IS434G7 GuessTheWordSG Results

From Social Analytics and Applications
Jump to: navigation, search
GuessTheWordSG Home
GuessTheWordSG Home
GuessTheWordSG Team
GuessTheWordSG Client
GuessTheWordSG Campaign
GuessTheWordSG Results

Past References

After sleepless nights of programming, we managed to launch the GuessTheWord (GTW) Web Portal, that allows the community to create and vote for custom decks. On that same date, GTW's latest Android release was also approved by Android PlayStore. This release consisted of a feature that allowed users to "Pay with a Share".

This timely approval from Android PlayStore served as a perfect synergy to the GTW Web Portal, whereby users would be able to unlock a paid deck by sharing the GTW Web Portal on their Facebook Page. As mentioned in our strategy, this is the main trigger point of our campaign.

Within 24 hours, 186 users were paying with a Facebook Share on their wall. Their friends were now exposed to GuessTheWord. The posts led to an astonishing 4426 reach on GuessTheWord's Facebook page and a spike in downloads to a daily average of 524 (original daily average is 296)..



On the 19th of March, one day after our application launched and was reaching its peak outreach, Facebook sent us a notification informing us that we had violated the Facebook platform policy. We had unintentionally breached the policy in the following ways:

Automatically publishing Stream Stories
When a user activated the pay-by-share feature to unlock a deck, our application would send a share on the user’s behalf to his Facebook wall. We implemented this feature with the idea to make the ‘payment experience’ seamless for the user. However, Facebook does not allow the share feature to be implemented in this way as it requires the user to explicitly press the share button instead.

Prefilling messages for Stream Stories
Again, to make the process seamless, we pre-populated the sharing content to share onto the user’s wall page, directing traffic to our website. However, Facebook also does not allow this feature to be implemented in this way and it requires the user to be able to change the content of his share.

Incentivizing users to share stories
One of the key components of our campaign was to leverage on the paid decks by making them free if they share. The idea was to incentivise users to share the news of the new campaign and direct traffic to the campaign microsite. Facebook does not allow such behaviour.

As a result of our violations, Facebook sent in a notice stating that we had 48 hours to comply with the rules or risk application suspension. Unfortunately, Facebook took action within 3 hours of their initial notice instead, and we were left with very little time to react.

Facebook’s action of suspending our application had dire implications for us. In the long run, it meant that we were unable to collect crucial social data such as the identity of sharers and re-sharers. i.e. we would not be able to track whose shares were the most influential and prompted the most re-shares, even with our website workaround.

However, at that time the more immediate concern was that Facebook’s suspension caused our application to crash every time a user tried to pay by share. This was a very bad experience for the user especially since it was just a day after we released the update and we knew that if we did not act fast, we would lose customer loyalty.

We also had to mirror the fix in the iOS version of the application which in turn reset the approval counter and delayed the iOS version launch further.

Crisis Management
In order to minimize the impact of the disruption, the team immediate proceeded to change the codes in the application. Instead of using Facebook’s SDK for Android which allows us to retrieve vital information from the user through the token exchange process, we moved the actual sharing process to Facebook’s default sharing feature.

In this implementation, the control of the application goes straight to the user’s native Facebook application on his android device, helps the user to share and then passes control back to our application. This allowed us to bypass Facebook’s suspension of our application as this method does not require the identity of our Facebook application to be used. (i.e. Facebook does not know it is our application that is sharing). The limitation to this is that we were unable to capture the user’s permission that will allow us to crawl his friends and his shares of our campaign.

However, given the nature of the problem, we decided to implement it this way to minimise disruption for our users. As a result, the user merely had to update his application to the latest version in order to use the pay-by-share feature properly.

Separately, as were working on the backend fix, we crafted apology posts acknowledging the issue and apologizing for the inconvenience. Later on, when the issue was fixed, we used a Facebook post to advise users to update to the latest version of the application and explained to the users exactly why our application has met with an issue.



Public Apology on Facebook


After due consideration, the team decided to release onto the Facebook page a time-sensitive coupon that entitles users to 50% off the premium bundle. That would mean that users would only be paying $4.99 instead of the current $8.98.

