Hubs: Hubs Data

From edegan.com
Revision as of 14:20, 21 July 2016 by ToddR (talk | contribs) (→‎Group 4)
Jump to navigation Jump to search

Background

This page represents the work used for mechanical turks for the paper: Hubs (Academic Paper). As of Spring 2016, a list of potential Hubs with a set of characteristics was created. Many of these are not what will be defined as Hubs. We will be creating a scorecard to help subjectively define Hubs based on certain characteristics.

For more information on Mechanical Turks in general, see Mechanical Turk (Tool).

The main goal of the mechanical turk is to automate the collection of variables for potential hubs as much as possible. The key steps for the project are:

  1. Creating a comprehensive list of potential hubs
  2. Determining the best variables for the scorecard
  3. Building "filters" for automating the collection
  4. Running and auditing of the automation
  5. Collecting the remaining manual data


Variables to be Used

Current Complete List

As of Week of 7/11

  1. Onsite Venture Capital
    • Assets Under Management
    • Number
  2. Onsite Angel Investors
  3. Onsite Mentors
  4. Founding Date
  5. Site URL
  6. Office hours investors
  7. Office hours mentor/advisors
  8. Onsite temporary workshops
  9. Onsite mentors
  10. Networking Meetups
  11. Sponsors/Partners
    • University
    • Corporate
  12. Curriculum
  13. Onsite code school
  14. Alumni Network
  15. Nonprofit status
  16. Mission statement
  17. Specific Industry
  18. Price for a space
  19. Price for office
  20. Twitter activity
  21. Size (sqft)
  22. Size (# companies)
  23. Onsite accelerator
  24. Community membership??
  25. Franchise
  26. Multiple locations within city

Grouping of Variables

There are a few categories the majority of the variables fall under

Group 1: Low Hanging Fruit Variables in this group are very easy to find and automate.

  1. Price for a space + office
  2. Twitter Activity
  3. Founding Date
  4. URL
  5. Mission Statement
  6. Nonprofit
  7. Sponsors/Partners
  8. Specific Industry


Group 2: The Difficult to Find There are certain variables where the information is not readily available online or difficult to find.

  1. Size (can try to find press releases)


Group 3: In Between 1 and 2 Variables that aren't too easy or difficult to find and automate.

  1. Onsite accelerator
  2. Alumni mentor---vs. other mentors???


Group 4: The Hard to Differentiate The key property of this group is that there are several similar variables, which would be difficult for a turk to differentiate. In order to fix this, we will need to create filters akin to the DSM5 scorecard. See the below section.

  1. Onsite VC v. Angel Investors
  2. Onsite OH Investors v. mentors
  3. Onsite temporary workshops v. networking events
  4. Curriculum v. code school


Group 5: The Need further Discussion Before Collection Variables that need to be developed more prior to collection.

  1. Franchise and multiple locations within a city
  2. Community Membership

Filters/Scorecard

General Approach

The Scorecard will be broken down into three main parts: description, characteristics, andTBD parts. The procedure for creating these will be as follows: the description will be determined, develop the characteristics after looking over examples, the creation of possible mechanical turks that have complete accuracy even if not comprehension (e.g. a task will that always guarantees that there is an onsite mentor that covers only 40% of firms, but never misspecifies the existence of mentors), and auditing of the results.

Example

Curriculum

  • Desc: The potential hub provides training programs for the founders of startups that might have human capital deficits that will lead to them not being about to adequately implement their ideas.
  • Characteristics:
    • Education that is for a founder (as opposed to code schools which can be for people who just want to join a startup)
      • Code schools are for startup labor supply
    • Active input into a current entrepreneurial endeavor
      • e.g. " The program is designed to augment and support the real-life business experiences that the students are facing every day in their entrepreneurial endeavors"
    • Not an ad hoc session, not a one time meeting but a full "course", evidence of this could be
    • Has evidence of a integrated curriculum leading to a new compentance
    • Has evidence of a set fixed start and end dates that last XXX long
    • Is a session linked to others that regularly occurs
  • TBD points
    • Do we care about outsourcing?
  • Potential Turk

