Collaboration in the cloud at Google

Collaboration in the Cloud at Google

Yunting Sun, Diane Lambert, Makoto Uchida, Nicolas Remy

Google Inc.

January 8, 2014

Abstract

Through a detailed analysis of logs of activity forall Google employees1, this paper shows how theGoogle Docs suite (documents, spreadsheets andslides) enables and increases collaboration withinGoogle. In particular, visualization and analysisof the evolution of Google’s collaboration net-work show that new employees2, have startedcollaborating more quickly and with more peopleas usage of Docs has grown. Over the last twoyears, the percentage of new employees who col-laborate on Docs per month has risen from 70%to 90% and the percentage who collaborate withmore than two people has doubled from 35% to70%. Moreover, the culture of collaboration hasbecome more open, with public sharing withinGoogle overtaking private sharing.

1 Introduction

Google Docs is a cloud productivity suite and itis designed to make collaboration easy and nat-ural, regardless of whether users are in the sameor different locations, working at the same or dif-ferent times, or working on desktops or mobiledevices. Edits and comments on the documentare displayed as they are made, even if many peo-ple are simultaneously writing and commentingon or viewing the document. Comments enablereal-time discussion and feedback on the docu-

ment, without changing the document itself. Au-thors are notified when a new comment is madeor replied to, and authors can continue a con-versation by replying to the comment, or endthe discussion by resolving it, or re-start the dis-cussion by re-opening a closed discussion stream.Because documents are stored in the cloud, userscan access any document they own or that hasbeen shared with them anywhere, any time andon any device. The question is whether this en-riched model of collaboration matters?

There have been a few previous qualitative anal-yses of the effects of Google Docs on collabora-tion. For example, the review of Google Docs in[1] suggested that its features should improve col-laboration and productivity among college stu-dents. A technical report [2] from the Universityof Southern Queensland, Australia argued thatGoogle Docs can overcome barriers to usabilitysuch as difficulty of installation and documentversion control and help resolve conflicts amongco-authors of research papers. There has alsobeen at least one rigorous study of the effect ofGoogle Docs on collaboration. Blau and Caspi[3] ran a small experiment that was designed tocompare collaboration on writing documents tomerely sharing documents. In their experiment,118 undergraduate students of the Open Uni-versity of Israel were randomized to one of fivegroups in which they shared their written assign-ments and received feedback from other studentsto varying degrees, ranging from keeping texts

1Full-time Google employees, excluding interns, part-times, vendors, etc2Full-time employees who have joined Google for less than 90 days

1

2 COLLABORATION VISUALIZATION

private to allowing in-text suggestions or allow-ing in-text edits. None of the students had usedGoogle Docs previously. The authors found thatonly students in the collaboration group per-ceived the quality of their final document to behigher after receiving feedback, and students inall groups thought that collaboration improvesdocuments.

This paper takes a different approach, and looksfor the effects of collaboration on a large, diverseorganization with thousands of users over a muchlonger period of time. The first part of the paperdescribes some of the contexts in which GoogleDocs is used for collaboration, and the secondpart analyzes how collaboration has evolved overthe last two years.

2 Collaboration Visualization

2.1 The Data

This section introduces a way to visualize theevents during a collaboration and some simplestatistics that summarize how widespread col-laboration using Google Docs is at Google. Thegraphics and metrics are based on the view, editand comment actions of all full-time employeeson tens of thousands of documents created inApril 2013.

2.2 A Simple Example

To start, a document with three collaboratorsAdam (A), Bryant (B) and Catherine (C) isshown in Figure 1. The horizontal axis repre-sents time during the collaboration. The verti-cal axis is broken into three regions representingviewing, editing and commenting. Each contrib-utor is assigned a color. A box with the con-tributor’s color is drawn in any time interval inwhich the contributor was active, at a verticalposition that indicates what the user was doingin that time interval. This allows us to see whencontributors were active and how often they con-tributed to the document. Stacking the boxes al-lows us to show when contributors were acting at

the same time. Only time intervals in which atleast one contributor was active are shown, andgaps in time that are shorter than a thresholdare ignored. Gray vertical bars of fixed widthare used to represent periods of no activity thatare longer than the threshold. In this paper, thethreshold is set to be 12 hours in all examples.