The intention of the coupon code was not only to reward users for voting decks through our Web Portal. Using the element of urgency, some of our users were prompted to take action and make a purchase, which resulted in a spike in average daily in-app purchases to $82.34 (original average daily in-app purchase $37.16).

The purpose of increasing in-app purchases was to convert more non-paying customers to paying customers. Once a user makes an in-app purchase, it becomes easier to convince him to make a subsequent purchase, and this will lead higher revenue in the long run.


Tehkor-segment3-facebook-edm.png Facebook EDM on Good Lobang!

Tehkor-segment3-voting.png Coupon after Voting!


In the 5th day of the campaign, we felt the need to reach out to more than just our circle of friends. We needed the handful of influential players that could get the message out to the masses - the law of the few.

Armed with a customised press-kit, we attached and sent specialised emails to 38 influential bloggers (e.g. Xiaxue, Yutakis) with special coupon codes to try out the game. We also contacted high traffic singapore news sites (e.g. VulcanPost, Mothership, The Straits Times). Within 2 days, 5 out of 38 bloggers responded. Unfortunately, the press kit wasn't very well-received.

  • Jessica, from Tippytapp mentioned that there could be a possibility that she would be helping us spread the word - but could only do so in May.
  • Another blogger declined by saying that she does not play games.
  • The other 3 bloggers belong to agencies and request for payment before they would spread the word to their followers. The fees requested were not within our budget for the campaign. For instance, Yutakis, from quoted $600 for one Instagram post.

We consulted the stakeholders of GTW on the sales quotations and were advised to not take up the offer as GTW is a relatively new startup with budget constraints. The team, however, continued to reach out to other bloggers with follow-up emails but have still not received a response till date.

Fortunately, 4 days after sending out the emails to top new sites, Azhar Aziz from VulcanPost, one of Singapore's high traffic news site, agreed to do a quick interview with GTW! Thrilled, the team informed GTW stakeholders of the great news. Preparation for a news article started, and more details about this will be discussed later on.


Tehkor-segment4-presskit.png Customised PressKit

Tehkor-segment4-email.png Customised Email


On 25 March, GTW’s latest app release was available on Apple Appstore. The long approval time for the Apple AppStore prevented the update from being released at the start of the campaign.

Similar to Android users, iOS users are now able to "Pay with a Facebook Share". iOS users were also able to share via Twitter. This allowed us to track the effectiveness of using Twitter to attract traffic. After the iOS update was released, many unlocked decks with shares and their friends were exposed to the posts. This exposure led to a huge spike in downloads the following day.


Tehkor-segment5-facebook-edm.png Announcement of iOS Release on Facebook


After going through a gamification lecture by Prof Kyong, we decided to increase the stickiness of our campaign through gamification. Hence, we implemented a simple leaderboard which displays the top 5 decks with the most votes at the moment. The leaderboard allows deck creators to know their current standings. Deck creators can know how far they are from winning, which increases their competitiveness and they will find means of driving traffic to their own decks.

Since the leaderboard was implemented, the daily number of votes increased two-fold from 39 to 71 the next day. Within the last 3 days, competition was intense as deck creators were anxious for their victory and we saw a huge increase in votes, moving from an average 39 daily votes to an average of 100 daily votes.



After a week of collaboration with Azhar Aziz from VulcanPost, GuessTheWord was finally featured on VulcanPost! This well written article increased awareness of GTW amongst the public and it attracted a huge spike in Facebook Likes and app downloads.

The next day, the article was also featured on YahooSG which led to more brand exposure, resulting in 1363 downloads within 2 days and a total of 515 Facebook Likes. Through the help of 3 influential news sites, we are able to reach out to a wide group of audience faster than we could when we were targeting existing players and our close friends. That's where we truly understand the Law of the Few.

Tehkor-segment7-chart-fb-likes.png Tehkor-segment7-chart-downloads.png


Now that the campaign has concluded, we will evaluate if our campaign goals have been met.


