Pollster
Shubham Goel
Xavier Malina
Advisor
Marti Hearst
A micro-survey platform
Need
Process
Solution
Impact
■ Finding a real world problem to solve
Problem
• Performance-driven ad Optimization
• A/B Testing
• “Crow d-sourced creativity of Humans”
Our partner &
sponsor
Need
Proc...
Need
Process
Solution
Impact
Problem Version
A
Which ad will drive more users to my website?
Need
Process
Solution
Impact
Problem
Which ad will drive more users to my website?
Version
A
Version
B
Need
Process
Solution
Impact
Problem
Version
A
Version
B
Show Version A
100,000 times
Show Version B
100,000 times
??? Cli...
Time
Version
A
Version
B
Week 1 800 Clicks 30 Clicks
Week 2
Week 3
Result
Total Cost: Best Ad Result:
Which ad will drive
...
Time
Version
A
Version
B
Week 1 800 Clicks 30 Clicks
Week 2 700 Clicks 10 Clicks
Week 3
Result
Total Cost: Best Ad Result:...
Time
Version
A
Version
B
Week 1 800 Clicks 30 Clicks
Week 2 700 Clicks 10 Clicks
Week 3 500 Clicks 60 Clicks
Result
Total ...
Time
Version
A
Version
B
Week 1 800 Clicks 30 Clicks
Week 2 700 Clicks 10 Clicks
Week 3 500 Clicks 60 Clicks
Result 2000 C...
Time
Version
A
Version
B
Week 1 800 Clicks 30 Clicks
Week 2 700 Clicks 10 Clicks
Week 3 500 Clicks 60 Clicks
Result 2000 C...
Time
Version
A
Version
B
Week 1 800 Clicks 30 Clicks
Week 2 700 Clicks 10 Clicks
Week 3 500 Clicks 60 Clicks
Result 2000 C...
Time
Version
A
Version
B
Week 1 110 Clicks 30 Clicks
Week 2 40 Clicks 100 Clicks
Week 3 100 Clicks 120 Clicks
Result 250 C...
Time
Version
A
Version
B
Week 1 110 Clicks 30 Clicks
Week 2 40 Clicks 100 Clicks
Week 3 100 Clicks 120 Clicks
Result 250 C...
“Which photograph should I place on my dating profile?”
“Which slogan resonates most with my customers?”
“Which offer woul...
Need
Process
Solution
Impact
Problem
“The biggest challenge in A/B testing is having the
ability to get quality answers fr...
Research
Need
Process
Solution
Impact
Problem
Need
Process
Solution
Impact
Problem
Data Exploration
Hypothesis 1: A crowd of respondents can identify the superior
version
Survey 1: 30 Boost Writers
How would the
crowd respond?
Our Surveys
Survey 1: 30 Boost Writers
Survey 1: 30 Boost Writers
53%
Correctly Identified
Survey 2: 100 Survey Monkey Respondents
56%
Correctly Identified
A crowd of respondents can identify the superior
version
Hypothesis 1:
Some individuals have a talent for identifying
the superior version
Hypothesis 2:
I think Version A
will attract
more visi...
Some individuals have a talent for identifying
the superior version
A crowd of respondents can identify the superior
versi...
Training respondents can increase a group’s
ability to identify the superior version
Hypothesis 3:
Our Surveys
Training
Survey 3: 35 Boost Writers
70%
Correctly Identified
Survey 4: 50 Survey Monkey Respondents
“This survey is very important to develop the ability to know which ad may have more clicks
than others.
Good exercise. It...
Some individuals have a talent for identifying
the superior version
A crowd of respondents can identify the superior
versi...
Results
• Respondents get better at identifying winner with
training
• Training results can help filter out poor
performer...
Results
• A crowd of respondents can correctly identify the
winning version
• Useful Human feedback
Need
Process
Solution
...
Need
Process
Solution
Impact
Problem
“The biggest challenge in A/B testing is having the ability
to get quality answers fr...
Two groups of users.
Joan
Two groups of users.
Joan Alex
Two groups of users.
Joan Alex
Two groups of users.
• Creates micro-surveys
• Selects Opinion Leaders to poll
• Receives Targeted Feedback
Joan
Is a Poll...
Two groups of users.
• Creates micro-surveys
• Selects Opinion Leaders to poll
• Receives Targeted Feedback
Joan
Is a Poll...
Place your screenshot here
Solution: Navigation Flow (Wireframe)
Solution: Navigation Flow (Interactive Mock)
Pollsters Work Flow
Signup
Login
Pick
Experts
Monitor
Results
Create
Micro-
survey
Opinion Leader Work Flow
Signup
Login
Train
Demonstrate
Expertise
Respond to
Micro-Surveys
Voting Page
■ Respond to Micro-Surveys
■ Train
■ Give Written Feedback
MEAN Stack
Savings
Standard
A/B Test
Pollster
Total Cost $500 - $5,000 $10 - $500
Time 3 Weeks 2 - 7 Days
Feedback • Number of Clicks...
■ Impact
• Micro-survey platform
• Cost-effective
• Quick, Targeted, Human Feedback
• Expandable
Need
Process
Solution
Imp...
“Having [the] type of feedback [gathered by Pollster] allows us to
understand what marketing messaging will work before pu...
Future Work & Potential Uses
Decisions where there is a high cost for making the wrong decision
• Dating profile feedback
...
Pollster Final Project Presentation
Pollster Final Project Presentation
Pollster Final Project Presentation
Pollster Final Project Presentation
Pollster Final Project Presentation
Pollster Final Project Presentation
Pollster Final Project Presentation
of 60