In Figure 1, an interval represents an hour.Adam and Bryant edited the document togetherduring the hour of 10 AM May 4 and Bryantedited alone in the following hour. The collab-oration paused for 8 days and resumed duringthe hour of 2 pm on May 12. Adam, Bryant andCatherine all viewed the document during thathour. Catherine commented on the documentin the next hour. Altogether, the collaborationhad two active sessions, with a pause of 8 daysbetween them.

Figure 1: This figure shows an example of thecollaboration visualization technique. Each coloredblock except the gray one represents an hour and thegray one represents a period of no activity. The Yaxis is the number of users for each action type. Thisdocument has three contributors, each assigned a dif-ferent color.

Although we have used color to represent col-laborators here, we could instead use color torepresent the locations of the collaborators, theirorganizations, or other variables. Examples withdifferent colorings are given in Sections 2.5 and2.6.

2 Google Inc.

2 COLLABORATION VISUALIZATION 2.3 Collaboration Metrics

2.3 Collaboration Metrics

To estimate the percentage of users who concur-rently edit a document and the percentage ofdocuments which had concurrent editing, we dis-cretize the timestamps of editing actions into 15minute intervals and consider editing actions bydifferent contributors in the same 15 minute in-terval to be concurrent. Two users who edit thesame document but always more than 15 minutesapart would not be considered as concurrent, al-though they would still be considered collabora-tors. Edge cases in which two collaborators editthe same document within 15 minutes of eachother but in two adjacent 15 minute intervalswould not be counted as concurrent events.

The choice of 15 minutes is arbitrary; however,metrics based on a 15 minute discretization anda 5 minute discretization are little different. Thechoice of 15 minute intervals makes computationfaster. A more accurate approach would be tolook for sequences of editing actions by differ-ent users with gaps below 15 minutes, but that

requires considerably more computing.

2.4 Collaborative Editing

Collaborative editing is common at Google. 53%of the documents that were created and sharedin April 2013 were edited by more than one em-ployee, and half of those had at least one concur-rent editing session in the following six months.Looking at employees instead of documents, 80%of the employees who edited any document con-tributed content to a document owned by othersand 65% participated in at least one 15 minuteconcurrent editing session in April 2013. Concur-rent editing is sticky, in the sense that 76% of theemployees who participate in a 15 minute con-current editing session in April will do so againthe following month.

There are many use cases for collaborative edit-ing, including weekly reports, design documents,and coding interviews. The following three plotsshow an example of each of these use cases.

Figure 2: Collaboration activity on a design document. The X axis is time in hours and the Y axis is thenumber of users for each action type. The document was mainly edited by 3 employees, commented on by18 and viewed by 50+.

Google Inc. 3

2.5 Commenting 2 COLLABORATION VISUALIZATION

Figure 2 shows the life of a design document cre-ated by engineers. The X axis is time in hoursand the Y axis is the number of employees work-ing on the document for each action type. Thedocument was mainly edited by three employ-ees, commented on by 18 employees and viewedby more than 50 employees from three major lo-cations. This document was completed withintwo weeks and viewed many times in the subse-quent month. Design documents are common atGoogle, and they typically have many contribu-tors.

Figure 3 shows the life of a weekly report doc-ument. Each bar represents a day and the Yaxis is the number of employees who edited andviewed the document in a day. This documenthas the following submission rules:

• Wednesday, AM: Reminder for submissions

• Wednesday, PM: All teams submit updates

• Thursday, AM: Document is locked

The activities on the document exhibit a pro-nounced weekly pattern that mirrors the sub-mission rules. Weekly reports and meeting notesthat are updated regularly are often used by em-ployees to keep everyone up-to-date as projectsprogress.

