How I did an English simulator, which is used not only by my mother
I will tell you how we invented, made and developed the service of mini-lessons of the English language, which is used by 4,000 people every day. For the seed, the growth in the number of active users:
Studying English at the institute is remembered for daily homework (otherwise, the tail, and three tails already threaten with expulsion) and the control at each lesson. I tried to prepare by textbooks - everyone praised the blue book of Murphy, but I quickly abandoned it because of a simple problem - I was tired of constantly climbing to the end of the book to check myself.
The problem of convenience is solved by online services, but they have no feedback. If I make a mistake, at best I will see the message "You were mistaken, you can continue or re-read the materials." And I want to quickly get an explanation and move forward.
Here I will also add a constant lack of time - it seems that it is difficult to set aside 30 minutes or an hour for classes, and I want to fill up the time between meetings, traveling in transport or having a lunch break. I’m not talking about independent study (I don’t believe in it in principle, you need to learn the first foreign language with the teacher), but about “homework”. Even if you study 3 times a week with a teacher, there is a lot of time between classes that would be good to use.
')
Total: I want to learn English words and grammar with explanations and examples in my free time.
A few years ago, we would have immediately rushed to program and tell our friends what a cool thing was invented. Now only tell friends. Hearing from friends about the same problems in classes on textbooks or in online services, it was decided to take the next step - to assemble a prototype, spending a minimum of effort and money.
Googling "landing page template for education" and after reviewing 200 options, we chose one, on which we simply edited the texts and deleted the extra blocks. Even the name of the service - Educa - remains from this template. We offered to subscribe to the test mailing list for the program (which was copied from Murphy) for 99 rubles per month.
In order not to cut the backend, the mailing list was simply made through Mailchimp Automation - after registration, the user received an email with a link to the test with explanations. The test was made on the Gistroll platform (our test platform). To receive payments screwed Yandex.Cash.
All took 2 days. We bought ads on Facebook and sat down to wait.
During the week, 790 people registered, of which 130 paid for the subscription. In the next 30 days, my daily morning rituals added the creation of an English test for mailing and constant correspondence with users - why they paid / didn’t pay for the subscription, what they like / don’t like, what to add, and so on. I managed to interview more than half of the subscribers and the following picture was formed:
The main result of the experiment - 60% of subscribers have fully completed the monthly course and wanted to study further. Inspired by such results, we decided to make a product, but again - with minimal effort. I must say that for two we have quite a decent experience in web development (more than 12 years) and the first version of the application we collected from scratch on Ruby on Rails in a week. After a simple registration (even without entering a password), an endless test immediately began, made up of all the materials that were ready by that time.
Particular attention in this minimum version of the product we paid viral mechanics. I subscribed to the growthhackers.com newsletter and I once remembered an article about how 140,000 people registered on a half-empty page of a non-existent product (supposedly the simplest task manager in the world). It was decided to simply repeat the mechanics: to give a subscription for a year to any user who brought 5 friends and send an encouraging Email after each registration by the link. Invented - done. The task took about an hour.
We also paid attention to the analyst. The product is no longer just a mailing list with clear metrics that Mailchimp perfectly shows. On the one hand, we constantly communicate with users (a couple of hours a day) and try to understand their emotions, hidden intentions, in order to take all this into account in the product. But it is also important to build statistics - whether we are growing or not (DAU, the number of users who pass at least one lesson per day), whether our users like the product (retention, number of lessons per week from the user), whether we spend money on wasted advertising (ROI, whether the investment in advertising pays off), whether users are using new features that appear in the service (for example, how many people use the pronunciation of words, go through the lessons again, study particular topics, and do not work on the program). It seems that you can use free solutions (Google Analytics or Yandex Metrika), but there are restrictions on the number of possible types of events that do not suit us. There are all sorts of services like Mixpanel, but 200,000 free events are spent on us in a couple of days, and we obviously will not pull the paid version. As a result, everything was reduced to logging events to a separate server (part JS from the client, part in the backend) and SQL queries. It's easy, cheap, and from the console you can get any statistics:
The beginning of February 2015. After a week of insane coding, we slept, looked at what had been a fresh look and gave ourselves the go-ahead to launch - we made a new group on Facebook and started a lot of advertising campaigns. We, of course, are still marketers and, perhaps, we approach the matter without a soul - we constantly test different banners: with pictures / colored background, with buttons / without buttons, with even font / handwriting, etc. Every day we download campaign statistics from Facebook, combine them with the data in our database, and look at the ROI (payback) chart. If it starts to fall - turn off the campaign. It seems that this approach helps to spend money on advertising effectively.
