How we improved the support service in Yandex.Mail
Yandex.Mail is developing very rapidly: over the past two years, its audience has grown almost one and a half times - to more than 18M people per month. For such growth to be possible, technical development should also go very fast - every week we roll out two releases with improvements and changes.
But in order to make a good service, there is not enough work with the technical side - you need to adjust the organizational part. Including a fast and efficient support service.
Throughout the past year, we have been implementing a project to reorganize it. We want to talk about how we conceived it, what metrics were set and what happened.
How would you explain to your grandmother how to set up an IMAP email client? And try to find out on the phone the reason why the parents do not turn on the computer or this or that site is not loading.
Yandex.Mail user support staff do this for hundreds of different grandmothers, parents and children, and not by telephone or in person, but in a multi-volume email correspondence. We ask clarifying questions, ask for a screenshot, and ask questions again. So we work incognito seven days a week under the pseudonyms Timofei Zhuravlev, Lev Dobrov and Julia Cloud.
Today in the Yandex.Mail and Mail support service for the domain there are 6 people who respond to user letters. We do not have a division into beginners and specialists: everyone can set a task to develop a new feature or come to the head of the service and demand to urgently fix something for users.
The average age of support children is 24 years. Almost all with higher education or still studying. There are those who work remotely. For example, from Kiev and St. Petersburg.
Schedule is standard: five days a week for eight hours. But employee shifts are distributed so that letters are looked after from early morning until late at night seven days a week. Every weekday at the workplace there are from three to five people (depending on the day of the week), and on weekends - two or three. Usually their strength is enough to answer hundreds of letters arriving per day or process almost a thousand if an accident occurs.
This has not always been the case.
A year ago - in the spring of 2012 - the support service consisted of 8 employees working from Monday to Friday; Saturday and Sunday were days off. All the guys answered letters and at the same time tried to solve additional problems. Within 24 hours, we managed to answer only 60-70% of letters, the rest were answered on the second or even third day.
Given the rapid pace of Yandex.Mail development, it becomes obvious that the 2012 support service did not match the volume and scale of the service.
To change the situation, the first thing we needed was to set goals and answer the basic questions: “Why is our support service for our service? After all, there are many projects without it - and nothing. Who needs support more - users or development team? ”
For ourselves, we found this answer: the quality of service directly depends on the quality of the support service. The faster the user receives a response, the more loyal he is to the service, the better he becomes. People want to be heard and help the service team improve it.
As a result, our new indicators were response time to users and regular feedback for the service team.
To achieve the goals, we needed to solve several problems:
As already mentioned, a year ago we answered 60-70% of letters within a day after receiving. Now we have set an ambitious goal - to respond to 100% of letters within a day. But since you just can't take it and do it “tomorrow”, the task was divided into trimesters: by September 1, 2012, the goal was 70% of letters within a day, by December - 80%, and by March 2013 - more than 90%.
Then they drew a graph with a forecast.
To improve the response time, it was necessary to develop a new KPI.
Previously, the KPI of each support was the number of responses per day and, of course, user ratings. The response rate was very poorly correlated with the deadlines: the support team did not understand why they were fulfilling the norm, but the deadlines were still bad. Therefore, the time for the opening of the ticket became the new KPI - that is, at the end of the working day there should not be any letters in the queue that will become daily allowance before the employee arrives at work tomorrow.
In addition to the improved response time with such KPIs, the new norm also had a positive effect on their quality: support no longer chases quantity, but can thoughtfully and unconventionally respond to a user’s letter.
Work monitoring has also changed. Earlier, in real time we saw only the number of answered tickets for each employee, which was not connected with the new KPI. Therefore, we began to daily take statistics on the response time for each queue. This allowed us to better see the problematic lines, exclude natural marriage in statistics and quickly respond to changes in the situation.
We had the usual feedback form, which gave only one opportunity - to write a letter to the Support Service. The simple form generated a huge number of letters with low information content, and, as a result, we had to have 2 support lines in order to somehow structure this flow:
First of all, we redid the feedback structure - it became tree-like. Using hints, the user can detail his problem and at the same time give us the opportunity to put his letter directly in the hands of a second-line specialist. Based on the existing knowledge base on the main problems of users, we have compiled branches, each of which ends with a certain recipe and form for communication, if the proposed solution does not help.
As a result of the transition to such a feedback, it was possible to reduce the daily number of incoming letters by almost 1.5 times - despite the fact that the audience of Yandex in general and Yandex.Mail in particular continues to grow.