Figure 3: Collaboration on a weekly report. TheX axis is time in days and the Y axis is the numberof users for each action type. The activities exhibita pronounced weekly pattern and reflect the submis-sion rules of the document.

Finally, Figure 4 shows the life of a documentused in an interview. The X axis represents timein minutes. The document was prepared by a re-cruiter and then viewed by an engineer. At thebeginning of the interview, the engineer editedthe document and the candidate then wrote codein the document. The engineer was able to watchthe candidate typing. At the end of the inter-view, the candidate’s access to the document wasrevoked so no further change could be made, andthe document was reviewed by the engineer. Col-laborative editing allows the coding interview totake place remotely, and it is an integral part ofinterviews for software engineers at Google.

Figure 4: The activity on a phone interview docu-ment. The X axis is time in minutes and the Y axisis the number of users for each action type. The en-gineer was able to watch the candidate typing on thedocument during a remote interview.

2.5 Commenting

Commenting is common at Google. 30% of thedocuments created in April 2013 that are sharedreceived comments within six months of creation.57% of the employees who used Google Docs inApril commented at least once in April, and 80%of the users who commented in April commentedagain in the following month.

4 Google Inc.

2 COLLABORATION VISUALIZATION 2.6 Collaboration Across Sites

Figure 5: Commenting and editing on a design document. The X axis is time in hours and the Y axisis the number of user actions for each user location. There are four user actions, each assigned a differentcolor. Timestamps are in Pacific time.

Figure 5 shows the life of a design document.Here color represents the type of user action (cre-ate a comment, reply to a comment, resolve acomment and edit the document), and the Y axisis split into two locations. The document waswritten by one engineering team and reviewedby another. The review team used commentingto raise many questions, which the engineeringteam resolved over the next few days. Collabora-tors were located in London, UK and MountainView, California, with a nine hour time zone dif-ference, so the two teams were almost ”takingturns” working on the document (timestampsare in Pacific time). There are many similarcommunication patterns between engineers viacommenting to ask questions, have discussionsand suggest modifications.

2.6 Collaboration Across Sites

Employees use the Docs suite to collaborate withcolleagues across the world, as Figure 6 shows.In that figure, employees working from nine lo-cations in eight countries across the globe con-tributed to a document that was written within aweek. The document was either viewed or editedwith gaps of less than 12 hours (the threshold forsuppressing gaps in the plot) in the first sevendays as people worked in their local timezones.After final changes were made to the document,it was reviewed by people in Dublin, MountainView, and New York.

Figure 7 shows one month of global collabora-tions for full-time employees using Google Docs.The blue dots show the locations of the employ-ees and a line connects two locations if a docu-ment is created in one location and viewed in theother. The warmer the color of the line, movingfrom green to red, the more documents sharedbetween the two locations.

Google Inc. 5

2.6 Collaboration Across Sites 2 COLLABORATION VISUALIZATION

Figure 6: Activity on a document. Each user location is assigned a different color. The X axis is time inhours and the Y axis is the number of locations for each action type. Users from nine different locationscontributed to the document.

Figure 7: Global collaboration on Docs. The blue dots are locations and the dots are connected if there iscollaboration on Google Docs between the two locations.

6 Google Inc.

3 THE EVOLUTION OF COLLABORATION 2.7 Cross Device Work

2.7 Cross Device Work

The advantage of cloud-based software and stor-age is that a document can be accessed from anydevice. Figure 8 shows one employee’s visits toa document from multiple devices and locations.When the employee was in Paris, a desktop orlaptop was used during working hours and a mo-bile device during non-working hours. Appar-ently, the employee traveled to Aix-En-Provenceon August 18. On August 18 and the first part ofAugust 19, the employee continued working onthe same document from a mobile device whileon the move.

Figure 8: Visits to a document by one user workingon multiple devices and from multiple locations.

