The document discusses various pitfalls of data-driven development from a product management perspective, focusing on agile software development. It outlines five main pitfalls: 1) having the wrong mindset for failure when conducting experiments, 2) difficulties perceiving progress, 3) overreliance on collecting more data when facing deadlocks, 4) unexpected downsides of democracy in decision making, and 5) lack of domain experts. It emphasizes embracing failures, setting clear success criteria, allowing for random changes, empowering domain experts, and the product manager making the final call.
The Role of FIDO in a Cyber Secure Netherlands: FIDO Paris Seminar.pptx
Pitfalls of Data-Driven Product Management
1. a few words about
Pitfalls of
Data-Driven Development
Product Management
with focus on
2. Product Management
with focus on
a few words about
Customers
a few words about
Developers
Product Management
with focus on
a few words about
Remote Collaboration
Product Management
with focus on
a few words about
Pitfalls of
Data-Driven Development
Product Management
with focus on
10. Where is the data?
Customer Interviews
Usage Analytics
Support TeamsSales Teams
Business PartnersGrowth Experiments
Competitive Analysis Surveys
In Product Feedback
Collection
Social Media
Conferences
Performance
Analytics
Industry Trends
Public backlogUser groups
12. Where is the data?
Qualitative
Quantitative
Surveys
Customer Interviews
Usage Analytics
In Product Feedback Collection
Conferences
Business Partners
48. hmmm…..
- „Is there really no progress?”
- „Or we do not see the progress?”
A
B
49. hmmm…..
- „Is there really no progress?”
- „Or we do not see the progress?”
we’ll get back to it
A
B
50.
51. Snorkels, when used by
vehicles with air-
breathing internal
combustion engines,
sometimes allow
limited deep
fording capability for river
crossing or amphibious
landing operations,
particularly in the case of
tanks and other armored
vehicles.
52. How to tell if snorkel is a
needed feature in your
customers’ car?
53.
54. Pitfall #2B
We do not see the progress
Set your success criteria right
55. Pitfall #2B
We do not see the progress
Set your success criteria right
ahead of time
56. Set your success criteria right
ahead of time
KPI IMPACT HYPOTHESIS (TEXT)
SUCCESS CRITERIA (TEXT)
58. What happens when you hit a deadlock
in DDD-powered Agile project?
Reaction to deadlock
59. - „Let’s collect more data.”
- „Let’s make safe decisions.”
- „Let’s work faster.”
What happens when you hit a deadlock
in DDD-powered Agile project?
60. What happens when you hit a dead
end in DDD-powered Agile project?
more, safer, faster data
Qualitative
Quantitative
- „Let’s collect more data.”
- „Let’s make safe decisions.”
61. What happens when you hit a dead
end in DDD-powered Agile project?
more, safer, faster data
Qualitative
Quantitative
- „Let’s collect more data.”
- „Let’s make safe decisions.”
unreliable
68. What are mutations?
- Seemingly random acts of change
- Dramatic changes as compared
to the rest of the genetic representation
- Chaotic, and potentially unsafe
69. What are mutations?
- Seemingly random acts of change
- Dramatic changes as compared
to the rest of the genetic representation
- Chaotic, and potentially unsafe
70. - Seemingly random acts of change
- Dramatic changes as compared
to the rest of the genetic representation
- Chaotic, and potentially unsafe
- „Let’s collect more data.”
- „Let’s make safe decisions.”
- „Let’s work faster.”
CONFLICT
INTERESTS
OF
73. A role of Domain Expert
- Can extrapolate
from past onto the future
- Someone with the thorough
knowledge of the domain
- Has seen the past and understands
the dynamics of changes
74. What happens when you hit a dead
end in DDD-powered Agile project?
better data
Qualitative
Quantitative
75. What happens when you hit a dead
end in DDD-powered Agile project?
better data
Qualitative
Quantitative
actionable
80. Evolution of APM
let’s hope
for the best
let’s collect
lots of data
and hope
for the best
let’s collect
lots of data,
improve search,
and hope
for the best
let’s collect
lots of data,
improve search,
add custom reports
and hope
for the best
81. Evolution of APM
let’s hope
for the best
let’s collect
lots of data
and hope
for the best
let’s collect
lots of data,
improve search,
and hope
for the best
let’s collect
lots of data,
improve search,
add custom reports
and hope
for the best
let’s provide
answers
based on
the data
we collect
let’s provide
solutions
based on
the answers
we have in
the data
we collect
91. separate noise from information
„outspoken” does not equal „knowledgeable”
as a PM, listen to your peers.
No Domain Experts
92. Your best Domain Expert friend might be
someone you least expect it to be.
check your customer support specialists
check your oldest and grumpiest developers
check the guys who never wear ties
93. We have no Domain Experts
Can a Visionary prove
his/her vision?
94. Can a Visionary prove
his/her vision?
- It is really, really hard.
- Especially if you only
have cheap data.
Qualitative
Quantitative
- Trust your Domain Experts
109. representation fitness function
If this is an algorithm,
why not automate it?
Natural Language Understanding
Automatic verbose feedback collection
Prioritised backlog of experiments
A/B tests for validation
110. representation fitness function
If this is an algorithm,
why not automate it?
Natural Language Understanding
Automatic verbose feedback collection
Prioritised backlog of experiments
A/B tests for validation
80%of development effort by 2025
111. representation fitness function
If this is an algorithm,
why not automate it?
Natural Language Understanding
Automatic verbose feedback collection
Prioritised backlog of experiments
Self-organising software
A/B tests for validation
80%of cost by ??????