What resources make or would make the user-to-user support in your language more effective?

We already saw the average responses to this question in one of our previous post in the series. Here there are again:

As you can see, resources based on text were ranked higher than any others. On the other hand, Instant Messaging or Tutorial videos were ranked much lower, ending slightly above ‘Neutral’. It might be interesting to see what impact on these results had the profile of the survey takers. Thanks to the demographic variable that we discussed previously, we were able to learn how each profile responded to the question.

The above chart shows how many respondents of each profile took the survey. 19% of the survey takers were end-users and 32% were community members actively contributing to the project. Is it possible that the responses given by advanced users (who were more numerous) outweigh those of end-users in this question? What preferences do the end-users have and, consequently, what is the best way to improve their support experience?

In order to compare the average responses between the user profiles, we reduced the number of analyzed variables from 10 to 3, by means of factor analysis. The new extracted variables were:

• Textual support, representing:

Tutorials/How-to’s
Troubleshooting articles
Forums

• Visual support, representing:

Screenshots
Screencasts/videos

• Interactive support, representing:

Live Chat
IM
VoIP

Average values of these new variables are visualized in the following chart.

We clearly notice that the end-users are much less enthusiastic towards textual support, and in the same time they prefer interactive help slightly more than other users. The visual support has been similarly rated across all user profiles (the differences weren’t statistically significant—hence the grey color in the chart).

In summary, thanks to the factor analysis, it was possible to make the comparison of preferences between user profiles easier. It turned out that even though the interactive resources, such as Instant Messaging and LiveChat, were ranked lowest by all respondents, there were significant differences between user profiles in the responses. We learned that the end-users favor textual support much less than other groups and show a slight preference of the interactive support.

Good support is an essential part of the product experience in its holistic approach. It is even more important for end-users who may be even using the product for the first time. This is why it was important for us to see what this particular group though about Mozilla support experience. Our study may serve to confirm that indeed the introduction of LiveChat on support.mozilla.com was a valuable addition which will help users looking for support.

* * *

This is our last post in the series about the SUMO survey. The series included:

• the first post, in which we presented the objectives of the survey and the questions,
• the second post, in which we discussed who the survey takers where and what languages they spoke,
• the third post, in which we analyzed the overall average responses and dove into the differences between the user profiles,
• the fourth post, in which we focused on the differences between locales,
• and this post, the fifth one, in which we looked closer at the responses to the last questions of the survey.

We hope you liked the series and found the data as fascinating as we did. We encourage you to post your thoughts in the comments or join us on #surveys on irc.mozilla.org.

In our series about the support experience survey, we already discussed who the survey takers were, as well as their average responses to all the questions (and the differences between user profiles). Today, it’s time to look into the differences between locales.

Question 1: Support experience rating

As we have often mentioned previously, doing an analysis of the differences between locales can be tricky. There are some cultural factors that can play significant role in how people from each locale respond to a question. We will not, however, explore these factors now. We will stick to the data we have from the survey, but keep in mind that some differences may be cultural ones.

This is how John Williams, who offered his help with the analyses we’re doing, explains this using the example of the first question of the support survey (”How would you rate the support that community provides for Mozilla products in your language?”):

The table below shows how many times each locale rated a product the lowest, and how many time each rated a product the highest. Note that because of the overlapping feature of multiple comparisons some the counts do not sum to the number of products.

It seems clear from this summary that German and French respondents generally rate the products lower than do respondents from other locales; conversely Polish and Italian respondents rate the products higher.

Let’s take a look at the average responses to this question:

You can certainly see that Firefox support has been rated the highest. There are two observation to be made here:

• Firefox is the main Mozilla product, so its support resources are the most developed
• Firefox is also the most known Mozilla product, and as such, many users are familiar with its support resources. There were very little missing values (’I don’t know’ responses) among the responses to the Firefox rating question

We also notice (just as John predicted) that the average responses to all three questions were generally lower for such locales as French and German, and higher for Hungarian, Italian and Polish.

The average responses to this question continue in the next chart.

Generally, the responses to these products were lower than to Firefox and Thunderbird. It is still interesting to see that the support for add-ons has been rated well (slightly below 4 for all locales), similar to Mozilla overall and Thunderbird from the first chart. And even if Calendar and SeaMonkey support are rated lowest in the chart, it’s important to remember that in general, all products have been rated on the positive side of neutral (above 3).

In this chart we’re showing the average responses for a product for each locale. You may remember that we also asked about Camino, which is not shown in the chart. This is because Camino had a lot of missing values (almost 75% of all responses) and the differences in responses for it between locales are not significant.  If you’re interested in the overall average across all locales, see our previous post about the SUMO survey.

Question 2: Characterizing support

Nine statements in our survey asked survey takers to characterize the support experience, which translated into nine variables. With nine different locales to compare, this gives us 81 figures to analyze, which was too many!  In order to make comparisons easier, we first looked at two variables covering quality of localization and then, we reduced the others to one variable that covered quality of content.

