ModCloth uses Tableau to enable stakeholders across the company to access and analyze data independently. By training stakeholders in Tableau, the data team is able to focus on more complex analyses while stakeholders can answer questions with same-day data. Some challenges in training include different skill levels and goals amongst stakeholders. ModCloth addresses this through tailored trainings and office hours. Since implementing stakeholder training, the data team spends less time on routine tasks and more on modeling and products while stakeholders complete over 200 additional requests per quarter in Tableau.
Aspirational Block Program Block Syaldey District - Almora
Tableau Conference 2014 Presentation
1. From Marketing to Merchandising: Using
Tableau to Enable ModCloth Stakeholders
with the Power of Data
Presented by:
Krystal St. Julien
Data Analyst, ModCloth
3. More than a fashion retailer…
Our Mission:
To inspire personal style and help
customers feel like the best
version of themselves.
Our Purpose:
To democratize fashion and
decor around the world.
4. A place where data inspires fashion
“It would be
PERFECT – if it
wasn’t for the
weird ruffle by
the waist……?”
~ Morgan
5. A place where data inspires fashion
“It would be
PERFECT – if it
wasn’t for the
weird ruffle by
the waist……?”
~ Morgan
6. Krystal St. Julien
Analyst
Link Doucedame
Junior Analyst
Shawn Davis
VP of Analytics
Julia King
Sr. Mgr. of
Analytics
Aiyesha Ma
Data Scientist
Lauren Anderson
Sr. BI Analyst
Anna Peterson
Analyst
ModCloth
Data Team
Julia Kirkpatrick
Sr. Researcher
Cherie Yagi
Researcher
Christine Wu
Sr. Web Analyst
Andy
Sevastopoulos
Lead Analyst
7. Jobs currently executed by ModCloth Data Team
• Data pulling and Data Delivery
• Ad Hoc Analysis (for business/strategy
recommendations)
• Dashboard/Automated Analysis Development
• Data Warehousing (creating and storing data)
• Data Modeling and Prediction
• Development of Data Products
• Teaching Stakeholders About Data, How to Use
it, and How to Present it
8. Jobs currently executed by ModCloth Data Team
• Data pulling and Data Delivery
• Ad Hoc Analysis (for business/strategy
recommendations)
• Dashboard/Automated Analysis Development
• Data Warehousing (creating and storing data)
• Data Modeling and Prediction
• Development of Data Products
• Teaching Stakeholders About Data, How to Use
it, and How to Present it
Jobs currently
executed by
Analysts AND
stakeholders!
9. Jobs currently executed by ModCloth Data Team
• Data pulling and Data Delivery
• Ad Hoc Analysis (for business/strategy
recommendations)
• Dashboard/Automated Analysis Development
• Data Warehousing (creating and storing data)
• Data Modeling and Prediction
• Development of Data Products
• Teaching Stakeholders About Data, How to Use
it, and How to Present it
Jobs currently
executed by
Analysts AND
stakeholders!
10. Why invest the time and energy in making data
stakeholder-friendly?
• Our current backlog:
over 100 requests
• Wait time for an analyst:
a couple of days to
several months
• Access to a user-friendly
analytics tool means
stakeholders can have
same-day data delivery!
• Communicating about
algorithms can be
difficult in the abstract.
14. Hurdles to overcome when teaching non-technical
stakeholders
• Some common stakeholder challenges include:
• Misunderstood jargon/misaligned data
communication
• Different stakeholders will have different
goals/needs
• Lack of knowledge of the tool’s full capability
and data available
15. Hurdles to overcome when teaching non-technical
stakeholders
• Some common stakeholder challenges include:
• Misunderstood jargon/misaligned data
communication
• Different stakeholders will have different
goals/needs
• Lack of knowledge of the tool’s full capability
and data available
21. Hurdles to overcome when teaching non-technical
stakeholders
• Some common stakeholder challenges include:
• Misunderstood jargon/misaligned data
communication
• Different stakeholders will have different
goals/needs
• Lack of knowledge of the tool’s full capability and
data available
22. ModCloth has MANY data use-cases
Merchandising
Assortment planning – What are
customers purchasing?