Pollster Final Project Presentation

Pollster is a system that seeks to bring together two markets of users with complementary needs and abilities. Pollsters and Voters. Pollsters are individuals with a decision to make, for whom specific feedback from informed individuals would help them greatly. Pollsters format any question into a small ‘micro-survey’ and select who in our Voter Community they’d like to hear back from. Pollsters can pick which Voters based on filtering criteria like location, experience and age, or other ‘learned’ features like proficiency at math, experience in the retail sector or proven ability to identify winning energy stocks. While our system can be used to poll a crowd indiscriminately, the real power is in its ability to ask questions of individuals who can demonstrate expertise in some ability or area of knowledge. Voters, are individuals with spare time and the desire to earn a little extra income. Voters are notified whenever a new poll arrives in their inbox, and are rewarded by Pollsters for answering polls and providing helpful feedback. Voters wishing to demonstrate some expertise may opt to take training polls aimed at showing Pollsters they know their stuff. In its most recent iteration, this system has been tailored to support Pollsters trying to decide on which version of an online ad to run, in the hope that a smarter decision will drive more users to their website.
Published on: Mar 4, 2016
Published in: Business      
Source: www.slideshare.net


Transcripts - Pollster Final Project Presentation

  • 1. Pollster Shubham Goel Xavier Malina Advisor Marti Hearst A micro-survey platform
  • 2. Need Process Solution Impact ■ Finding a real world problem to solve Problem
  • 3. • Performance-driven ad Optimization • A/B Testing • “Crow d-sourced creativity of Humans” Our partner & sponsor Need Process Solution Impact Problem
  • 4. Need Process Solution Impact Problem Version A Which ad will drive more users to my website?
  • 5. Need Process Solution Impact Problem Which ad will drive more users to my website? Version A Version B
  • 6. Need Process Solution Impact Problem Version A Version B Show Version A 100,000 times Show Version B 100,000 times ??? Clicks ??? Clicks Which ad will drive more users to my website? $5 per 1,000 views $1 per click Minimum cost $1,000
  • 7. Time Version A Version B Week 1 800 Clicks 30 Clicks Week 2 Week 3 Result Total Cost: Best Ad Result: Which ad will drive more users to my website?
  • 8. Time Version A Version B Week 1 800 Clicks 30 Clicks Week 2 700 Clicks 10 Clicks Week 3 Result Total Cost: Best Ad Result: Which ad will drive more users to my website?
  • 9. Time Version A Version B Week 1 800 Clicks 30 Clicks Week 2 700 Clicks 10 Clicks Week 3 500 Clicks 60 Clicks Result Total Cost: Best Ad Result: Which ad will drive more users to my website?
  • 10. Time Version A Version B Week 1 800 Clicks 30 Clicks Week 2 700 Clicks 10 Clicks Week 3 500 Clicks 60 Clicks Result 2000 Clicks 100 Clicks Total Cost: Best Ad Result: Which ad will drive more users to my website?
  • 11. Time Version A Version B Week 1 800 Clicks 30 Clicks Week 2 700 Clicks 10 Clicks Week 3 500 Clicks 60 Clicks Result 2000 Clicks 100 Clicks Total Cost: $3,100 Best Ad Result: Which ad will drive more users to my website?
  • 12. Time Version A Version B Week 1 800 Clicks 30 Clicks Week 2 700 Clicks 10 Clicks Week 3 500 Clicks 60 Clicks Result 2000 Clicks 100 Clicks Total Cost: $3,100 Best Ad A Result: Run Version A Which ad will drive more users to my website?
  • 13. Time Version A Version B Week 1 110 Clicks 30 Clicks Week 2 40 Clicks 100 Clicks Week 3 100 Clicks 120 Clicks Result 250 Clicks 250 Clicks Total Cost: $1,500 Best Ad Which ad will drive more users to my website?
  • 14. Time Version A Version B Week 1 110 Clicks 30 Clicks Week 2 40 Clicks 100 Clicks Week 3 100 Clicks 120 Clicks Result 250 Clicks 250 Clicks Total Cost: $1,500 Best Ad ??? Which ad will drive more users to my website?
  • 15. “Which photograph should I place on my dating profile?” “Which slogan resonates most with my customers?” “Which offer would bring the most new patrons to my store?” Version A ? Version B
  • 16. Need Process Solution Impact Problem “The biggest challenge in A/B testing is having the ability to get quality answers from a crowd of people.” - Bill Sherby, Analytics Manager at Boost Media Inc. The Problem