Viral mechanics also worked, although we did not break the record of 140,000 people. It turned out that users from ads and users from sheros for a free account are different:
Indeed, the invited friends are less active, but this is the effect of a “free” subscription for a year. However, 14% of our user base is invited friends. Invited users also invite friends (we have friends of friends of friends of friends of friends of friends).
Every day we registered 400-500 users, but many simply fell off. Metrics metrics, but from communication with users, it became clear that we have either very simple materials or very complex. Thus, we got to only 20% of users who continued to take lessons every day.
It was necessary to somehow properly organize our materials and we asked for help from a graduate / graduate student of the Institute of Foreign Languages, who was also a very experienced teacher. It turned out that there is a European system of levels of English , so we divided our materials into levels and individual courses in grammar / vocabulary / vocabulary. At each course we made a program. In order to check the desire of users to take courses that are not yet open and not to lose them, we have made it possible to subscribe to a notification about the appearance of the course. This brought an excellent result - a large number of users signed up for 3 courses at once, and we not only collected an impressive list for distribution, but also checked the demand for future courses.
It’s simply impossible to make so many materials ourselves, so with the money we’ve accumulated we hired two Methodists who were assigned to do several lessons a day. We are trying to automate the methodologists as much as possible, but for now our passion for automation is breaking on the blank wall of human factors. Explanations for the exercises should be written in human language, so each is read several times, and when published on the site, we monitor whether they read it and how often they click on the “I do not understand” button. If the explanation reaches the critical level - rewrite.
Now 4000 people are engaged daily. Already 243 people have completed a course in beginner grammar and received certificates of passing. Our most important indicators are the number of lessons passed per person and the conversion into payment, which we are constantly increasing. We now have an average of 25 lessons per person, and the conversion to pay after the launch of the program and the separation of courses has grown by more than 50%.
About mom - it was not a joke, she works as a programmer and for the last 5 years I want to make her pull up English. I offered her various textbooks, downloaded serials with subtitles, but it didn’t work until we did our service.
On her example, you can look at the learning statistics - how erroneous questions are corrected and new rules are learned in the work on errors:
The last task we worked on is the teacher’s office, where, based on the statistics collected, we show the user's progress and the topics that need to be analyzed in the lesson. The idea came to us from both sides - the teachers, who invited their students to the service, wanted to see how they were doing, and the parents of schoolchildren write that they want to show the mistakes to the teachers:
Another important task is to analyze the data of 800,000 lessons that users have already passed (and we collect a lot of statistics) and take this into account in our algorithm for compiling lessons. Now they are compiled in such a way that one lesson takes 5-10 minutes. It consists of three parts - repetition, work on the bugs and new materials. Thus, for each user, we want to find the optimal complexity of classes.
I must say that all our ideas come from users - we categorize absolutely all the “desires” and collect them in a table in order to see the “trend” and do the most important tasks. We are asked to make courses for another 7 languages, keyboard control, “leaderboards” and much more.
It seems that our experiment can be called successful. We "invented" the format of questions with explanations and turned out a product that users like. We finished the grammar course for beginners on our own (we called for help a student of the Moscow State Linguistic University). Now we need to do more courses - in vocabulary, grammar, listening, some users even ask us to do courses for children. The money that the service brings is enough to pay for the work of one methodologist, but at this rate we will make all the materials only in a year. Therefore, we decided to ask for help from our users and their friends to launch a crowdfunding campaign.
To launch a full-fledged campaign on KickStarter or its analogues for a long time, especially since we will see real money in a few months and only part of it. Therefore, it was decided to make your KickStarter withblackjack preference and courtesans:
I must say that our current users have very coolly perceived our offer. We were literally flooded with letters with words of support, offers to help, and most importantly, many people supported us by purchasing a paid subscription to themselves or friends.
It seems that you can not stop and continue to write for many more hours - as we made our vocabulary from 15 million sentences and use it to compile examples, about a bunch of A / B tests that we did, about new features, over which we now We work, about participation in Tolstoy Startup Camp (without which nothing would have happened). But let's leave it for now for future articles and questions.