Code School

  • Desc: training programs that teach coding, data processing, webpage building and other technical skills.
  • Characteristics:
    • Target group are the developers or people who want to join the startups but not the founders themselves
    • Scheduled classes, not a one time meeting (as opposed to workshops)

Temporary Workshops

  • Desc:a discussion/learning of a group of people on specific subjects
  • Characteristic:
    • One time
    • Have a topic/subject/goal
      • e.g. learn to code workshop: Java script 101

Additional Resources

  1. Mechanical Turk (Tool)
  2. Veeral has created a google automating procedure for different lists


Work in Progress

Goals for WIP

  1. For GROUP 1, creation of mechanical turk steps:
    • EXAMPLE:
    • Twitter Activity
      • STATUS: Complete/In Progress/Not Started
      • Previously Collected: Yes/No
      • Published on Mechanical Turk: Yes/No
      • Audited: Yes/No
      • Updates:
      • Code:
  2. For GROUP 4:
    1. Scorecard Example
    2. Potential Mechanical Turk Steps (e.g. if specific organization is on website)
    3. Mechanical Turk Example (GROUP 1)
    4. Add Comments on:
      1. How much manual work remains/What is missing
      2. Any remaining difficulties
  3. For GROUPS 2 and 3:
    1. Brainstorm potential ways to find data
    2. Follow Steps in Group1

Steps Needed to Complete

  1. Establish automation process for Groups 1-3
    • Status (7/21): G/Y: Founding date issues
    • Begin Date: Started
    • Reach Goal: Complete By Friday 7/22
  2. Differentiate variables in Group 4
    • Status (7/21): Green: much progress has been made
    • Begin Date: Started
    • Reach Goal: Complete by Wednesday 7/27
  3. Test processes and audit
    • Status (7/21): NS
    • Begin Date: Thursday 7/21
    • Reach Goal: Complete by Friday 7/28
  4. Have a comprehensive list of potential hubs
    • Status (7/21): NS
    • Begin Date: Thursday 7/21
    • Reach Goal: Complete by Tuesday 7/26
  5. Fill in Remaining Data Manually
    • Status (7/21): NS
    • Begin Date: Monday 7/25
    • Reach Goal: Complete by Friday 7/29