Localization of the support resources

We asked two questions related to the quality and sufficiency of localized support resources:

• Most of the relevant information can be found in my language
• Support in English is sufficient

The average responses to these two questions are show in the chart below:

Although statistically different, the average responses are very similar for all locales: in general, the survey takers agreed that most of the relevant information can be found in their language, as well as disagreed that support in English is sufficient (in the chart, value 2 means ‘Disagree’).

Support experience assessment

6 variables related to the quality of content were strongly correlated which allowed us to replace them with a single variable using the factor analysis. These six variables were:

• Most of the relevant information can be found in my locale
• Navigation on the support sites is easy
• The support sites are up-to-date
• Most common issues are covered well
• The support sites load quickly
• New issues are quickly addressed

We extracted one new factor (one new variable summarizing all six others) with similar loadings from all 6 variables. Consequently, we named it “support experience assessment” and used it to easily compare the support experience between locales.

Again, we see differences between locales that are most probably due to cultural factors. It is still interesting, however, to see that Italian and Russian locales score highest, whereas Polish, Portuguese and German score lower than other analyzed locales. We should also notice that in the beginning of the post, we included the Polish locale to the group of the ‘eager’ ones, so probably its low score in the last chart should be investigated in more detail.

What’s next?

In the next and last post about the SUMO survey, we will take a closer look at the last question, in which we asked what resources would make the user-to-user support more effective.

For this post, we will start looking into the responses to our questions. Remember that we asked the following:

1. Please rate the overall quality of the support for each product;
2. How much does the user agree or disagree with a list of statements describing the support experience;
3. Please rate the usefulness of particular content types (tutorials, screencasts, chats, etc.).

We can start by summarizing the means to each answer, typically the easiest statistic to interpret and visualize.  But, as a forewarning, the chart below is pretty long!

Darker bars are the mean responses and the thin vertical grey lines represent the levels of the scale for each question:

• ‘Lowest quality’ to ‘Highest quality’ for the first question
• ‘Strongly Disagree’, ‘Disagree’, ‘Neutral’, ‘Agree’ and ‘Strongly Agree’ for the remaining two questions

Perhaps it’s not surprising, but Firefox’s support was rated highest (our guess: because of its depth and completeness), whereas products like SeaMonkey, Calendar and Camino had a very high ratio of missing values (the ‘I don’t know’ responses).  As the support rating lowers for each product, the percentage of the missing values increases, which also constitutes an important piece of information.  60-70% of the survey takers didn’t know how to rate the quality of support for SeaMonkey or Calendar—probably because they hadn’t used support or the application itself.  For Firefox, we can say that its support was rated between 4 and 5, where 5 was “highest quality”.  NB, only about 6% of the survey takers didn’t know how to rate Firefox support.   Please keep in mind that every respondent was thinking about support in their language, so in our next post we will calculate the mean responses to this question for every locale.

For the next two questions, we should interpret the bars differently.  Each of the thin vertical grey lines corresponds to a level starting with ‘Strongly Disagree’ on the left, through ‘Disagree’, ‘Neutral’, ‘Agree’ and, for the rightmost line, ‘Strongly Agree’.  The length of the darker bar for the first statement means that, again on average, the respondents tended to agree that “the support sites loaded quickly.” As for the brighter bar, we interpret it in the same way as above: about 10% of the survey takers didn’t know how to react to this statement.

When you look at the average responses to the second question, you’ll see some interesting findings:

1. “Support in English is sufficient” scores lowest (between ‘disagree’ and ‘neutral’), and at the same time, it has very few missing values.  So, people certainly had an opinion on this one since there were not many “Don’t Know” responses.
2. Two to three other statements have neutral response means (the corresponding dark bars end close to the middle of the chart)
3. Six statements show that the survey takers well above the midpoint and reflecting agreement.

Is this helpful data if we see mostly “Agrees”, “Neutrals” and one “Disagree”?

Keep in mind that every person who responded to these two questions reflected on their unique support experience and then submitted their response. Also, remember that support of the Mozilla products

• can be accessed in different languages (and the quality of the support depends on quality of the localization),
• is often decentralized (each local community has its own discussion board),
• and aims to satisfy each users distinct needs.

If the cumulative responses were “Agree”, the means may still differ across certain groups of people. We can see that the “Most of the relevant support can be found in my language” statement scores high (respondents tended to agree with it), which is a good first step toward understanding how the local communities are doing in terms of the user-to-user support.  One interesting thing to note is the “The support sites make use of video” statement.  It scores low, but also has an outstanding ratio of missing values.  Could you argue that an “I don’t know” answer is close to “Disagree”?  If someone hadn’t ever seen a video on a support site, but was careful in their judgments and supposed that there might be other support sites where videocasts were available, they were probably more likely to answer “I don’t know” instead of “Disagree”.  In this survey, we assumed that an “I don’t know” answer equaled “I haven’t ever seen any of them”.  Therefore, we decided that the survey takers who described themselves as end-users were more likely to give an “I don’t know” answer to this statement.