Finance Sales reports and dashboards
Human Resources Reviewing company stats
Public Relations Data gathering for press cards
Product Mangers
Data diagnostics – What is going
well/failing on our site?
Operations/Shipping
What are customers ordering?
Identification of fraudulent orders
Marketing
Marketing channel performance
reporting
23. Implement team/topic specific training
Intro training:
Tableau navigation
Topic-specific:
Dashboards and Data sources
Super-user:
Tableau desktop
24. Results of team/topic specific training
“It was tailored to our specific
needs and demonstrated how to
access/utilize key reports. (Versus
previous training session that was
much more general and hard to
follow.)”
“I liked that this training was
specific to our category so we
could discuss our team’s needs.”
“I liked how we walked through
the specific reports that will be
most useful for our specific team.
I walked out of the training with a
clear understanding of the
information I can find in Tableau
and how to pull it.”
25. Hurdles to overcome when teaching non-technical
stakeholders
• Some common stakeholder challenges include:
• Misunderstood jargon/misaligned data
communication
• Different stakeholders will have different
goals/needs
• Lack of knowledge of the tool’s full capability
and data available
26. Training should include exercises showing
stakeholders what can be done
• Filter on date
• Bring in “Vendor Name” dimension
• Bring in “Count of Products” measure
• Multiple visualizations can be useful
• Allow ~5 minutes of individual
work time per question
• Go through the question as a
team using the following steps:
• What filters will we need?
• What dimensions do we want
to see?
• What are we trying to
measure?
• Which visualization would
you prefer to see if someone
were presenting this data to
you?
27. Lather, rinse, repeat… implement office hours
“[I want to get]
individual help running
[my] own reports.”
“Wish we spent more time
doing live scenarios,
practicing using the tool,
reviewing the metrics
available, how to pull ad
hoc reports, etc.”
• We currently host 4 hours of Tableau office hours a week
• ~50% of office hour time is scheduled and used
28. Stakeholders can learn from Super Users!
Finance Merchandising Marketing
• Provided with Tableau Desktop
• Allowed to create and modify dashboards
• First line of defense for team-questions
30. Pull data about the company without pinging a
database
Objective: Collate a list of Tops and their associated Lengths.
Click and drag metrics into place
31. Quick ad hoc analysis – answering a question
Question: Do customers consider reviews more helpful when the reviewer’s
measurements are associated?
Click and drag metrics into place
Use a quick calculation to get Avg Count of Helpful Votes per Review
Use “show me” to visualize data as bar chart
32. Trended analysis for dashboarding
Objective: Find the running sum of new customers that are placing repeat
orders over time.
Write logic to find the date difference between date when order
number = 1 and date when order number = 2
Click and drag metrics into place
Implement a quick calculation to produce running total
34. Pros and Cons
Pros Cons
• Stakeholders do not
have to wait for an
analyst to come
available
• Project iterations are
easily
accomplished/easy to
shift direction
• Analysts can focus on
more impactful
analyses, models, and
predictions
• Appropriate time for
teaching/training as
well as follow-up
training must be
allocated
• When tools are
updated/changed,
additional training is
required
• Tools come at a
monetary cost
35. Usage at ModCloth
• Last quarter, of ~200 potential Tableau users outside of the analytics
team,…
• MC analytics completed 2 hours of training and ~26 hours of office
hours, contributing to:
• 120 users logging-in
• 114 users looking at readily available dashboards
• 96 users accessing data via a data source
• 30 users publishing at least 1 workbook to share
• an estimated >220 additional “requests” being resolved by
teaching stakeholders how to use Tableau
Most of these users were trained in Jan or Feb of 2014 (8 hours offered)
36. My balance of time before/after training
implementation
JOB/SKILL BEFORE AFTER
Data pulling and Data Delivery 15 5
Ad Hoc Analysis (for business/strategy
recommendations)
35 20
Dashboard/Automated Analysis Development 25 20
Data Warehousing (creating and storing data) 10 10
Data Modeling and Prediction 5 20
Development of Data Products 0 5
Teaching Stakeholders About Data, How to Use it, and
How to Present it
10 20
37. My balance of time before/after training
implementation
JOB/SKILL BEFORE AFTER
Data pulling and Data Delivery 15 5
Ad Hoc Analysis (for business/strategy
recommendations)
35 20
Dashboard/Automated Analysis Development 25 20
Data Warehousing (creating and storing data) 10 10
Data Modeling and Prediction 5 20
Development of Data Products 0 5
Teaching Stakeholders About Data, How to Use it, and
How to Present it
10 20
39. Please take the session survey
1.Tap to this session on the Schedule tab of the
Data14 app
2. Scroll down to “Feedback” and tap through the
3-question survey
3.Tap Send Feedback
Hinweis der Redaktion
- But before I dive into that topic, I’d like to tell you a bit about the company I work for
How many of you have heard of MC or even purchased from our site? (Good!)
