3x Explore, Expand, Extract - Kent Beck

( https://www.youtube.com/watch?v=FlJN6_4yI2A )

Question, to be asked early on.

small investment - big investment (---)

return on investment (red line)

waterfall resurgence

1) let's first do a market survey

2) let's do a business model refinement

3) let's then develop

4) QA

new product development with waterfall

large investment, small chance of payout

Question: what do you have to lose?

first question I need to ask, to decide how to operate

EXPLORE

until you find a growth loop

scratch the right spot

same thing as before, extreme result

not a rational process, or someone else would have done it already

absurd ideas, crazy idea

nobody else is stupid enough to try a bad idea

"how hard would it be to try it"

laughter is a good sign

trade-off: short term, throw away, hack

opposition: indifference

economy of scope: "how little work can I do in order to validate some question"

all about learning, creativity, coverage of the space

EXPAND scale

people care about the idea

barriers to growth: money, floor space, network bandwith, ...

each obstacle is existential: fail one, no princess

economy of risk: "capital is cheap, risk is really expensive"

throwing money at problems

different discipline: engineering, management, ...

all about finding the next thing that's going to kill you

barriers to growth

focused, short period of time

extra hours makes a difference

sitting in the same room makes a difference

EXTRACT value

long term kicks in

automated testing

KPI

exploration is binary (either 0 or massive)

expansion is one number that's gonna kill us

extraction has balanced scorecard

Can you run Exploration and Extract projects at the same time?

-> Facebook

more complicated as a manager

explorers

expanders floating around

extracter

expand projects trump everything:

explore: small change of high payoff

expand: high change of high payoff

extract: high change of  small payoff

-> missing expand = missing the most valuable thing

Facebook everyone is measured on impact

specific numeric goals

works well in explore/expand

extract: smaller improvements, longer projects -> game the system

incentives are all about looking at the upsides, but nothing for looking at the potential downsides

extract: move from impact/outcome -> good decision making

"I'm not gonna add my feature, it's just gonna be too much"

---

question: What do I have to lose?

twitter is my platform to explore

"what do people want me to talk about?"

twitter -> engagement

20x

soundcloud to explore

record one after another

see what get's engagement

---

Explore mode:

don't analyse

figure out how to experiment

Expand mode:

figure out what the next barrier to growth is

address it

get ready for the next one

Extract mode:

how can I make a little improvement

---

Should you write tests?

That depends: do you have anything to lose?

No

Well, only write the tests that help you experiment quicker

Model: 3x for estimates

Model: 3x for estimates

No estimates

explore: estimates are waste

-> great idea in exploration, because you've never done it before

-> estimate is just a guess, no value, wasted effor

extraction:

-> estimates are important data for making decisions

-> do I do project A or project B