Professional Documents
Culture Documents
2
Presenter
ksmith@birst.com
@kevinmsmith
kevinmichaelsmith
Kevin Smith
VP, Embedded Solutions
ksmith@birst.com
What is a data product?
‹#›
A story of building analytics gone wrong
‹#›
Our mission
Make our
existing
SaaS
application
more
engaging by
adding
analytics
‹#›
We have smart
Engineers…
Let’s build it
ourselves and
save some
money! ‹#›
We had resources.
I’m an
I’m an analytics
+ +
analytics user story.
I’m an
user story. Please
analytics
Please I’m implement
an I’m an
user
I’mstory.
an analytics
analytics
implement me!
I’m an
I’m an Please
me! analytics
user story.
user story.
analytics
analytics
implement
user I’m
Please story.
Please
user
user story.
me! story.an
Please
implement implement
analytics
Please Please I’m an
I’m an implement
me!
I’m an me!story.
user
implementanalytics
implement
me!
analytics Please
analytics
me! me! user story.
user implement
user story. story.
Please
Please Please me!
implement
implement
implement
me!
me! me!
‹#›
We had a vision.
$ $ $ $ $ ‹#›
What we actually got.
‹#›
Why was this so painful?
‹#›
The Truth
about Buy
vs. Build for
Embedded
Analytics
‹#›
The BI bar has
1
been raised
‹#›
Time has changed the analytics game
1990 1995 2000 2005 2010 2015
It’s on It’s only I can sort A chart? Real-time I can’t drag this
the 30 days by In color? data? chart to a new
web? old? column NICE! NICE! location, apply
NICE! NICE! headers? filters and have it
NICE! notify me when it
exceeds the
targets I
uploaded?
FAIL. ‹#›
Because table stakes & delighters aren’t
static
Table Stakes Delighters
• Expected • Unexpected
• Can’t compete here • The place to
• Your competition compete
has them • Useful for
• Can’t charge for this differentiation
‹#›
We can build it for less!
• Pay for Highcharts
$100K
$250K $100K $100K times infinity
Buy
year 1 year 2 year 3 infinity
it
• Pay platform fee
• Pay platform fee • Pay platform fee
• Pay for
implementation
• Possibly buy
more storage
(but
it • Cost to build SSO
• Possibly buy • Possibly buy probably
• Cost to build pages more bandwidth more bandwidth even less)
Buy
• Pay for
implementation
Infinite
it • Pay for training
money
‹#›
What we all think we need to do…
Buy charting package
Build ETL
Build Charts
Build Dashboards
‹#›
In reality, there’s a bit more.
Aggregate data
Buy charting package Target setting
Build roll-ups
Build data load Transformations
User permissioning
Build Charts
UI controls
Drill Across
Connect via SSO Filters
QA
Target setting Dimensions
‹#›
What are you
$ 3 skipping in order
to build?
‹#›
The cost of missed core product
value
Cost to
Cost to build
support
analytics
analytics
Cost of NOT
working on your
core
application
‹#›
Your most talented people should
work on unsolved problems.
50% of companies base their decision to build on the fact that they
have the necessary talent to build analytics
From Wayne Eckerson, “Embedded BI: Putting Reporting and Analysis Everywhere”, TechTarget, December, 2014. ‹#›
Where can we add the most
Ask
differentiating value?
Core Product Analytical Platform
• Predictive modeling/forecasting
build
Capability
• Data exploration
• Why did this happen?
• Alerts
• What actions are needed?
Descriptive • Query/drill-down
Easy (er) to build
• What is the problem?
• Ad hoc reports/scorecards
• How many, often, where?
‹#›
Two ways to compete on analytics
Core Value Key Metric Main Challenge
Differentiate
(we’re the leaders!) Separation Unmatchable How far?
‹#›
Two ways to compete on analytics
Differentiate
(we’re the leaders!) Separation Unmatchable How far?
‹#›
It’s an equation, not a single number
‹#›
It’s an equation, not a single number
Will we be able
What’s the What aren’t we
What’s the keep up the
TCO for doing if we build
real cost to development pace
purchasing and how
build for the foreseeable
analytics important is it?
future?
‹#›
1 The cost to buy embedded analytics
Will we be able
What’s the What aren’t we
What’s the keep up the
TCO for doing if we build
real cost to development pace
purchasing and how
build for the foreseeable
analytics important is it?
future?
‹#›
2 The real cost to build
Will we be able
What’s the What aren’t we
What’s the keep up the
TCO for doing if we build
real cost to development pace
purchasing and how
build for the foreseeable
analytics important is it?
future?
‹#›
Capture all of the true costs
Task Type Task Title Description
Purchase of the software to make the charts + maintenance & support for Hi Charts (10 developer
Licensing Buy the software to make the visuals license) -- this ONLY includes production
ETL Build connector to data source Create processes which will connect the charting software to the data source(s)
ETL Perform transformations Transform the data into an analytic ready state for charting
Data Modeling Create data aggregations Perform the roll-ups of data so that you can compare to previous yrs , qtrs, etc.
UI Create dashboard page Create the page which will contain your analytics
QA Perform QA Inspect the analytics and all calculations for accuracy
UI Create dimensions Create the dimensions by which measurement can be examined
Data Modeling Create filters Create the filtering element to include/exclude data by dimension
Data Modeling Build drill-down/across paths Link analytics together so that users can drill down and across to explore causes
Security Build multi-tenancy model Develop model to ensure that customers can't see each other's data
Security Build security model Develop model to ensure that users see only the data they are allowed to see
Data Create data model for targets Build a model to store targets for the metrics
UI Build UI for target setting Create an interface to allow for the setting of targets by metric
UI Build UI for alerts Create the interface for setting alers and notifications for user self-service
Data Modeling Create visualizations Build the visualizations to display the data such as bar charts, line charts, infographics, etc.