Not surprisingly, the pattern of working on desk-tops or laptops during working hours and on mo-bile devices out of business hours holds generallyat Google, as Figure 9 shows. The day of weekis shown on the X axis and hour of day in lo-cal time on the Y axis. Each pixel is coloredaccording to the average number of employeesworking in Google Docs in a day of week andtime of day slot, with brighter colors represent-

ing higher numbers. Pixel values are normalizedwithin each plot separately. Desktop and lap-top usage of Google Docs peaks during conven-tional working hours (9:00 AM to 11:00 AM and1:00 PM to 5:00 PM), while mobile device usagepeaks during conventional commuting and otherout-of-office hours (7:00 AM to 9:00 AM and 6:00PM to 8:00 PM).

Figure 9: The average number of active users work-ing in Google Docs in each day of week and time ofday slot. The X axis is day of the week and the Yaxis is time of the day in local time. Desktop/Laptopusage peaks during working hours while mobile usagepeaks at out-of-office working hours.

3 The Evolution of Collabora-tion

3.1 The Data

This section explores changes in the usage ofGoogle Docs over time. Section 2 defined collab-orators as users who edited or commented on thesame document and used logs of employee edit-ing, viewing and commenting actions to describecollaboration within Google. This section definescollaborators differently using metadata on doc-uments. Metadata is much less rich than theevent history logs used in Section 2, but meta-data is retained for a much longer period of time.

Document metadata includes the document cre-ation time and the last time that the document

Google Inc. 7

3.2 Collaboration for New Employees 3 THE EVOLUTION OF COLLABORATION

was accessed, but no other information about itsrevision history. However, the metadata does in-clude the identification numbers for employeeswho have subscribed to the document, where asubscriber is anyone who has permission to view,edit or comment on a document and who hasviewed the document at least once. Here we usemetadata on documents, slides and spreadsheets.

We call two employees collaborators (or subscrip-tion collaborators to be clear) if one is a sub-scriber to a document owned by the other andhas viewed the document at least once and thedocument has fewer than 20 subscribers. Theowner of the document is said to have sharedthe document with the subscriber. The num-ber of subscribers is capped at 20 to avoid over-counting collaborators. The more subscribersthe document has, the less likely it is that allthe subscribers contributed to the document.

There is no timestamp for when the employeesubscribed to the document in the metadata, sothe exact time of the collaboration is not known.Instead, the document creation time, which isknown, is taken to be the time of the collabora-tion. An analysis (not shown here) of the eventhistory data discussed in Section 2 showed thatmost collaborators join a collaboration soon af-ter a document is created, so taking collabora-tion time to be document creation time is notunreasonable. To make this assumption evenmore tenable, we exclude documents for whichthe time of the last view, comment or edit is morethan six months after the document was created.This section uses metadata on documents cre-ated between January 1, 2011 and March 31,2013. We say that two employees had a subscrip-tion collaboration in July if they collaborated ona document that was created in July.

3.2 Collaboration for New Employees

Here we define the new employees for a givenmonth to be all the employees who joined Googleno more than 90 days before the beginning ofthe month and started using Google Docs inthe given month. For example, employees callednew in the month of January 2011 must have

joined Google no more than 90 days before Jan-uary 1, 2011 and used Google Docs in January2011. Each month can include different employ-ees. New employees are said to share a documentif they own a document that someone else sub-scribed to, whether or not the person subscribedto the document is a new employee. Similarly, anew employee is counted as a subscriber, regard-less of the tenure of the document creator.

Figure 10 shows that collaboration among newemployees has increased since 2011. Over thelast two years, subscribing has risen from 55% to85%, sharing has risen from 30% to 50%, and thefraction of users who either share or subscribehas risen from 70% to 90%. In other words, newemployees are collaborating earlier in their ca-reer, so there is a faster ramp-up and easier ac-cess to collective knowledge.

Figure 10: This figure shows the percentage of newemployees who share, subscribe to others’ documentsand either share or subscribe in each one-month pe-riod over the last two years.