After some experiments with the formulations and the visual arrangement of quick recipes, we finally saw that the tickets go to the specialists you need. Plus, we managed to separate questions about Mail from questions about other services.
We updated all the information about common user problems and collected it in one place. In addition, we have appointed a person responsible for the relevance and structure of our knowledge base. We sorted out the tasks for the service, described the rules of interaction with each development department, and also appointed a person responsible for this. Supporters stopped wasting time looking for the right information and updating clusters during business hours, which made it possible to respond to more letters.
Now that the marriage has become at a minimum, the knowledge base was relevant, and the tasks did not go out, we began work on the transition to a seven-day work week.
To do this, as it turned out, is not so simple. You can’t force a person to work seven days a week, since the attentiveness and, as a result, the quality of the response are the first to suffer. Therefore, it was necessary to distribute work schedules so that every day at the workplace there were at least two support (this is enough to parse the queue), but at the same time all the guys worked according to the usual 5/2 schedule. Moreover, employees who can answer the widest range of questions, not only from their field of knowledge, should work on weekends. After a month of experimenting, renegotiating contracts and even hiring new employees, we finally succeeded. The transition to support 7 days a week made it possible to respond to the vast majority of letters within a day,
We have achieved the goal!
According to statistics for March, more than 92% of Yandex.Mail users receive a response within 24 hours after contacting support.
Moreover!
Since the number of letters fell significantly and some time was freed up for us, we decided to spend it profitably and think: what else can be done to optimize our work. Namely, they collected statistics on tags (each letter in mail support is marked with a certain tag, which means its area of problems; for example, “Address Book” or “Downloading HTTPS”), highlighted the most common cases in several directions in mail support at once and set tasks whose solution could reduce the number of letters on these issues.
The challenge for the next few months is to start responding within hours.
But in order to make a good service, there is not enough work with the technical side - you need to adjust the organizational part. Including a fast and efficient support service.
Throughout the past year, we have been implementing a project to reorganize it. We want to talk about how we conceived it, what metrics were set and what happened.
Theory
How would you explain to your grandmother how to set up an IMAP email client? And try to find out on the phone the reason why the parents do not turn on the computer or this or that site is not loading.
Yandex.Mail user support staff do this for hundreds of different grandmothers, parents and children, and not by telephone or in person, but in a multi-volume email correspondence. We ask clarifying questions, ask for a screenshot, and ask questions again. So we work incognito seven days a week under the pseudonyms Timofei Zhuravlev, Lev Dobrov and Julia Cloud.
Today in the Yandex.Mail and Mail support service for the domain there are 6 people who respond to user letters. We do not have a division into beginners and specialists: everyone can set a task to develop a new feature or come to the head of the service and demand to urgently fix something for users.
The average age of support children is 24 years. Almost all with higher education or still studying. There are those who work remotely. For example, from Kiev and St. Petersburg.
Schedule is standard: five days a week for eight hours. But employee shifts are distributed so that letters are looked after from early morning until late at night seven days a week. Every weekday at the workplace there are from three to five people (depending on the day of the week), and on weekends - two or three. Usually their strength is enough to answer hundreds of letters arriving per day or process almost a thousand if an accident occurs.
This has not always been the case.
A year ago - in the spring of 2012 - the support service consisted of 8 employees working from Monday to Friday; Saturday and Sunday were days off. All the guys answered letters and at the same time tried to solve additional problems. Within 24 hours, we managed to answer only 60-70% of letters, the rest were answered on the second or even third day.
Task
Given the rapid pace of Yandex.Mail development, it becomes obvious that the 2012 support service did not match the volume and scale of the service.
To change the situation, the first thing we needed was to set goals and answer the basic questions: “Why is our support service for our service? After all, there are many projects without it - and nothing. Who needs support more - users or development team? ”
For ourselves, we found this answer: the quality of service directly depends on the quality of the support service. The faster the user receives a response, the more loyal he is to the service, the better he becomes. People want to be heard and help the service team improve it.
As a result, our new indicators were response time to users and regular feedback for the service team.
Condition
To achieve the goals, we needed to solve several problems:
- Reduce response time.
- To ensure that people themselves find answers to questions that could be found without the help of a support service representative, and do not write letters on such occasions.
- To increase the information content of complaints and narrow down the problem already at the stage of writing a letter to support (sometimes you have to ask clarifying questions: screenshot, browser, playback conditions).
- Separate letters directly about Mail from letters about other services (Disk, Subscriptions, SDA, etc.).
- Simplify the transfer of knowledge to support desk users.
- Go to the seven-day work week.