Mean responses by user profile

Thanks to the user profile variable, we were ablt to split the cases and plot new charts for 3 different user profiles. These were community members (49% of respondents), active community members (localizing, coding, etc.; 32%) and end-users (19% of the survey takers).

We observe the biggest discrepancies between two extreme profiles: end-users and active contributors. The community member profile seems to be in the middle for most of the questions.   For example, end-users rate the quality of the support for each product poorer than the contributors (green bars are longer than the yellow ones). More visible differences can also be observed in the “Most of the relevant support can be found in my language” statement. It seems that active contributors are more enthusiastic here.   Why is this the case?  Perhaps, it’s due to prior knowledge about the products (and computers in general), greater familiarity with the technology, and a better command of English.  What do you think?

Two other statements bring forward even bigger differences in the way the end-users and the active contributors responded.

• “The support sites make use of video”

Why do end-users agree more than active contributors that there are support sites using video?  This is probably a result of what we have discussed above: respondents might have preferred the “I don’t know” answer to this question over “Strongly disagree” or “Disagree”, even if these ones would have been more accurate. We can see that there were nearly 50% of end-users answering “I don’t know” to this statement, and only half as many of the active contributors. Some end-users did agree on this topic, but some didn’t and we can assume that some of them selected the “I don’t know” answers instead of “Disagree”. Using this theory, we can understand why end-users scored this higher: there were not enough “disagree” responses to counterbalance the “agree” ones.

• “Users rarely ask the same question more than once”

It seems understandable that active contributors are more likely to disagree. After all, they’d probably seen lots of duplicated questions in the forums and Bugzilla.

In the third question we see that end-users tend to favor direct communication (IM, VoIP) over text-based communication, compared to other user profiles. We also notice that forums turned out to be one of the most favorite channels of communication for active contributors—and much less so for the other profiles.

Finally, here is the graph of the missing values across different user profiles.

* * *
In the next posts we will reduce the number of variables by means of the factor analysis. This will allow us to better synthesize the analysis without loosing much of the information enclosed in the original set of variables.

Thanks for reading this to the end!!

Over the course of three weeks, 1,558 people took the second community survey.  We published the survey in 16 languages, thanks to the work of the localizers from mozilla-europe.org.   This step was critical for the survey because we wanted to see how the user-to-user support was doing in each local community and to what extent English resources were useful to each community.

Below is the frequency of survey takers across the various languages in which we offered the survey.

For five of the languages (English, German, French, Portuguese and Spanish) we asked the survey takers to indicate their country of origin, since those languages are spoken in multiple countries. Here’s that breakdown:

Some initial comments:

• As you can see, 60% of the people who took the survey in English came from non-English speaking countries. We’ve done a pretty good job localizing this survey, but there are obviously people from other countries who speak other languages who are interested interested in the project.
• More than a half of surveys taken in Spanish were taken in the Latin and South America.
• France and the Germany dominated their locales.
• Look at the global participation!  We want to extend these surveys to Southeast Asia, Japan, and China…working with Gen and Mozilla China to do that this summer.

(Originally, the Community Survey project was targeted the European communities (as Staś’s internship project at Mozilla Europe). We quickly shifted to a more global scope, and we were thrilled to see so many people from other parts of the world expressing their interest in the project and the survey. Thanks for your support!)

Here is another interesting question:  how many surveys were taken each day when the survey was open?  See the graph below:

As you can see, the peak happened two days after the launch of the survey. Quickly afterwards the number of surveys taken daily stabilized at around 50-60, but you can still see the Christmas and the New Year’s eve when we got the least responses. We hit a second burst two days before the end of the survey due to a friendly reminder posted on our blogs.

Demographics

This survey was the first one to feature a demographic question, a lesson learned from last time. Below is an illustration of the frequency of various responses to our question, “Who are you?”:

As you can see, we presented seven answers to the demographic question and allowed for multiple responses. To avoid difficulties in the analysis, we then aggregated this data into three separate variables:

• End-users — those who responded that they were users of Mozilla products, but didn’t select any other answer
• Community members — those who responded that they used Mozilla products and followed the news on the Mozilla project
• Active community members — those who respond that they were localizers, developers, users helping other users, and authors of the support articles for Mozilla products.

Keep in mind that creating these groups was an arbitrary decision we made to help with analysis. We assumed that people who selected “I follow the news on Mozilla” were part of what we call the “Mozilla community” or considered themselves members of this community.  What do you think? Should we simply ask another question “Do you consider yourself a member of the Mozilla community”? And, what about all those people who follow the news Mozilla just because they follow IT-related or Internet-related news in general? These are the kind of questions we face every time when working on our surveys. Feel free to join the discussion.

