Effects Gap

Intent Hypothesis Feedback
Knowledge Gap

Most people focus mostly on this thing
Knowledge Gap

Platforms always win. App stores are disproportionately
shifting innovation to Google and Apple. If you give
developers API’s, they will come.
Spotify needs to become the platform for music. We also
need to become faster at shipping our internal software.
By mimicking an Operating System and using web tech to
power apps, we can hit two flies with one rock.
We need to rewrite our desktop client to support web
apps, and use this to power an app store.
People will use apps to listen to music and discover new
stuff. Developers will innovate and we’ll benefit.
STRATEGY
PLAN
OUTCOME
ACTION

App
Platform
2011
DATES ARE MOSTLY ILLUSTRATIVE. VERY LIKELYTHEYARE SOMEWHAT OFF.

App
Platform
2011
Started building team

App
Platform
2011
Started building team Shipped the platform

Many apps on launch
App
Platform
2011

Many apps on launch
App
Platform
2011 2012

Many apps on launch
App
Platform
2011 2012
100’s of apps shipped

Many apps on launch
App
Platform
2011 2012
Specialized team to 
help apps launch 
quicker

Many apps on launch
App
Platform
2011 2012
help apps launch 
quicker
Still 100’s of apps shipped

Many apps on launch
2013
App
Platform
2011 2012
help apps launch 
quicker

Many apps on launch
2013
App
Platform
2011 2012
help apps launch 
quicker
Sigh. Still 100’s of apps shipped

Many apps on launch
2013
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard

Many apps on launch
2013
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Started consider options

Many apps on launch
2013
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.

Many apps on launch
2013
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.
Start retraining team

Many apps on launch
2013 2014
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.

Many apps on launch
2013 2014
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.
We tell developers

Many apps on launch
2013 2014
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.
We tell developers
Implement it all

Many apps on launch
2013 2014
App
Platform
2011 2012
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.
We tell developers
All of team retrained
Implement it all

Many apps on launch
2013 2014
App
Platform
2011 2012 2015
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.
We tell developers
Implement it all

Many apps on launch
2013 2014
App
Platform
2011 2012 2015
help apps launch 
quicker
Mobile turns out
to be very hard
Decisions to
shut it down.
We tell developers Apps Platform RIP
Implement it all

Plan Action Outcome
We shipped the product on time.
Hundreds of apps launched on the
platform.
However, over time, interest started to
wane. I believe the initial inabilityto
monetise on the platform, plus the
comparatively few users (~30m MAU at
peak) lead to interest petering out.
As the shift to mobile was in full swing,
reproducing the platform on mobile
turned out to be very hard.

Plan Action Outcome
Knowledge Gap
Special purpose app stores have never succeeded at
the scale we envisioned (outside of games).
Running arbitrary apps inside an app is much harder on
mobile than desktop.
Hire for generalists, even with in specialised roles.
The API’s and incentives put in place did
not lead to long term engagement from
developers with the platform.

What to do with those learnings?

What does
innovation look like?

Innovation
1. The action or process of innovating

Innovation
1. The action or process of innovating
2. A new method, idea, product, etc.

Innovation
It’s also what most of us here do, in pursuit of making our
users lives better and building better product.

Here’s the real maxima
You’re here

“App stores disproportionately shift innovation to Google & Apple”
- Me, just a few slides ago

So can we learn anything about innovation in the app store?

Assume: Successful innovation will eventually generate revenue.

On average, more money means you’ve innovated better.

So how does the app store distribute revenue, financially?

http://dazeend.org/2015/01/the-shape-of-the-app-store/

That puts 99.93% of apps in the “long tail” of the App Store.
The “head” of the App Store, those 870 top grossing apps that
make up 0.07% of the App Store population, collect over 40%
of the App Store revenue.
http://dazeend.org/2015/01/the-shape-of-the-app-store/

You’re not creating something from nothing, you have a
constraining context. You have strategic intent and you can
propagate it more effectively.

Next Thing
App Platform
Next Next Thing

Plan Action Outcome
= Strategic Intent

Plan Action Outcome
Propagate intent
= Strategic Intent

Plan Action Outcome
Propagate intent
= Strategic Intent
Answering this is figuring out “Where do we compete?”

Plan Action Outcome
Propagate intent
= Strategic Intent Platforms drives down time,
experimentation culture drives down scope
Minimize inventory
Measure outcomes
Answering this is figuring out “Where do we compete?”