  • 17. Research Need Process Solution Impact Problem
  • 18. Need Process Solution Impact Problem Data Exploration
  • 19. Hypothesis 1: A crowd of respondents can identify the superior version
  • 20. Survey 1: 30 Boost Writers
  • 21. How would the crowd respond?
  • 22. Our Surveys
  • 23. Survey 1: 30 Boost Writers
  • 24. Survey 1: 30 Boost Writers 53% Correctly Identified
  • 25. Survey 2: 100 Survey Monkey Respondents 56% Correctly Identified
  • 26. A crowd of respondents can identify the superior version Hypothesis 1:
  • 27. Some individuals have a talent for identifying the superior version Hypothesis 2: I think Version A will attract more visitors because…
  • 28. Some individuals have a talent for identifying the superior version A crowd of respondents can identify the superior version Hypothesis 1: Hypothesis 2:
  • 29. Training respondents can increase a group’s ability to identify the superior version Hypothesis 3:
  • 30. Our Surveys
  • 31. Training
  • 32. Survey 3: 35 Boost Writers 70% Correctly Identified Survey 4: 50 Survey Monkey Respondents
  • 33. “This survey is very important to develop the ability to know which ad may have more clicks than others. Good exercise. It also gives me tips and reminders on how to write effective ad copies for Boost clients. Thank you for this opportunity. :) I thought the survey was very interesting and really made me think about why people click on certain ads and not others. Actually It was hard to choose. I believe that I need to learn more [sic.]. But tips & test were useful for me. I hope my answers [sic.] will be useful also. It's been exciting and wonderful. Training Feedback
  • 34. Some individuals have a talent for identifying the superior version A crowd of respondents can identify the superior version Hypothesis 1: Hypothesis 2: Training Opinion Leaders can increase a group’s ability to identify the superior version Hypothesis 3:
  • 35. Results • Respondents get better at identifying winner with training • Training results can help filter out poor performers Need Process Solution Impact Problem
  • 36. Results • A crowd of respondents can correctly identify the winning version • Useful Human feedback Need Process Solution Impact Problem
  • 37. Need Process Solution Impact Problem “The biggest challenge in A/B testing is having the ability to get quality answers from a crowd of people.” - Bill Sherby, Analytics Manager at Boost Media Inc. Solving the Problem
  • 38. Two groups of users. Joan
  • 39. Two groups of users. Joan Alex
  • 40. Two groups of users. Joan Alex
  • 41. Two groups of users. • Creates micro-surveys • Selects Opinion Leaders to poll • Receives Targeted Feedback Joan Is a Pollster Alex Is an Opinion Leader
  • 42. Two groups of users. • Creates micro-surveys • Selects Opinion Leaders to poll • Receives Targeted Feedback Joan Is a Pollster Alex Is an Opinion Leader • Trains to improve abilities • Gives insightful feedback • Receives rewards for responding to micro-surveys
  • 43. Place your screenshot here
  • 44. Solution: Navigation Flow (Wireframe)
  • 45. Solution: Navigation Flow (Interactive Mock)
  • 46. Pollsters Work Flow Signup Login Pick Experts Monitor Results Create Micro- survey
  • 47. Opinion Leader Work Flow Signup Login Train Demonstrate Expertise Respond to Micro-Surveys
  • 48. Voting Page ■ Respond to Micro-Surveys ■ Train ■ Give Written Feedback
  • 49. MEAN Stack
  • 50. Savings Standard A/B Test Pollster Total Cost $500 - $5,000 $10 - $500 Time 3 Weeks 2 - 7 Days Feedback • Number of Clicks • Number of Clicks • Written Feedback from • Humans Need Process Solution Impact Problem
  • 51. ■ Impact • Micro-survey platform • Cost-effective • Quick, Targeted, Human Feedback • Expandable Need Process Solution Impact Problem Process Solution Impact Problem
  • 52. “Having [the] type of feedback [gathered by Pollster] allows us to understand what marketing messaging will work before putting it into market. [“ This] will save us money in the long run and allow us to experiment more with marketing messaging overall.” - Bill Sherby Analytics Manager at Boost Media Inc .
  • 53. Future Work & Potential Uses Decisions where there is a high cost for making the wrong decision • Dating profile feedback • Fashion advice • Brick & Mortar Marketing Initiatives

Related Documents