Try the Educa class and share your impressions: educa.io
I would be very grateful if you support the creation of new materials on Educa or tell your friends about the service: educa.io/starter
The idea of ​​an English simulator
Studying English at the institute is remembered for daily homework (otherwise, the tail, and three tails already threaten with expulsion) and the control at each lesson. I tried to prepare by textbooks - everyone praised the blue book of Murphy, but I quickly abandoned it because of a simple problem - I was tired of constantly climbing to the end of the book to check myself.
The problem of convenience is solved by online services, but they have no feedback. If I make a mistake, at best I will see the message "You were mistaken, you can continue or re-read the materials." And I want to quickly get an explanation and move forward.
Here I will also add a constant lack of time - it seems that it is difficult to set aside 30 minutes or an hour for classes, and I want to fill up the time between meetings, traveling in transport or having a lunch break. I’m not talking about independent study (I don’t believe in it in principle, you need to learn the first foreign language with the teacher), but about “homework”. Even if you study 3 times a week with a teacher, there is a lot of time between classes that would be good to use.
')
Total: I want to learn English words and grammar with explanations and examples in my free time.
Idea check
A few years ago, we would have immediately rushed to program and tell our friends what a cool thing was invented. Now only tell friends. Hearing from friends about the same problems in classes on textbooks or in online services, it was decided to take the next step - to assemble a prototype, spending a minimum of effort and money.
Googling "landing page template for education" and after reviewing 200 options, we chose one, on which we simply edited the texts and deleted the extra blocks. Even the name of the service - Educa - remains from this template. We offered to subscribe to the test mailing list for the program (which was copied from Murphy) for 99 rubles per month.
In order not to cut the backend, the mailing list was simply made through Mailchimp Automation - after registration, the user received an email with a link to the test with explanations. The test was made on the Gistroll platform (our test platform). To receive payments screwed Yandex.Cash.
All took 2 days. We bought ads on Facebook and sat down to wait.
Experimental results
During the week, 790 people registered, of which 130 paid for the subscription. In the next 30 days, my daily morning rituals added the creation of an English test for mailing and constant correspondence with users - why they paid / didn’t pay for the subscription, what they like / don’t like, what to add, and so on. I managed to interview more than half of the subscribers and the following picture was formed:
- all 20+ years
- everyone likes that the course has a program
- everyone likes that one lesson takes 5-10 minutes and you can get it over your morning coffee
- everyone is just delighted that every exercise has an example and an explanation
Product development
The main result of the experiment - 60% of subscribers have fully completed the monthly course and wanted to study further. Inspired by such results, we decided to make a product, but again - with minimal effort. I must say that for two we have quite a decent experience in web development (more than 12 years) and the first version of the application we collected from scratch on Ruby on Rails in a week. After a simple registration (even without entering a password), an endless test immediately began, made up of all the materials that were ready by that time.
Particular attention in this minimum version of the product we paid viral mechanics. I subscribed to the growthhackers.com newsletter and I once remembered an article about how 140,000 people registered on a half-empty page of a non-existent product (supposedly the simplest task manager in the world). It was decided to simply repeat the mechanics: to give a subscription for a year to any user who brought 5 friends and send an encouraging Email after each registration by the link. Invented - done. The task took about an hour.
We also paid attention to the analyst. The product is no longer just a mailing list with clear metrics that Mailchimp perfectly shows. On the one hand, we constantly communicate with users (a couple of hours a day) and try to understand their emotions, hidden intentions, in order to take all this into account in the product. But it is also important to build statistics - whether we are growing or not (DAU, the number of users who pass at least one lesson per day), whether our users like the product (retention, number of lessons per week from the user), whether we spend money on wasted advertising (ROI, whether the investment in advertising pays off), whether users are using new features that appear in the service (for example, how many people use the pronunciation of words, go through the lessons again, study particular topics, and do not work on the program). It seems that you can use free solutions (Google Analytics or Yandex Metrika), but there are restrictions on the number of possible types of events that do not suit us. There are all sorts of services like Mixpanel, but 200,000 free events are spent on us in a couple of days, and we obviously will not pull the paid version. As a result, everything was reduced to logging events to a separate server (part JS from the client, part in the backend) and SQL queries. It's easy, cheap, and from the console you can get any statistics:
Launch
The beginning of February 2015. After a week of insane coding, we slept, looked at what had been a fresh look and gave ourselves the go-ahead to launch - we made a new group on Facebook and started a lot of advertising campaigns. We, of course, are still marketers and, perhaps, we approach the matter without a soul - we constantly test different banners: with pictures / colored background, with buttons / without buttons, with even font / handwriting, etc. Every day we download campaign statistics from Facebook, combine them with the data in our database, and look at the ROI (payback) chart. If it starts to fall - turn off the campaign. It seems that this approach helps to spend money on advertising effectively.