Decision
1. Reduce response time
As already mentioned, a year ago we answered 60-70% of letters within a day after receiving. Now we have set an ambitious goal - to respond to 100% of letters within a day. But since you just can't take it and do it “tomorrow”, the task was divided into trimesters: by September 1, 2012, the goal was 70% of letters within a day, by December - 80%, and by March 2013 - more than 90%.
Then they drew a graph with a forecast.
To improve the response time, it was necessary to develop a new KPI.
Previously, the KPI of each support was the number of responses per day and, of course, user ratings. The response rate was very poorly correlated with the deadlines: the support team did not understand why they were fulfilling the norm, but the deadlines were still bad. Therefore, the time for the opening of the ticket became the new KPI - that is, at the end of the working day there should not be any letters in the queue that will become daily allowance before the employee arrives at work tomorrow.
In addition to the improved response time with such KPIs, the new norm also had a positive effect on their quality: support no longer chases quantity, but can thoughtfully and unconventionally respond to a user’s letter.
Work monitoring has also changed. Earlier, in real time we saw only the number of answered tickets for each employee, which was not connected with the new KPI. Therefore, we began to daily take statistics on the response time for each queue. This allowed us to better see the problematic lines, exclude natural marriage in statistics and quickly respond to changes in the situation.
2, 3, 4. We increase the information content of complaints, reduce the number of letters to support and share them by project
We had the usual feedback form, which gave only one opportunity - to write a letter to the Support Service. The simple form generated a huge number of letters with low information content, and, as a result, we had to have 2 support lines in order to somehow structure this flow:
- On the first line, attendants answered the simplest questions and shared letters on projects, each of which described its area of knowledge. For example, "Mail programs" or "Problems downloading Mail."
- Supporters on the second line answered questions in their directions, set tasks for the service and resolved problem situations. Two support lines are almost always a heavy load, which means poor response times and lengthy correspondence.
First of all, we redid the feedback structure - it became tree-like. Using hints, the user can detail his problem and at the same time give us the opportunity to put his letter directly in the hands of a second-line specialist. Based on the existing knowledge base on the main problems of users, we have compiled branches, each of which ends with a certain recipe and form for communication, if the proposed solution does not help.
As a result of the transition to such a feedback, it was possible to reduce the daily number of incoming letters by almost 1.5 times - despite the fact that the audience of Yandex in general and Yandex.Mail in particular continues to grow.
After some experiments with the formulations and the visual arrangement of quick recipes, we finally saw that the tickets go to the specialists you need. Plus, we managed to separate questions about Mail from questions about other services.
5. Unified knowledge base
We updated all the information about common user problems and collected it in one place. In addition, we have appointed a person responsible for the relevance and structure of our knowledge base. We sorted out the tasks for the service, described the rules of interaction with each development department, and also appointed a person responsible for this. Supporters stopped wasting time looking for the right information and updating clusters during business hours, which made it possible to respond to more letters.
6.24/7
Now that the marriage has become at a minimum, the knowledge base was relevant, and the tasks did not go out, we began work on the transition to a seven-day work week.
To do this, as it turned out, is not so simple. You can’t force a person to work seven days a week, since the attentiveness and, as a result, the quality of the response are the first to suffer. Therefore, it was necessary to distribute work schedules so that every day at the workplace there were at least two support (this is enough to parse the queue), but at the same time all the guys worked according to the usual 5/2 schedule. Moreover, employees who can answer the widest range of questions, not only from their field of knowledge, should work on weekends. After a month of experimenting, renegotiating contracts and even hiring new employees, we finally succeeded. The transition to support 7 days a week made it possible to respond to the vast majority of letters within a day,
Result
We have achieved the goal!
According to statistics for March, more than 92% of Yandex.Mail users receive a response within 24 hours after contacting support.
Moreover!
- The daily number of support letters fell 1.5 times compared to the same period last year.
- The number of specialists responsible only for replies to letters has decreased. Some of them have grown into managers and now help their colleagues work better and solve problems faster.
- The information content of the support response, according to the results of user voting, increased by 10 percent (from 75 to 85).
- We have a stock of employees in case of emergency. And this is despite the fact that in a year there have been fewer specialists in the support service.
homework
Since the number of letters fell significantly and some time was freed up for us, we decided to spend it profitably and think: what else can be done to optimize our work. Namely, they collected statistics on tags (each letter in mail support is marked with a certain tag, which means its area of problems; for example, “Address Book” or “Downloading HTTPS”), highlighted the most common cases in several directions in mail support at once and set tasks whose solution could reduce the number of letters on these issues.
The challenge for the next few months is to start responding within hours.