Below is a look at the number of survey takers in each of these new buckets that we created.

19% of the survey takers selected the ‘I am a user of a Mozilla product’ option as their only answer (the end-user profile), and 49% additionally responded that they followed the news on the project (the community member profile). Therefore, 68% of the survey takers (19% + 49%) don’t localize, don’t develop, don’t write documentation and don’t consider themselves actively participating in the forums. In fact, we reached a very wide group of people who feel connected somehow to the Mozilla project and were able to learn their opinions on support.

Finally, let’s take a look at the structure of the user profiles among the survey takers for each locale.

And the composition of locales by percentage…

The important next step, which we will start to discuss in the next post, is to look at how each question was answered by these profiles.  It should be pretty interesting to see how end-users, community members, and active community members are answering support-related questions across locales.

Thanks for your comments and reading this.  More to come on the second survey soon.

Where did we get this idea?

The topic for this survey came from Jesper Kristensen, a Danish localizer, who participated in a discussion on the Mozilla Europe’s mailing list about new survey ideas.  He wrote:

“SUMO (support.mozilla.com) is a great project, and it is probably going to support localization. But the project relies on volunteers from the local communities. I think it would be interesting to know if the local communities are interested in participating in that project. [...] David and Axel speculates if support of Firefox only vs. all Mozilla will influence how much volunteers will contribute to the project. Maybe a survey could be used to find the answer.”

We took this idea and expanded it to include Mozilla support in general.  It made most sense to then contact David Tenser, the project manager of SUMO, and invite him to help us with the objectives and questions.  Kindly, he agreed to help.

Preparing the questions

Just as we did in the first survey, we kept the questions short to help increase responses rates while still providing valuable information to us.  After a couple of long calls and e-mails, we agreed on a set of objectives listed below:

1. To understand and measure how Mozilla is doing at support in the different locales;
2. To get a better understanding on people’s perception of the quality of support;
3. To find out how can we make Mozilla support more useful.

We asked four questions in total, three of which directly related to our objectives.

In the first question,we asked the survey takers to gauge overall support on different Mozilla products.  In order to identify support-related issues in local communities, we provided a list of possible experiences for people to rate on a scale from one to five.

Question 1: How would you rate the support that community provides for Mozilla products in your language? (1- lowest quality, 5- highest quality)

• Mozilla overall
• Firefox
• Thunderbird
• Calendar
• SeaMonkey
• Camino
• Add-ons

Question 2: How accurate are the following statements for your community? (1 – least accurate, 5 – most accurate)

In my community…

• most of the relevant information can be found in my locale
• the support sites have a structure that is easy to read and navigate
• users ask questions in the right channels
• users rarely ask the same question more than once
• the support sites are up-to-date
• the support sites cover well the most common issues
• the support sites load quickly in my browser
• the support sites make use of screencasts and tutorial videos to help our users
• new issues/problems with Mozilla software are quickly added to support sites

Unlike last time, we made sure to include the “Don’t Know” option.  When we closed the survey, we turned up with a few missing values in our results, but it was still interesting to see how they were distributed, and how they could be interpreted.

For the third question, we asked about how support content might be delivered.

Question 3: What resources make or would make the user-to-user support in your language more effective?

• Tutorials/How-to’s
• Troubleshooting articles
• Support available in my language
• Support available from a single site
• Screenshots
• Screencasts/videos
• Specific versions of support depending on operating system (Windows/Mac/Linux)
• Forums
• Live Chat

And, you’ll remember the lesson learned from last survey:  Next time, ask a demographic question!  We did this time.  Without a specific target for this survey or control of who would take it, we could discover a bit more about who took the survey by asking the following:

Question 4: What’s your involvement in the Mozilla project:

[  ]  I’m a user of a Mozilla product
[  ]  I follow the news from the project
[  ]  I help other users in the discussion boards
[  ]  I write documentation
[  ]  I localize documentation
[  ]  I localize Mozilla products
[  ]  I develop Mozilla code

We then pushed these questions live in Stas’s survey tool and started to collect answers.  What do you think?  Do you think the questions are composed in a way that will elicit information that will meet our objectives?  Your feedback is appreciated.

By the way, a couple of weeks ago, Seth described on his blog the process of creating the second community survey. If you’re interested in a peek behind the scenes of the Community Surveys project, you might find his post interesting.

In closing, we realize this is coming a bit late, but with all the things we learned about survey analysis and presenting the results, time was not lost. Now that we know a few more of the challenges we may encounter, we can publish much more quickly.  A big thank you to everyone for supporting our efforts and keeping us motivated as you have done so far.  We hope it’s been enjoyable and informative.