Not only do new employees start collaboratingmore often (as measured by subscription andsharing), they also collaborate with more people.Figure 11 shows the percentage of new employ-ees with at least a given number of collabora-tors by month. For example, the percentage of

8 Google Inc.

3 THE EVOLUTION OF COLLABORATION 3.3 Collaboration in Sales and Marketing

new employees with at least three subscriptioncollaborators was 35% in January 2011 (the bot-tom red curve) and 70% in March 2013 (the topblue curve), a doubling over two years. It is in-teresting that the curves hardly cross each otherand the curves for the farthest back months liebelow those for recent months, suggesting thatthere has been steady growth in the number ofsubscription collaborators per new employee overthis period.

Figure 11: This figure shows the proportion of newemployees who have at least a given number of col-laborators in each one-month period. Each period isassigned a different color. The cooler the color of thecurve, moving from red to blue, the more recent themonth. The legend only shows the labels for a subsetof curves. The percentage of new employees who haveat least three collaborators has doubled from 35% to70%.

To present the data in Figure 11 in another way,Table 1 shows percentiles of the distribution ofthe number of subscription collaborators per newemployee using Google Docs in January 2011 andin January 2013. For example, the lowest 25% ofnew employees using Google Docs had no suchcollaborators in January 2011 and two such col-laborators in January 2013.

25% 50% 75% 90% 95%

January 2011 0 1 4 7 11

January 2013 2 5 10 17 22

Table 1: This table shows the percentile of numberof collaborators a new employee have in January 2011and January 2013. The entire distribution shifts tothe right.

3.3 Collaboration in Sales and Mar-keting

Section 3.2 compared new employees who joinedGoogle in different months. This section followscurrent employees in Sales and Marketing whojoined Google before January 1, 2011. That is,the previous section considered changes in newemployee behavior over time and this sectionconsiders changes in behavior for a fixed set ofemployees over time. We only analyze subscrip-tion collaborations among this fixed set of em-ployees and collaborations with employees notin this set are excluded.

Figure 12: This figure shows the percentage of cur-rent employees in Sales and Marketing who have atleast a given number of collaborators in each one-month period.

Figure 12 shows the percentage of current em-ployees in Sales and Marketing who have at least

Google Inc. 9

3.4 Collaboration Between Organizations 3 THE EVOLUTION OF COLLABORATION

a given number of collaborators at several timesin the past. There we see that more employeesare sharing and subscribing over time becausethe fraction of the group with at least one sub-scription collaborator has increased from 80%to 95%. And the fraction of the group withat least three subscription collaborators has in-creased from 50% to 80%. It shows that many ofthe employees who used to have no or very fewsubscription collaborators have migrated to hav-ing multiple subscription collaborators. In otherwords, the distribution of number of subscrip-tion collaborators for employees who have beenin Sales and Marketing since January 1, 2011 hasshifted right over time, which implies that collab-oration in that group of employees has increasedover time.

Finally, the number of documents shared by theemployees who have been in Sales and Marketingat Google since January 1, 2011 has nearly dou-bled over the last two years. Figure 13 shows thenumber of shared documents normalized by thenumber of shared documents in January, 2011.

Figure 13: This figure shows the number of shareddocuments created by employees in Sales and Market-ing each month normalized by the number of shareddocuments in January 2011. The number has almostdoubled over the last two years.

3.4 Collaboration Between Organiza-tions

Collaboration between organizations has in-creased over time. To show that, we considerhundreds of employees in nine teams within theSales and Marketing group and the Engineer-ing and Product Management group who joinedGoogle before January 1, 2011, were still activein March 31, 2013 and used Google Docs in thatperiod. Figure 14 represents the Engineering andProduct Management employees as red dots andthe Sales and Marketing employees as blue dots.The same dots are included in all three plotsin Figure 14 because the employees included inthis analysis do not change. A line connects twodots if the two employees had at least one sub-scription collaboration in the month shown. Thedenser the lines in the graph, the more collabora-tion, and the more lines connecting red and bluedots, the more collaboration between organiza-tions. Clearly, subscription collaboration has in-creased both within and across organizations inthe past two years. Moreover, the network showsmore pronounced communities (groups of con-nected dots) over time. Although there are nineindividual teams, there seems to be only threemajor communities in the network. Figure 14indicates that teams can work closely with eachother even though they belong to separate de-partments.