ModCloth is the leading social shopping e-tailer of women's fashion and décor.
Co-founder Susan Gregg Koger found an early love of vintage fashions, and that passion has grown into a company of over 500 employees with 3 office locations (site)
MC strives to cultivate a unique catalog of products as well as a strong commitment to customer care, leading to customers who are more than shoppers, but really part of our community!
As leaders of this social retail community, we hope to be more than a retailer, we want to
inspire personal style and help customers feel like the best versions of themselves as well as to
disrupt the fashion industry by democratizing fashion and décor around the world.
To do this, we allow our data and customer involement to inspire us.
We allow data we collect from our customer help us influence the direction of our catalog through several means, including
interactive programs that provide us data about our consumers and their style:
2 of these programs are Make the cut and Be the Buyer
… So who at modcloth actually uses this data?...
Make the Cut, our crowdsourced design program where our community is invited to submit clothing sketches we may produce and sell on our site.
You can see here an example. In this case we received the sketch from a member of the ModCommunity, and the design received enough support to go into production.
Be The Buyer, our pick it/skip it program that allows our customers to tell us why they do or don’t love our new or upcoming styles
In this example, you can see that the ModCommunity was not a huge fan of the original design, community members submitted reviews like this one from Morgan… and the modified style was much more popular
These are just 2 of the many ways ModCloth has decided to use data to direct fashion and customer experience…
To do this, we allow our data and customer involement to inspire us.
We allow data we collect from our customer help us influence the direction of our catalog through several means, including
interactive programs that provide us data about our consumers and their style:
2 of these programs are Make the cut and Be the Buyer
… So who at modcloth actually uses this data?...
Make the Cut, our crowdsourced design program where our community is invited to submit clothing sketches we may produce and sell on our site.
You can see here an example. In this case we received the sketch from a member of the ModCommunity, and the design received enough support to go into production.
Be The Buyer, our pick it/skip it program that allows our customers to tell us why they do or don’t love our new or upcoming styles
In this example, you can see that the ModCommunity was not a huge fan of the original design, community members submitted reviews like this one from Morgan… and the modified style was much more popular
These are just 2 of the many ways ModCloth has decided to use data to direct fashion and customer experience…
Our current data team is split into 4 major factions that allow us to take on a large breadth and depth of data projects:
Our BI analyst is our go-between for data warehousing and analytics
Our Data Scientist engineers data products in collaboration with both analytics and engineering teams at MC
Experience research focuses on consumer insights and trying to gather direct from consumer data via surveys and face-to-face interaction
Analytics focuses on analyzing data gathered through interactions with the MC site: (next slide)
None of the analysts are truly embedded with any given team, although we do have specific spheres or liaisonships that we take responsibility for (for example I work closely with our Social, Merchandising, and Operations teams)
From here on in, I will use the word stakeholder multiple times. By stakeholder, I mean the intra-company partners that the analysts are working with or for.