To raise the brand awareness of GuessTheWord, we have identified 2 key performance indicators - the number of Facebook Page Likes and the number of shares and re-shares on the page posts. Although Facebook Page Likes might be superficial, it is still relatively a good measurement for brand awareness as it shows the number of people who are interested in app updates. On the other hand, the number of shares and re-shares measures the outreach of the brand, as well as determine the popularity of content posted.

Before the start of the campaign, GuessTheWord’s Facebook Page has only 166 Facebook Page Likes. On the average, each post on the Facebook Page has only 3.5 shares per post. Given that there are 100,000 existing downloads, these numbers are rather disappointing. We aimed to increase the number of Facebook Page Likes by 200% and the number of shares and re-share by at least 50%.

After 2 weeks of intense campaigning, we have met the targets that we had set for ourselves. The number of Facebook Likes rocketed from 166 to 515 likes. This increase is mainly attributed to the news coverage on various technology sites such as Vulcan Post and Games in Asia. More significantly, the numbers of shares and re-shares have tripled, increased from 3.5 per post to 11 shares per post.

Before Planned Actual  % Change
No. of Facebook Page Likes 166 489 515 + 210 %
No. of Shares and Re-Shares 3.5 7 11 + 214 %

The number of downloads is clearly an important key metric for our campaign. An increase in number of downloads is a good indication that our campaign managed to acquire new users and reach a wider group of audience. Currently, the average daily downloads have been falling drastically and this trend is worrying. Our group aim to increase average daily downloads by at least 50%.

After 2 weeks of campaign, we fell slightly below the original targets we have set for ourselves. Nevertheless, we have managed to reverse the downward trend, increasing average daily download by 35%. This increase is also largely attributed to the news coverage.

Before Planned Actual  % Change
Average Daily App Downloads* 296 444 398 + 35 %

*Average Daily App Downloads calculated based on 2-week time period from Mar 5 - Mar 18 2014


Since GuessTheWord is a free game, the revenue for the company comes mainly from the in-app purchases. Users can buy premium decks individually at $1.28 or discounted deck bundles at $8.98. According to AppAnnie, the revenue earned from the game has been decreasing. Our group targets to increase daily in-app purchase revenue by 100%.

In the course of campaign, we introduced 50% off discount coupons and this has met with relatively positive results. The discount coupons have encouraged the price-sensitive users to purchase the deck bundles. Although the daily in-app revenue did not meet our set targets, it still managed to increase by a slight 5%, despite slashing 50% off the deck bundles. This indicates that although the profit margin per sale is lower due to the discount, the increase in quantity of purchases made up for it and shows that our campaign has successfully converted non-paying users to paying users.

Before Planned Actual  % Change
Average Daily In-App Purchase Revenue* $37.16 $74.32 $39.12 + 5.27 %

*Average Daily In-App Purchase Revenue calculated based on 2-week time period from Mar 5 – Mar 18 2014


We recognized the fact that GuessTheWord is fundamentally a game requiring a lot of interaction as it can only be played with a minimum of 2 players. Therefore, we want to extend this interaction to beyond the game and have different groups of players interact with each other on a different platform. Hence, we built an external web portal for users to create decks and vote for their favourite community decks. Our group aims to achieve at least 10,000 page views on our web portal within our 2-week campaign.

During our campaign, we held our first community deck challenge and the response was favourable. We had 27 decks submitted, 794 votes casted and successfully met our goals with a web view of 12,392.

Before Planned Actual  % Change
No. of Pageviews on Web Portal No Results 10,000 12,392 -


Tehkor-header-analytics-methodology.png Tehkor-campaign-all-tools.png

In this part of the report, we will explain the tools used to collect and analyze our metrics, as well as the purpose of each metric tracked. We will also be covering the differences between our interim proposal and our actual implementation.


In our interim proposal, our purpose of tracking the “pay with share” actions was to allow us to determine the influential GuessTheWord SG players who could be potential brand ambassadors. Sharing was initially done with the help of Facebook SDK, but due to the complication with Facebook's policy (also further elaborated in the Timeline Section), we had to move away from the use of Facebook’s SDK.

The benefit of using Facebook’s SDK was that it allowed us to retrieve both the permission (access token) and postid after the user has shared the link. This would have allowed us to retrieve and aggregate key metrics on the number of shares/likes/comments of a particular post.