We also sampled 187 teams within the Sales andMarketing group and the Engineering and Prod-uct Management group. Figure 15 representsteams in Engineering and Product Managementas red dots and teams in Sales and Marketingas blue dots. Two dots are connected if the twoteams had a least one subscription collaborationbetween their members in the month. Figure15 shows that the collaboration between thoseteams has increased and the interaction betweenthe two organizations has becomed stronger overthe past two years.

10 Google Inc.

3 THE EVOLUTION OF COLLABORATION 3.4 Collaboration Between Organizations

Figure 14: An example of collaboration across orga-nizations. Red dots represent employees in Engineer-ing and Product Management and blue dots representemployees in Sales and Marketing

Figure 15: An example of collaboration betweenteams. Red dots represent teams in Engineering andProduct Management and blue dots represent teamsin Sales and Marketing

Google Inc. 11

3.5 Cultural Changes in Collaboration 4 CONCLUSIONS

3.5 Cultural Changes in Collabora-tion

Google Docs allows users to specify the accesslevel (visibility) of their documents. The de-fault access level in Google Docs is private, whichmeans that only the user who created the docu-ment or the current owner of the document canview it. Employees can change the access level ona document they own and allow more people toaccess it. For example, the document owner canspecify particular employees who are allowed toaccess the document, or the owner can mark thedocument as public within Google, in which caseany employee can access the document. Clearly,not all documents created in Google can be vis-ible to everyone at Google, but the more docu-ments are widely shared, the more open the en-vironment is to collaboration.

Figure 16: This figure shows the percentage ofshared documents that are ”public within Google”created in each month. Public sharing is overtakingprivate sharing at Google.

Figure 16 shows the percentage of shared doc-uments in Google created each month betweenJanuary 1, 2012 and March 31, 2013 that arepublic within Google. The red line, which is acurve fit to the data to smooth out variability,shows that the percentage has increased about

12% from 48% to 54% in the last year alone. Inthat sense, the culture of sharing is changing inGoogle from private sharing to public sharing.

4 Conclusions

We have examined how Google employees collab-orate with Docs and how that collaboration hasevolved using logs of user activity and documentmetadata. To show the current usage of Docs inGoogle, we have developed a visualization tech-nique for the revision history of a document andanalyzed key features in Docs such as collab-orative editing, commenting, access from any-where and on any device. To show the evolutionof collaboration in the cloud, we have analyzednew employees and a fixed group of employeesin Sales and Marketing, and computed collabo-ration network statistics each month. We findthat employees are engaged in using the Docssuite, and collaboration has grown rapidly overthe last two years.

It would also be interesting to conduct a similaranalysis for other enterprises and see how long itwould take them to reach the benchmark Googlehas set for collaboration on Docs. Not only hasthe collaboration on Docs changed at Google,the number of emails, comments on G+, calen-der meetings between people who work togetherhas also had significant changes over the past fewyears. How those changes reinforce each otherover time would also be an interesting topic tostudy.

Acknowledgements

We would like to thank Ariel Kern for herinsights about collaboration on Google Docs,Penny Chu and Tony Fagan for their encour-agement and support and many thanks to JimKoehler for his constructive feedback.

12 Google Inc.

REFERENCES REFERENCES

References

[1] Dan R. Herrick (2009). Google this!: usingGoogle apps for collaboration and productiv-ity. Proceeding of the ACM SIGUCCS fallconference (pp. 55-64).

[2] Stijn Dekeyser, Richard Watson (2009). Ex-tending Google Docs to Collaborate on Re-

search Papers. Technical Report, The Uni-versity of Southern Queensland, Australia.

[3] Ina Blau, Avner Caspi (2009). What Typeof Collaboration Helps? Psychological Own-ership, Perceived Learning and OutcomeQuality of Collaboration Using Google Docs.Learning in the technological era: Proceed-ings of the Chais conference on instructionaltechnologies research (pp. 48-55).

Google Inc. 13