Now, before I get into the main structure of my talk, I’d like to know a couple of things about the audience…
In doing this, our day-to-day can include any of the following:…
Some of these tasks can be heavily aided and made simpler by the use of data visualization tools such as Pentaho, Omniture, Tableau and others…
From here forward, I will talk about my experiences in using Tableau, but I wanted to make it clear that the principles I will discuss should be applicable for any good data vis gui
(examples on next slides)
None of the analysts are truly embedded with any given team, although we do have specific spheres or liaisonships that we take responsibility for (for example I work closely with our Social, Merchandising, and Operations teams)
From here on in, I will use the word stakeholder multiple times. By stakeholder, I mean the intra-company partners that the analysts are working with or for.
Now, before I get into the main structure of my talk, I’d like to know a couple of things about the audience…
In doing this, our day-to-day can include any of the following:…
Some of these tasks can be heavily aided and made simpler by the use of data visualization tools such as Pentaho, Omniture, Tableau and others…
From here forward, I will talk about my experiences in using Tableau, but I wanted to make it clear that the principles I will discuss should be applicable for any good data vis gui
(examples on next slides)
None of the analysts are truly embedded with any given team, although we do have specific spheres or liaisonships that we take responsibility for (for example I work closely with our Social, Merchandising, and Operations teams)
From here on in, I will use the word stakeholder multiple times. By stakeholder, I mean the intra-company partners that the analysts are working with or for.
Now, before I get into the main structure of my talk, I’d like to know a couple of things about the audience…
In doing this, our day-to-day can include any of the following:…
Some of these tasks can be heavily aided and made simpler by the use of data visualization tools such as Pentaho, Omniture, Tableau and others…
From here forward, I will talk about my experiences in using Tableau, but I wanted to make it clear that the principles I will discuss should be applicable for any good data vis gui
(examples on next slides)
But the ModCloth analytics team has been able to extract the most power out of these data tools by teaching non-analyst stakeholders at our company (people to whom we often deliver data and recommendations) how to perform quick analyses themselves
{Animate the point-by-point order and maybe set up photo shoot to replace this stock art}
Subscribing by the old adage of teaching a man how to fish
- This is important because our teams current backlog has over 100 requests, and not every request can be prioritized immediately
Rather than focusing the rest of my talk on the benefits of data visualization in teaching others, I thought I would go through the benefit of teaching others in creating data visualizations
To do this, I will discuss some of my own personal insights I have gathered in training non-technical individuals how to use a data visualization GUI
Teaching non-technical individuals how to use a data visualization GUI is not as simple as showing them the functionality and letting them loose. As a teacher of these tools, I have identified my top hurdles to overcome: (as listed on slide)
The following set of slides will detail my experiences in trying to overcome these hurdles
Teaching non-technical individuals how to use a data visualization GUI is not as simple as showing them the functionality and letting them loose. As a teacher of these tools, I have identified my top hurdles to overcome: (as listed on slide)
The following set of slides will detail my experiences in trying to overcome these hurdles
Data is clean – We expect stakeholders to be able to use them with minimal input from the analytics team
How many data sources do we have – how many are truly optimized??
Data is clean – We expect stakeholders to be able to use them with minimal input from the analytics team
Data is clean – We expect stakeholders to be able to use them with minimal input from the analytics team
How many data sources do we have – how many are truly optimized??