However, as our application had unintentionally violated Facebook policy by incentivizing shares, Facebook blocked our application id that is calling the SDK. As a result, it caused our application to crash. We had to do an immediate fix and find a workaround.

In the end, we had to use the default Facebook share dialog in order to implement the pay-by-share feature. This method gives control to the native Facebook application to complete the sharing process. The difference is that we are not able to retrieve any access token or postid and therefore are blind to the metrics of the post.


Google Analytics provided us with an easy way of tracking information from our web visitors. Besides tracking the Visitor Retention rate, as mentioned in our interim proposal, we also managed to track the following information from Google Analytics:

  1. Entry point
  2. Drop-off and visit duration
  3. Site which referred users to GuessTheWord's Website
  4. iPhone & Android App Usage statistics

Google Analytics for websites is relatively easy to implement. After setting up Google Analytics account and site, we only need to copy a short javascript code to every page of our GuessTheWord website.

Drop-off and Visit Duration

This functions as an indicator of interest for each website page. From the analytics, we found that 30% of the traffic of visitors dropped off after visiting our root site ( This result may indicate that our root site may require improvements on its layout so as to effectively bring users through the intended actions and retain their interest.

iPhone & Android App Usage Statistics

These were tracked by implementing Google Analytics inside the application to measure vital statistics such as usage time, number of instances of application crash as well as the most popular category of decks to see if they have increased as a result of our campaign.


App Annie works as an external analytics tool that provides Google Play Store and Apple App Store related information, such as downloads, app ranking and revenue analytics. These data can be analysed to determine whether our campaign has successfully increased downloads, rankings or in-app purchases. There is no implementation involved on our end as AppAnnie receives their data from public information external to our application. Furthermore, our clients had already set up their revenue account with AppAnnie so we can start retrieving data from AppAnnie immediately.


To measure the effectiveness of each user for spreading content, our team wanted to keep track of both the amount of shares generated by a post and how exactly the shared posts are spread. For example, if a person shared our link, we wanted to know how many people viewed that particular shared post, and then went on to share the link himself.

This was critical to us for 2 reasons:

  1. Gauge how much reach our campaign has generated
  2. By the law of the few, determine our most effective brand ambassadors

Unfortunately, Facebook does not natively allow us to track such information. This is because for every user who uses our application, we required his explicit permission (in the form of access token exchange) to identify him and get metrics on his posts such as likes or shares.

However, even if the user gives us the permission, we would not have the opportunity to request for permission from the other users who re-shares his post as it does not go through our application. Therefore, there is currently no way to get enough information to draw a network graph connecting users and shares.


To work around this limitation, we implemented a special functionality in our web application to track the shares. For every user that logs into our website or to our android/iphone application, we extract his facebook id and append it to the link in the user’s pre-populated share dialog. For example (assuming his id is 12345), instead of sharing, the user will share <12345><12345>. On the server side, we record and process the facebook id into our database.

This helped us in 2 ways:

  • Track Share Count
    • Making use of Facebook’s FQL link_stat table, we are able to keep track of the number of users entering the website as a result of a particular user’s shared link .
  • Tracking Share Footprint
    • We designed it in this manner to keep track of the sharing footprint. For example, if someone saw the link<12345> on his timeline and decided to click it, she would enter the website and we will record his visit. If he decides to share the same link, our website would generate<12345>&id?=<45678> for him to share.
    • Thus, if we query how many users shared/liked/entered the link, we would know that they have gotten the reference from user <45678> who in turn got the reference from user <12345>. We would then be able to map out our sharing tree and also detect users with high centrality. For example, if we realize that most traffic to our site comes from<12345>, we would know that user <12345> is an influential user and he is valuable to us.


For all the other information not tracked by Google analytics or App Annie, we keep track of such information in our server for future analysis using FQL and R. Some of the information we track are as follows.