Measure like a mad man. Rinse and repeat.

http://www.slideshare.net/MrChrisJohnson/from-idea-to-execution-spotifys-discover-weekly

When Spotify first debuted its Discover Weekly personalized
playlist in July, it felt like a hit. Well, now there are some
numbers to back up the hype: In its first five months, Discover
Weekly resulted in 1.7 billion streams.
http://www.fastcompany.com/3054176/elasticity/why-spotifys-discover-weekly-playlists-are-such-a-hit

TAKE-AWAY #2
Iterate with Purpose To Beat Darwin

The Shift to Mobile (2011 onwards)

from: CTO
to: technology
received: Thu, 28 Feb 2013
subject: How do we become a mobile company?
This is the million dollar question... We all know that we must become a mobile company to win, but how do we make it
happen? A group of 25 people have spent the last 2 months on this topic. The group includes engineers, PO's & designers from
all levels of the company and we also reached out to several external companies for advice.
There are still many unanswered questions, we will continue to attack questions one after another. But the fundamentals are
now widely agreed.
Please read the attached to see what we came up with. This will be presented verbally at an all hands soon, but if you can't wait
until then please read this doc. You can direct any questions to your senior manager, myself or any mobile developer.

Mary Meeker, 2010 + 2011
http://www.kpcb.com/blog/2011-internet-trends

http://www.kpcb.com/blog/2011-internet-trends

Mobile phones overtook desktops. Really really quickly.

Shift to
Mobile

Shift to
Mobile
2010
/
2011

Shift to
Mobile
2010
/
2011
Meeker Decks

Shift to
Mobile
2010
/
2011
2012
Meeker Decks

Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps Platform Strategy

Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Hiring frenzy forweb

Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
iOS and Android,  
very seldomly and 
from two teams

Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Work on mobile strategy

Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Demo latest flagship on web

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Mobile strategy launched 
“One app on both platforms”

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
First Interim  
Release Manager
> 1 team contributes

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
First Interim  
Release Manager
Start hiring craze

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
First Interim  
Release Manager
Start trying to ship everythree
weeks
Start hiring craze

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
First Interim  
Release Manager
weeks
Start hiring craze
Start trying to 
retrain devs

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
First Interim  
Release Manager
weeks
Start hiring craze
iOS 7 announced: Abandon web on mobile
Start trying to 
retrain devs

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
chapters contributing.
Start trying to 
retrain devs

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
blocking most teams.
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Start trying to 
retrain devs

2013
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
Ship our mobile first product
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Start trying to 
retrain devs

2013 2014
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Start trying to 
retrain devs

Dedicated release / arch teams
2013 2014
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Start trying to 
retrain devs

2013 2014
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Start trying to 
retrain devs
Hiring starts working

2013 2014
Shift to
Mobile
2010
/
2011
2012
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Mobile chapters most places, 
guilds are effective.
Start trying to 
retrain devs

2013 2014
Shift to
Mobile
2010
/
2011
2012 2015
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
Start trying to 
retrain devs

2013 2014
Shift to
Mobile
2010
/
2011
2012 2015
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
chapters contributing. ~April 2014
Several thousands
Single hundreds
JAN 2011 JUN 2016
Start trying to 
retrain devs

Monthly active developers
2013 2014
Shift to
Mobile
2010
/
2011
2012 2015
Meeker Decks
Apps shipping on 
from two teams
Lack of mobile is
First Interim  
Release Manager
weeks
Start hiring craze
Multiple squads and
chapters contributing. ~April 2014
Several thousands
Single hundreds
JAN 2011 JUN 2016
Start trying to 
retrain devs

Repo
Android
App
Squad
Users
Squad Squad Squad Squad
SingleTeam Writes & Ships WholeApp
1-10 developers in one app
50-500 commits per month
Other squads provide backends for UI team to work with
Nice segregated backend architecture to playwith

Repo
Android
App
Squad
Users
Squad
Squad
Squad
Squad
AllTeams ContributeToThe OneApp
OneTeam Releases
10-100 developers in one app
1000’s-10 000’s commits / month
Squad
Squad

Knowing this would be painful, we should have started way earlier.

Squad
Users
Squad
Squad
Squad
Squad
Squad
Squad

Squad
Squad
Squad
Squad
Users
Squad
Squad
Squad

Who owns the user experience
Who owns the quality of that experience?
Who owns holistic quality metrics?
Are teams set up with the proper skill to own this?
If not, how will they get there?
Will this lead to a better product for the user in the end?
How do you know that?

