Adventures in growing a startup dealing with chatbot training datasets
If you don’t know us yet (most probably you won’t), we are a young Swiss startup. We decided a few months ago to build a platform dedicated to one simple task: build training datasets for chatbots.
Long story short, while building a love relationship coach chatbot, we found out how difficult it is to properly train it to understand what users are saying. Building a proper training dataset for an NLU (Rasa, Dialogflow, Luis, you name it) was a painful task.
While solving this problem, we started to see a business opportunity. So, as they say in the startup world, we pivoted. During the last two months, we have been keeping our heads down our keyboards. As a result, we are now the proud provider of chatbotstrap.io, a platform dedicated to help you build proper chatbot training datasets.
By the way, if you want to build a chatbot, don’t hesitate to have a look. We would love to have your feedback.
Anyway, back to the marketing subject. We feel that we have build a good product. Our friends and former colleagues (at ELCA and Serial two great agencies in Switzerland) gave us a good feedback. They are now starting to use the tool.
Are customers flocking in and subscribing to our paying plan ?
Thanks to our startup friendly coworking space La Forge at EPFL innovation park and the recent meetups organized by them, we know the answer. NO. Or, as @laurentbalmelli did put it during his recent talk “beyond marketing bs”:
Successful products are usually the ones with the best marketing rather than the ones with the best features.
This fact is unfortunately confirmed by our platform analytics: During the last week, we did not drive a lot of traffic:
So we decided to step up our marketing game and came up with a plan. As agile versed engineers, our plan consists in weekly iterations:
1. Implement actions from our marketing backlog
Vincent Quero from freshmilk.ch gave a really good presentation at our co-working space about content marketing and how a startup can implement it. This did clearly kickoff our marketing backlog (full of ideas now), kudos to him !
2. Measure our actions success or failure rate
Our marketing efforts will clearly be data driven. To get started, we will monitor the number of visits on our platform and the conversion rate driven by each one of the implemented marketing actions. If the numbers look good, we will continue to invest our time in these actions.
3. Report them on this blog
We are really curious to understand which marketing actions benefit our business. And if you kept reading this article until here, it probably means you are too. So keep following us, this is only the beginning !
4. Restart the process
And what about this week ?
We decided to start with basic and simple actions. Our plan for the next 7 days will consist in:
- Answering 5 to 10 questions on Quora or Stackoverflow like this one.
- Creating a newsletter with our latest articles and features. Sending it to all the persons who have shown an interest in the platform idea.
- Creating a Linkedin Chatbotstrap account to communicate.
Stay tuned to see if these actions were helpful or not and to find out about our next marketing experiments. See you next week !