Deck Information

These information are stored in our database when the decks are created. Some analysis that can be derived from the information about the decks are as follows:

  1. Users who voted on the deck, frequency of votes and the popularity of the deck. Since social competitions such as this are usually popularity contests, this information is critical in identifying people with high centrality and know a lot of people who will vote for them.
  2. To identify who is genuinely interested in the concept of the GuessTheWord campaign. We identify this group based on their willingness to vote for other decks despite having created their own decks.
  3. Tehkor-custom-tracking-votes.png
    Find out how the top 3 winners won the challenge. We wanted to see if any one of the winners actually won because they returned constantly to vote for their own decks. Interestingly, the person who voted for himself the most was only in fourth place whereas the deck creator in second place did not even vote for herself at all. Upon further analysis, we then discovered through google analytics and custom tracking tools, that her success was due to the effectiveness of her friends’ facebook shares which had many re-shares and a high clickthrough rate.


Coupons system creates and monitors the usage of coupons. We issued unique coupon codes to each blogger in the press kit. Using this tool, we can track if any particular blogger had used the given coupon. Hence, we will be able to evaluate the effectiveness of giving coupon codes to bloggers. Users also received a coupon code upon voting, which would enable them to purchase a combo deal at 50% of the price. We were also able to determine if the 50% coupon code was attractive enough to encourage users to make in-app purchases.


We created a multi-purpose keyhash value pair table and many web services that will save web requests into the database. One of the use of this tracking tool is to track any url with postfix extensions, e.g. We regularly posted posters on GuessTheWord’s Facebook page and we wanted to track each individual facebook post and evaluate the click through rate of each poster. The keyhash allowed us to determine the facebook post that garnered the highest traffic to the site.


Degree and Betweenness Centrality

To analyze our GuessTheWord audiences, we gathered our GuessTheWord Facebook page likers and commentors as well as Facebook users from GuessTheWord competition website. Using their Facebook id and their connection, we created an adjacency matrix and using the matrix we generated a network graph in Gephi to visualize it. We also used the adjacency matrix to calculate the degree and betweenness centrality using R.

Top Degree Score Betweenness Score
1 Lee.Xiangrui 229 Lee.Xiangrui 2322.8777
2 Jo.Essential 183 Kean.Kwok 867.5259
3 Kelvin.Tay 183 Zixuan.Lee 796.8208
4 Zixuan.Lee 183 Kelvin.Tay 690.3218
5 El.malique.Ghazali 176 Jo.Essential 601.4254
6 Lim.Yi.Sheng 176 Ng.Mel.Ling 534.7045
7 Ng.Mei.Ling 176 El.malique.Ghazali 411.9666
8 Kean.Kwok 174 Lim.Yi.Sheng 411.9666
9 Elizabeth.Chua 169 Law.Yifu 262.6205
10 Lisa.Tan 169 Elizabeth.Chua 219.5612

The main difference in this network graph is that the data is based on existing GuessTheWord users and we do not have their friends who are not involved with GuessTheWord.

In this scenario, GuessTheWord would consider users with higher degree and betweenness centrality to be more valuable as brand ambassador. Users with higher degree centrality have more friends who have interacted with GuessTheWord. The reason why they have so many friends on GuessTheWord might be because they convinced their friends to join. They would be likely to be more receptive to becoming our brand ambassador and will have an easier time to convince his other friends to become GuessTheWord users. Users with high betweenness have friends from differnt social groups playing GuessTheWord. They are likely to be abe to help us spread the game to different groups of people so as to increase our outreach. Users with high betweenness are similar that it also have a lot of friends interacting with GuessTheWord but have lesser common friends if multiple ambassadors are selected.

Share Influence Index

In our previous analysis of betweenness centrality and degree centrality, we measured how influential a person is within our network. This was good to identify individuals who had a greater outreach to a greater number of people. However, the data we gathered was not weighted and we would not know the extent at which an individual was influential. An individual like Xiang Rui may have a high degree centrality but the extent of his influence may not be enough to prompt his connections take an additional action such as a like or re-share. For example, Xiang Rui has a high degree centrality but only 3 people re-shared or liked his post. This can be compared to Scott Ng who may have a relatively lower degree centrality but garnered 20 people to re-share or like his post.