For a stakeholder to create a powerful visualization, it must be meaningful, and often the language used by the analytics team may not be the most straightforward or intuitive
To make sure the stakeholder has the best chance at creating a meaningful vis, our team goes to lengths to make sure data is stored and presented with non-technical individuals in mind
(anecdote about stakeholders using the wrong data and being confused about why the bar chart looked so weird)
{Look into tableau comments bubbles and maybe show labels in dashboards themselves as examples?}
Reviews data source
Teaching non-technical individuals how to use a data visualization GUI is not as simple as showing them the functionality and letting them loose. As a teacher of these tools, I have identified my top hurdles to overcome: (as listed on slide)
The following set of slides will detail my experiences in trying to overcome these hurdles
Teaching non-technical individuals how to use a data visualization GUI is not as simple as showing them the functionality and letting them loose. As a teacher of these tools, I have identified my top hurdles to overcome: (as listed on slide)
The following set of slides will detail my experiences in trying to overcome these hurdles
Since these tools can be very far-reaching with many, many functionalities, it is a logistically sound strategy to entice and intrigue stakeholders by making trainings team or topic specific rather than general
Anecdote about how trainings began as general, but we found we were often repeating ourselves during out-of-training hours to people who were interested but didn’t gain the appropriate knowledge during training. Now, trainings are much more useful for stakeholders as attested to by the stakeholders themselves (slide)
Since these tools can be very far-reaching with many, many functionalities, it is a logistically sound strategy to entice and intrigue stakeholders by making trainings team or topic specific rather than general
Anecdote about how trainings began as general, but we found we were often repeating ourselves during out-of-training hours to people who were interested but didn’t gain the appropriate knowledge during training. Now, trainings are much more useful for stakeholders as attested to by the stakeholders themselves (slide)
Teaching non-technical individuals how to use a data visualization GUI is not as simple as showing them the functionality and letting them loose. As a teacher of these tools, I have identified my top hurdles to overcome: (as listed on slide)
The following set of slides will detail my experiences in trying to overcome these hurdles
-
(slide)
- Write logic to find dates when order_num = 1 and dates when order_num = 2
Write logic to find the date diff between order_date_1 and order_date_2
Plot date difference (days between) vs Sum of New Customers (variable we provide)
Implement a quick calc to produce running total
I’m not going to act like there are no downsides to teaching others how to become basic data vis productionists using a data vis GUI,
but I hope to convince you in the next couple of slides that the trade-offs are well worth it
- Before I go into some of data showing the success MC has experienced by using this model of empowering “non-technical analysts”, I’d like to go through a quick pros/cons list
(slide)
(slide)
- Write logic to find dates when order_num = 1 and dates when order_num = 2
Write logic to find the date diff between order_date_1 and order_date_2
Plot date difference (days between) vs Sum of New Customers (variable we provide)
Implement a quick calc to produce running total
I’m not going to act like there are no downsides to teaching others how to become basic data vis productionists using a data vis GUI,
but I hope to convince you in the next couple of slides that the trade-offs are well worth it
- Before I go into some of data showing the success MC has experienced by using this model of empowering “non-technical analysts”, I’d like to go through a quick pros/cons list
First, a quick pros/cons list
{MC analytics completed ~75 requests in ~943 hours of working time (~12 working hours per request)}
By teaching non-technical individuals at our company how to use data vis GUIs in an effective manner, we have been able to add only a few hours of additional work per week to increase our “request” delivery by nearly 2x
In addition, the majority of data pulls/simple data delivery can now be done by the stakeholders themselves, giving the analysts on the team more latitude to pursue topics they deem more interesting such as data modeling and prediction
Even more “requests” have been filled through use of additional tools such as Omniture and even google analytics (which has many of the data housing and visualization functions)
None of the analysts are truly embedded with any given team, although we do have specific spheres or liaisonships that we take responsibility for (for example I work closely with our Social, Merchandising, and Operations teams)
From here on in, I will use the word stakeholder multiple times. By stakeholder, I mean the intra-company partners that the analysts are working with or for.
Now, before I get into the main structure of my talk, I’d like to know a couple of things about the audience…
In doing this, our day-to-day can include any of the following:…
Some of these tasks can be heavily aided and made simpler by the use of data visualization tools such as Pentaho, Omniture, Tableau and others…
From here forward, I will talk about my experiences in using Tableau, but I wanted to make it clear that the principles I will discuss should be applicable for any good data vis gui
(examples on next slides)
None of the analysts are truly embedded with any given team, although we do have specific spheres or liaisonships that we take responsibility for (for example I work closely with our Social, Merchandising, and Operations teams)
From here on in, I will use the word stakeholder multiple times. By stakeholder, I mean the intra-company partners that the analysts are working with or for.
Now, before I get into the main structure of my talk, I’d like to know a couple of things about the audience…
In doing this, our day-to-day can include any of the following:…
Some of these tasks can be heavily aided and made simpler by the use of data visualization tools such as Pentaho, Omniture, Tableau and others…
From here forward, I will talk about my experiences in using Tableau, but I wanted to make it clear that the principles I will discuss should be applicable for any good data vis gui
(examples on next slides)