Who owns the user experience
Who owns the quality of that experience?
Who owns holistic quality metrics?
Are teams set up with the proper skill to own this?
If not, how will they get there?
Will this lead to a better product for the user in the end?
How do you know that?
Plan
Action
Outcome

Plan Action Outcome
Knowledge Gap
Assess technical trends with a long view and bet earlier
on emerging platforms.
Web and mobile are distinct skills. Mobile engineers are
highly sought after and competitive.
Leverage guilds and RFPs to drive high quality decision
making in a Spotifyway.
Platformisation is an extremely heavy-weight maneavuer.
Hiring and onboarding tons of mobile
engineers doesn’t immediately make you
quick. You need to platformise.

Now that we have this, how do we maximise it’s advantage?

Three rituals and tools we use

1. DIBBs
Data Insights Beliefs Bets
Plan Outcome

Things we believe about the world and why we
believe them.

Bet
WHY
Belief
WHY
Insight
WHY

Bet
WHY
Belief
WHY
Insight
WHY
Data

Applied for both strategy and culture.

DIBBs
Data Insight Belief Bet
http://blog.crisp.se/2016/06/08/henrikkniberg/spotify-rhythm
Desktop
Mobile
USAGE TRENDS
Desktop
Mobile
FRONT END STAFFING
Mobile is overtaking desktop
as primary music consumption
device, and it’s happening quickly.
We’re staffed as a desktop first
company.
The technology stacks between
desktop and mobile are wildly
different.
The future is mobile, and the
companythat wins music will
be a companythat wins mobile.
Hire mobile developers
Retrain desktop developers
Go from 1 team doing mobile for
all features, to mobile being a
concern of everyteam.
STRATEGY

DIBBs
CULTURE

DIBBs
All modern engineering
processes are dialects of
agile/lean which relies on
short iterations with constant
learning.
Bugs are more expensive to
fix the later they are found.
We have had many great ideas
but many of our ideas don’t
work in practice.
.
.
.
CULTURE

DIBBs
We can only learn when the code is
in front of a consumer. Therefore the
more often we show code to
users/customers the fasterwe will
learn.
Code is onlyworth something when
used by consumers, hence we must
ship often to gain value.
learning.
work in practice.
.
.
.
CULTURE

DIBBs
learn.
Speed of iteration beats quality
of iteration.
learning.
work in practice.
.
.
.
CULTURE

DIBBs
learn.
Speed of iteration beats quality
of iteration.
Our organisation, culture,
architecture and process is optimised
for learning and execution speed as
opposed to say developer resource
utilisation or cost per built feature.
.
.
.
.
learning.
work in practice.
.
.
.
CULTURE

Giving away all the
worlds music for free
can bring the music
industry back to health.

2. Innovation Propagation / Bets Board
Action

The Company Bets Kanban
The main priorities of the company
Evaluated and re-prioritised quarterly
Strictly stack ranked
Each bet has an assigned RM
Open to the entire company and  
debated and questioned endlessly
Expresses intent
NOW NEXT

3. Innovation Execution / Taps
Action

A meeting every 6 weeks of everyone involved in an area ofwork.
Two hours. Free for all questioning and asking for help.
Bythe team, for the team.
Taps

Tap Tap
Strategy
Tap Tap
S S S S S S S S S

Taps create peer accountability, natural
pressure to ship and transparency by default.

Putting those who do closest to the effects of doing

TAKE-AWAY #3
Over-reacting beats under-reacting

What happened
after Apps Platform?

How long did it take to build?

How long did it take to build?
4 MONTHS. FROM IDEA
TO MVP IN HAND.

Next Thing
Apps Platform
Next Next Thing

No Apps Platform, No PlayStation

Minimum Viable
Tons and tons and tons of data - Right MVP

1. PlayStation
2. Less important
Number 1 priority - Block and tackle

Plan Action Outcome
Number 1 Priority
Technical Capability
Ubiquity

Plan Action Outcome
Number 1 Priority
Technical Capability
Right MVP
Ubiquity

A whole mess of stuff, really.

Autonomy & Alignment Interact
Empower peer-based teams
Make sure failure travels fastest
Vocabulary shapes culture

Iterate with Purpose To Beat Darwin
Make failure expected and retrospected
Copy with intention
Be Bayes not Darwin

Over-Reacting Beats Under-Reacting
When in doubt, Ship It
Put doers closest to value delivery
Hire for capability to change

#sparkconf
#sparklondon
@marcusf

To Speed Through Failure

Recommended

More Related Content

Similar to To Speed Through Failure (20)

Recently uploaded (20)

To Speed Through Failure