For these reasons, our team wanted to identify individuals who were more influential than others and who were able to prompt others to re-share our advertisement. The mechanism of how we collected this data has been covered in the Analytics Tracking section previously where we talked about the use of Facebook Query Language. Below are the results of our analysis.


The value of the Share Influence Index represents the total number of likes and shares garnered from each user’s shared post.

Unfortunately, due to Facebook suspending our application, we only managed to secure data from the start of the campaign to the day it was suspended. (18-March to 19-March). However, even with the limited data, we are immediately able to see who has a higher Share Influence Index than others. Excluding the co-owners of the company and the project team, the top 5 individuals with a high Share Influence Index is Kelvin Toh, Dwight Lee, Danial Sim, Marcus Lee and Liong Koon Awyeong. After further investigation, we realised that these individuals are not connected to the project team in any way and are purely customers of the application outside of SMU.

Our results show that these individuals are genuinely interested in the Guess The Word application and are good contenders for the company to focus on in the future as Brand Ambassadors. By leveraging on them, the company would be able to more effectively extend its impression and outreach outside its current customer base.



From the overview we can see a huge increase in new user logins, especially from the android platform. Another interesting metric to note is the 11.8k impressions generated throughout the lifetime of our campaign


Based on the demographic data we retrieved, we realised that majority of our users are females and there are almost twice as many females as per males in the ages of 18-24. This could potentially be useful for the company to segment the market in the future.

The Sprint!

Out of 784 votes made in total on all GuessTheWord decks, 433 votes occurred on the last 4 days of the competition, 27th March 2014. One of the possible explanation for this occurrence could be our newly implemented leaderboard on 27th March 2014. We believe that this leaderboard made participants or participants-to-be more competitive after viewing the top five current leaders of the competition, thereby pushing them to drive more traffic to their deck.


Targeting Second-Tier Bloggers

We might consider targeting second-tier bloggers for free advertising features. We contacted 38 influential bloggers and only 5 replied, most of which were asking for fees to be paid for social media features. As many of the influential bloggers operated under agencies and charged for advertisement, it was difficult to get their help with our lack of budget.

Much Earlier Marketing Efforts

User engagement started rising in the second week of the campaign, when the campaign was already half complete. If time permitted, we would have started marketing efforts a week before the start of the campaign so that there will be more people attracted.


Press Releases

The company might want to consider targeting second-tier bloggers instead of the well-known bloggers. We found out that most famous bloggers are managed by advertising agencies and they charge a rather exorbitant price for writing an article or sharing news on their social networks. The second-tier bloggers are more open to collaboration and are more willingly to do free coverage.

The company should also target journalists and technology sites. This is because they are more open to publish new articles. Also, sending a press release is not a one-time thing. Building a relationship with the press is a long-term activity. Nectary Studio should keep the press updated with their PR-worthy events and upcoming releases.

Compulsory Facebook Login

Moving forward, we will like to suggest to Nectary Studio to incorporate compulsory Facebook login. With users’ data, they are able to make better business decisions and have the capability to understand the behavior of their users. They can also make use of the data to identify users with high centrality and engage them as brand ambassadors to promote the game.

Launch in Other Markets

From our campaign, our group felt that the game might have reach its maximum market potential in Singapore. Nectary Studio should consider launching other localized versions of GuessTheWord in other markets like Malaysia, Japan or Korea. Particularly Korea and Japan, these markets have high smartphone penetration and people are more willingly to spend on in-app purchases according to a recent report by Tech-In-Asia.


Social Media Campaign - Not an Easy Feat

From this campaign, our group learnt that running a social media campaign is not easy at all. Very often, things do not always go as smoothly as what we have envisioned. Hence, it is important to monitor your various social media channels so that we can respond to a crisis quickly and correctly. When Facebook banned our application due to a violation of their policy, the game crashed on many Android users. This is an unexpected crisis that our group did not foresee and would have been disastrous if we did not monitor the campaign actively.

Also, our group learnt that a good social media campaign does not necessarily require a big budget. As seen from our campaign, we managed to achieve a satisfactory result with no marketing dollars at all. With a bit of ingenuity and knowledge, we can create a viral social media campaign with little or no budget. We just have to be more resourceful, or shameless.