Actual WIP

Group 1

  1. Twitter Activity
    • STATUS: Complete
    • Previously Collected: YES/NO - Recorded 2/1/0 to represent activity level, but not same as we are
    • Published on Mechanical Turk: Yes
    • AUDITED: Yes
      • Audit Results: Comparing to 30 that manually done, for twitter handle, all 3 turkers agreed with our results 81% of the time, but at least 2 turkers agreed with our results 98% (the exception was a company that had several twitter handles based on location). Results were 52% and 89% respectively.
    • UPDATES:
      • UPDATE (7/14): Updated turk to reflect our desired formats
      • UPDATE (7/12): Audited
      • UPDATE (7/11): uploaded and published on amazon's mechanical turk site. Given the time cost to either record number of tweets in a month or look up more than 10 tweets, we decided to record the date of the last 10th tweet. Using a sample of ~10 companies, We noticed minimal differences in data observations among using 10,20, and 30 tweets.
    • CODE
      1. Copy the text in the Search Text into a search engine.
      2. Click on result from twitter.com with the company name. If the link does not appear on the first 3 pages, record DNE for both outputs
      3. Record the company's Twitter Handle into Twitter Handle
      4. Record the date (MM/DD/YY) of that tweet for Twitter Activity. If there are less than 10 tweets, record DNE.
  2. URL
    • STATUS: In Progress
    • Previously Collected: YES
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/18): Code written, expected time for each assignment is <15 seconds - pay rate, therefore, recommended $.04
    • CODE
      1. Copy the text in the Search Text into a search engine.
      2. Record the URL of the first result in the following format www.___.__/ (e.g. if url is example.us/other, record www.example.us/)
  3. Mission Statement
    • STATUS: In Progress
    • Previously Collected: YES
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/18): Code written, expected time for each assignment is 20-30 seconds - pay rate, therefore, recommended $.08
    • CODE
      1. Copy the text in the Search Text 1 into a search engine (will include site:__ from Company's URL).
      2. Click on first link that is a subsection (e.g. "Mission", "About") from company's website (see Company's URL)
      3. If this does not exist, repeat steps 1 and 2 with Search Text 2
      4. If this does not exist, got to Company's URL
      5. Record the main text on the page up to five paragraphs (some of these will be a single line). Do NOT record subsections.
      6. If locating the main text in the prior step is unclear, record "Unclear"
      7. If no text exists, record "DNE"
  4. Specific Industry
    • STATUS: In Progress
    • Previously Collected: YES/NO, based on LinkedIn identifier
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/21): Given that most companies include their specialty in mission statement and difficulty to turk, we will manually check each mission statement and mark it accordingly.
    • CODE
      1. NONE
  5. Nonprofit
    • STATUS: In Progress
    • Previously Collected: NO
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
      • REQUIRES ADDITIONAL STEPS: YES (need to double check results)
    • UPDATES:
      • UPDATE (7/19): Code written, code 2 of 2 is believed to be more accurate and efficient. Expected time to complete is 15 seconds - pay rate, therefore, recommended $.04
    • CODE 1 of 2
      1. Go to Company's URL.
      2. Go to links (sometimes will be sections of the URL page) that describe the company, usually they are labelled: 'About', 'Our Story,' 'Mission'.
      3. If none of these exist, record DNE for PAGES
      4. Look for the word 'profit'/'nonprofit'/'non-profit'/'not-for-profit' (with or without -)
      5. If any of the key words exist is identified, record as 1, otherwise 0 for EXISTS (1/0).
      6. If it is marked as 1, record all sentences that the word is found in under SENTENCES.
      7. If the links do exist, record the name of the link under PAGES
      8. Repeat steps 4, 5, and 6 on the pages that were linked.
    • CODE 2 of 2
      1. Copy the text from Search Text into the search bar at http://www.guidestar.org/.
      2. Record all Organization Names that appear
      3. If no results appear, record DNE
  6. Sponsors/Partners
    • STATUS: In Progress
    • Previously Collected: NO
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/21): Code written, but may require additional manual work. Expected time to complete is 45 seconds due to a potential list of a lot of sponsors/partners - pay rate, therefore, recommended $.12.
    • CODE
      1. Choose first result from Search Text 1 and Search Text 2 (allintext: Sponsors/Partnerrs site:URL)
      2. Record all Sponsors from Search Text 1 into SPONSORS. If there does not exist a list or the link was for only 1 sponsor, record DNE.
      3. If any Sponsors from Search Text 1 include a University or College (will be listed in name), record them into UNIVERSITY SPONSORS
      4. Record all Partners from Search Text 2 into PARTNERS. If there does not exist a list or the link was for only 1 partner, record DNE.
      5. If any Partners from Search Text 2 include a University or College (will be listed in name), record them into UNIVERSITY PARTNERS
  7. Price for a space + office
    • STATUS: Not Started
    • Previously Collected: YES
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/_): TBD
    • CODE
      1. Copy the text in the Search Text into a search engine.
      2. TBD
  8. Founding Date
    • STATUS: Not Started
    • Previously Collected: YES, but only year
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/_): TBD
    • CODE
      1. Copy the text in the Search Text into a search engine.
      2. TBD

Group 2

  1. Size
    • BRAINSTORM: (7/19) 1), 2), 3): search allintext: sqft/square foot/square feet site: company URL. 4) Company Name, city, square feet and then choose frist result. Process might be easier (and cheaper) if Veeral runs code firstto eliminate a bunch of 0 result returned.
    • STATUS: In Progress
    • Previously Collected: YES/NO, many missing
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/19): Brainstorm and code updated
    • CODE
      1. Copy the text in the Search Text 1 into a search engine.
      2. Record DNE if 0 results returned in SEARCH 1
      3. If there is a result, click first link in which result search text appears and record the sentence in which the text appears in SEARCH 1
      4. Repeat Steps 1-3 for Search Text 2 and 3 and record in respective SEARCH 2 and SEARCH 3 respectively