Build the pixel perfect reports that use the metrics and dimensions to display the data in a tabluar
Data Modeling Create reports format with rollups, sub-groups, totals, etc.
Create the functionality that allow you to add and remove customers and companies from the
Administrative Build user mangement capabilities analytical functionality
Develop the monitoring capabilities so that you can see the total usage by customer (for billing
Administrative Build monitoring purposes)
‹#›
And calculate both money & time The Powered by Birst Buy vs. Build Calculator
How much does it REALLY cost to build dashboards for your product on your own?
Building your dashboard in-house would cost
Variable Value Our expected cost to build:
at least:
Hourly rate $150.00
# of data sources
# of visualizations
2
15
Your cost to build using these parameters $226,350
# of reports 2 that's 1485 hours or 0.67 FTE years not
# of metrics 30 working on your core product
# of dashboards 1
* not including the time to manage the project
# Dimensions/metric 2
ETL Perform transformations Transform the data into an analytic ready state for charting 30 10 300 $45,000.00
Perform the roll-ups of data so that you can compare to
Data Modeling Create data aggregations previous yrs , qtrs, etc. 30 10 300 $45,000.00
UI Create dashboard page Create the page which will contain your analytics 1 20 20 $3,000.00
QA Perform QA Inspect the analytics and all calculations for accuracy 30 5 150 $22,500.00
Create the dimensions by which measurement can be
UI Create dimensions examined 60 5 300 $45,000.00 ‹#›
Create the filtering element to include/exclude data by
3 Opportunity costs & risks of building
Will we be able
What’s the What aren’t we
What’s the keep up the
TCO for doing if we build
real cost to development pace
purchasing and how
build for the foreseeable
analytics important is it?
future?
‹#›
Four parts to this side of the equation
1 2
enough? enough?
3 4
Do we want to KEEP
What ELSE could we
building it? build with the time?
‹#›
1
Can we build it FAST
enough?
Can we build it fast We’ve got a development team We have resources, but may We don’t have the resources/
dedicated to analytics, fully- have trouble building quickly don’t want to dedicate the
enough? trained in the entire stack, and enough to achieve table stakes. resources to build analytics.
can build quickly.
Can we build it good Yes — we can build all the Maybe — we can add some Nope — we’d have trouble
basics plus functionality to table stakes, not all. Maybe our getting to table stakes.
enough? differentiate ourselves from the delighters will outweigh the
competition. gaps in functionality.
Do we want to keep Yes — this is where we will Maybe — we could add some No — we’d prefer to use our
compete so we’ll devote equal functionality over time but it resources on other things.
building? resources to analytics develop would secondary in importance
as our core app. to the core app.
Could we be doing No — analytics are the app for Maybe — analytics are Yes — we can add more value
us. We consider this to be the important and our core app by working on our core
other things? core of what we do. roadmap is not full. application.
‹#›
The Buy vs. Build Decision Matrix
Low Risk Medium High Risk
(1 point) (3 points) (5 points)
TOTAL
Can we build it fast We’ve got a development We have resources, but may We don’t have the resources/
team dedicated to analytics, have trouble building quickly don’t want to dedicate the
enough? fully-trained in the entire stack, enough to achieve table resources to build analytics. 3
and can build quickly. stakes.
Can we build it good Yes — we can build all the Maybe — we can add some Nope — we’d have trouble
basics plus functionality to table stakes, not all. Maybe getting to table stakes.
enough? differentiate ourselves from the our delighters will outweigh 3
competition. the gaps in functionality.
Do we want to keep Yes — this is where we will Maybe — we could add some No — we’d prefer to use our
compete so we’ll devote equal functionality over time but it resources on other things.
building? resources to analytics develop would secondary in 2
as our core app. importance to the core app.
Could we be doing No — analytics are the app for Maybe — analytics are Yes — we can add more value
us. We consider this to be the important and our core app by working on our core
other things? core of what we do. roadmap is not full. application. 2
Can we build Yes — we can build all the Maybe — we can add some Nope — we’d have
basics plus functionality to table stakes, not all. Maybe trouble getting to table
it good differentiate ourselves from our delighters will outweigh stakes. 5 3 15
enough? the competition. the gaps in functionality.
Do we want Yes — this is where we will Maybe — we could add some No — we’d prefer to use
compete so we’ll devote functionality over time but it our resources on other
to keep equal resources to analytics would secondary in things. 3 2 6
building? develop as our core app. importance to the core app.
Could we be No — analytics are the app Maybe — analytics are Yes — we can add more
for us. We consider this to important and our core app value by working on our
doing other be the core of what we do. roadmap is not full. core application. 2 3 6
things?
GRAND TOTAL 37
(possible 60 points) points
‹#›
The Buy vs. Build Decision Spectrum
You likely will be able to You may be able to build It is unlikely you will get
build fast enough and fast enough and keep to market fast enough or
keep building fast enough building fast enough to be able to stay ahead of
to hold off the beat the competition in your competition
competition select areas
‹#›
Weigh the pros & cons to make the
decision that’s right for your situation
id e
Strategy S
igh risk
Medium H ep
ke
to build &
Cost Side building
$53K
May save
‹#›
In summary: don’t use “internal” criteria
1 The BI bar has been raised
TheThe
complete
complete
guide
guide
to designing,
to designing,
pricing,
pricing,
BEYOND
BEYOND
THE
THE
TECHNICAL
TECHNICAL& launching
& launching
embedded
embedded
analytic
analytic
products
products ©2015
©2015
Birst,
Birst,
IncInc
‹#›
Thank you!
ksmith@birst.com
@kevinmsmith
kevinmichaelsmith
‹#›