Viral mechanics also worked, although we did not break the record of 140,000 people. It turned out that users from ads and users from sheros for a free account are different:
Indeed, the invited friends are less active, but this is the effect of a “free” subscription for a year. However, 14% of our user base is invited friends. Invited users also invite friends (we have friends of friends of friends of friends of friends of friends).
Every day we registered 400-500 users, but many simply fell off. Metrics metrics, but from communication with users, it became clear that we have either very simple materials or very complex. Thus, we got to only 20% of users who continued to take lessons every day.
It was necessary to somehow properly organize our materials and we asked for help from a graduate / graduate student of the Institute of Foreign Languages, who was also a very experienced teacher. It turned out that there is a European system of levels of English , so we divided our materials into levels and individual courses in grammar / vocabulary / vocabulary. At each course we made a program. In order to check the desire of users to take courses that are not yet open and not to lose them, we have made it possible to subscribe to a notification about the appearance of the course. This brought an excellent result - a large number of users signed up for 3 courses at once, and we not only collected an impressive list for distribution, but also checked the demand for future courses.
It’s simply impossible to make so many materials ourselves, so with the money we’ve accumulated we hired two Methodists who were assigned to do several lessons a day. We are trying to automate the methodologists as much as possible, but for now our passion for automation is breaking on the blank wall of human factors. Explanations for the exercises should be written in human language, so each is read several times, and when published on the site, we monitor whether they read it and how often they click on the “I do not understand” button. If the explanation reaches the critical level - rewrite.
Product now
Now 4000 people are engaged daily. Already 243 people have completed a course in beginner grammar and received certificates of passing. Our most important indicators are the number of lessons passed per person and the conversion into payment, which we are constantly increasing. We now have an average of 25 lessons per person, and the conversion to pay after the launch of the program and the separation of courses has grown by more than 50%.
About mom
About mom - it was not a joke, she works as a programmer and for the last 5 years I want to make her pull up English. I offered her various textbooks, downloaded serials with subtitles, but it didn’t work until we did our service.
On her example, you can look at the learning statistics - how erroneous questions are corrected and new rules are learned in the work on errors:
Plans
The last task we worked on is the teacher’s office, where, based on the statistics collected, we show the user's progress and the topics that need to be analyzed in the lesson. The idea came to us from both sides - the teachers, who invited their students to the service, wanted to see how they were doing, and the parents of schoolchildren write that they want to show the mistakes to the teachers:
Another important task is to analyze the data of 800,000 lessons that users have already passed (and we collect a lot of statistics) and take this into account in our algorithm for compiling lessons. Now they are compiled in such a way that one lesson takes 5-10 minutes. It consists of three parts - repetition, work on the bugs and new materials. Thus, for each user, we want to find the optimal complexity of classes.
I must say that all our ideas come from users - we categorize absolutely all the “desires” and collect them in a table in order to see the “trend” and do the most important tasks. We are asked to make courses for another 7 languages, keyboard control, “leaderboards” and much more.
“Kickstarter”
It seems that our experiment can be called successful. We "invented" the format of questions with explanations and turned out a product that users like. We finished the grammar course for beginners on our own (we called for help a student of the Moscow State Linguistic University). Now we need to do more courses - in vocabulary, grammar, listening, some users even ask us to do courses for children. The money that the service brings is enough to pay for the work of one methodologist, but at this rate we will make all the materials only in a year. Therefore, we decided to ask for help from our users and their friends to launch a crowdfunding campaign.
To launch a full-fledged campaign on KickStarter or its analogues for a long time, especially since we will see real money in a few months and only part of it. Therefore, it was decided to make your KickStarter with
I must say that our current users have very coolly perceived our offer. We were literally flooded with letters with words of support, offers to help, and most importantly, many people supported us by purchasing a paid subscription to themselves or friends.
Conclusion
It seems that you can not stop and continue to write for many more hours - as we made our vocabulary from 15 million sentences and use it to compile examples, about a bunch of A / B tests that we did, about new features, over which we now We work, about participation in Tolstoy Startup Camp (without which nothing would have happened). But let's leave it for now for future articles and questions.
Try the Educa class and share your impressions: educa.io
I would be very grateful if you support the creation of new materials on Educa or tell your friends about the service: educa.io/starter
Source: https://habr.com/ru/post/255803/
All Articles