Group 3

  1. Mentors
    • BRAINSTORM: Current form of this variable seems to be too general.
    • STATUS: In Progress
    • Previously Collected: NO
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/19): Two possible codes written. First one requires more manual work
    • CODE 1 of 2
      1. Go to Company URL
      2. Look for links related to mentorship such as 'mentors', 'mentorship' or 'mentoring programs'.
      3. If the key words can be identified, record 1 in BINARY, copy the sentence it is included in SENTENCE, and record urlhome in PAGE.
      4. If there is no explicit 'mentoring' section, look for links related to a description of the company, such as: 'About,' 'Our Team,' 'Our Mission,' etc., and look for a subsection or mention of mentor/mentorship/mentoring.
      5. If these exist, record 1 in BINARY, copy the sentence it is included in SENTENCE, and record the link name clicked in PAGE.
      6. If not, go to links related to membership 'benefits,' 'perks,' or related and repeat Step 5.
      7. If none of these steps result in a mark of 1, mark as 0.
    • CODE 2 of 2
      1. Copy Search Text into search engine
      2. Mark as 1 if reliable site is populated, 0 otherwise
  2. Onsite Accelerator
    • BRAINSTORM: Need a count.
    • STATUS: Not Started
    • Previously Collected: YES/NO, only a binary variable
    • Published on Mechanical Turk: NO
    • AUDITED: NO
      • Audit Results: TBD
    • UPDATES:
      • UPDATE (7/_): TBD
    • CODE
      1. Copy the text in the Search Text into a search engine.
      2. TBD

Group 4

Curriculum and Code School

Curriculum

  • Desc: The potential hub provides training programs for the founders of startups that might have human capital deficits that will lead to them not being about to adequately implement their ideas.
  • Characteristics:
    • Education that is for a founder (as opposed to code schools which can be for people who just want to join a startup)
      • Code schools are for startup labor supply
    • Active input into a current entrepreneurial endeavor
      • e.g. " The program is designed to augment and support the real-life business experiences that the students are facing every day in their entrepreneurial endeavors"
    • Not an ad hoc session, not a one time meeting but a full "course", evidence of this could be
    • Has evidence of a integrated curriculum leading to a new compentance
    • Has evidence of a set fixed start and end dates that last XXX long
    • Cultivate leadership for entrepreneurs
    • Tagged "Business" as opposed to 'Tech' or 'Design'
    • Is a session linked to others that regularly occurs
  • TBD points
    • Do we care about outsourcing?
  • Potential Turks
    • Search Text: Fullbridge, leadership program, business academy, business course, aspiring entrepreneurs
  1. Google: "Search Text" site:URL
  2. Record 0 if no result returns.
  3. If there is a result, click first link in which result search text appears and record the sentence in which the text appears.

Code School

  • Desc: training programs that teach coding, data processing, webpage building and other technical skills.
  • Characteristics:
    • Bootcamps
    • Target group are the developers or people who want to join the startups but not the founders themselves
    • Scheduled classes, not a one time meeting (as opposed to workshops)
  • TBD points
  • Potential Turks
    • Search Text 1: website design, coding, web development, software, bootcamp
    • Search Text 2: General Assembly, Anyone Can Learn to Code, Umbraco, Designation, Boise CodeWorks, Grand Circus, DevMountain, Silicon Valley Data Academy, Academy Pittsburgh
  1. Google: "Search Text" site:URL
  2. Record 0 if no result returns.
  3. If there is a result, click first link in which result search text appears and record the sentence in which the text appears.


Onsite VC v. Angel Investors

Onsite OH Investors v. mentors

Thoughts (Ariel, 07/20): The names listed on 'mentor' page/sections must all be mentors, and the same applies for investors/OH investors although few companies list their investors. So here the only thing we are trying to differentiate is whether the mentor is a investor. maybe via checking whether they are from a VC firm?? But even they are from VC companies doesn't mean they are going to invest on the startups of the Hubs they are mentoring on. Or another way to think about it is differentiating between mentors/OH mentors. Mentors tend to give the particular startups long term support and available when needed while OH mentors only gives advice on the spot.

Mentors

  • Desc:
  • Characteristics:
    • Focus on improving entrepreneurial community through ongoing, recurring support
    • Help and guide the startups on: business plans and models, management, development, execution, technology innovation, marketing, sales
    • Common fields/occupations: founder/CEO of another company, business development, serial entrepreneur, marketing, sales, management consulting, technology and innovation, research professor etc.
    • Some companies offer mentor office hours
  • TBD Points:


Investors

  • Desc:
  • Characteristics:
    • Focus on investing on early stage or growth stage startups
    • Usually from VC firms
    • Common fields/ occupations: VC firm manager, VC firm partner, fund manager
  • TBD Points:
  • Potential Turks:
  1. Search allintext:"office hours" site:URL
  2. Mark office hours as 1 if there is a result, otherwise mark as 0.
  3. Click on the first five results
  4. On each of the five pages, search for two items:
    1. search for 'mentor'. (Ctrl + F) If 'mentor' appears in the description paragraph of office hours on any of the five pages, mark mentor OH as 1. Otherwise mark as DNE and copy the description paragraph of office hours of all five pages.
    2. search for 'fund'. (Ctrl + F) If 'fund' appears in the description paragraph of office hours on any of the five pages, mark investor OH as 1. Otherwise mark as DNE and copy the description paragraph of office hours of all five pages.

Onsite temporary workshops v. networking events

Temporary Workshops

  • Desc:
  • Characteristics:
    • The purpose is learning and discussing
    • Often have a specific topic: business issue (e.g. online marketing) or techniques learning (e.g. intro to Java script)
    • In the forms of: workshop, class, panel, project, XX session, seminar, series, intro to XX
    • Exception: tech meetup is usually a workshop(e.g. C++ programmer meetup, http://techranchaustin.com/events/)
  • TBD Points:
  • Potential Turks:
    • See Turk for Both Below

Networking Events

  • Desc:
  • Characteristics:
    • The purpose is to meet fellow entrepreneurs and experts and networking with them
    • Focus on experience sharing or communication as opposed to discussing a specific topic or technical subject
    • In the forms of: meetup, networking, happy hour, info session?, luncheon, XX night, socials, talks??, community XX
  • TBD Points:
  • Potential Turks:
    • See Turk for Both Below
  • Turk for Both 1 of 2:
    1. Search the Search Text 1 (allintext: events site: URL) and choose link to "Events", "Calendar", or related. Record 'url' on SOURCE If this does not exist, go to Step 6
    2. For all events that have dates, copy the events from today's date to the following month into ALL EVENTS
    3. For all events that are related to teaching or learning (e.g. contain "Training," "Seminar," "Class," "Learn," "Bootcamp," "Workshop," "Pitch Event", copy the name of the events into WORKSHOPS
    4. For all events that are related to scoial activities and networking (e.g. "Social," "Meet Up," "Breakfast"/"Lunch"/"Happy Hour", "Movie Night"/"Bowling"), copy the name of the events into NETWORKING
    5. If a message explicity says there are no events, mark as 0 for ALL EVENTS, WORKSHOPS, and NETWORKING
    6. If this does not exist, search Search Text 2 (allintext: Company Name site: meetup.com) and click on the meetup.com for the company if it exists. If it does exist, record meetup on SOURCE. If not, go to step 8.
    7. Repeat Steps 2-5.
      1. If this does not exist, search Search Text 3 (allintext: Company Name site: eventbrite.com) and click on the eventbrite.com for the company if it exists. If it does exist, record eventbrite on SOURCE. If not, mark DNE for all variables.
    8. Repeat Steps 2-5.
  • Turk for Both 2 of 2:

Companies Used for Auditing/etc.

Capital Factory, Austin
1871, Chicago
Rocket Space, San Francisco
1776, Washington D.C.
Betamore, Baltimore
Packard Place, Charlotte
The venture Center, Little Rock
GSV Labs, San Francisco
The Hive, Palo Alto
Innovation Pavilion, Denver
OSC Tech Lab, Akron
Speakeasy, Indianapolis
Riverside.io, Riverside
The Salt Mines, Columbus
InNEVation, Las Vegas
804 RVA
Impact Hub, Salt Lake
Awesome Inc, Louisville
Geekdom, San Antonio
Alloy26, Pittsburg
ReSET, Hartford
Ansir Innovation Center, San Diego
Domistation, Tallahassee
Atlanta Tech Village, Atlanta
Spark Labs, New York

Completed Work

Retrieved from "http://www.edegan.com/mediawiki/index.php?title=Hubs:_Hubs_Data&oldid=7352"