Value proposition Flashcards
(16 cards)
Who are you? (Fede)
Hello, my name is Fede. I’m from a small city named Salamanca in Guanajuato, a state of Mexico. Throughout my life, I have received public education up to university being a first-generation college student. I graduated from ITAM, a top university in Mexico, through a scholarship for tuition and living expenses. I obtained a bs degree in applied mathematics and economics. Prior to Nixtla, I worked at the Mexican central bank and at Banorte, one of the largest banks in Mexico. I am currently working as a machine learning engineer/data scientist at Nixtla.
Who needs what you’re making?
Data scientists, developers and companies. As a team, we have more than 7 years of experience in time series forecasting. And we have worked with data scientists and developers really close. We have seen their struggling trying to deploy these models to production. Data scientists and developers need a fast, scalable solution and need to know how it works and its running code. That’s why they need the open-source version of nixtla. Our clients need nixtla, too, especially when they need an end-to-end forecasting solution.
What are your users doing now?
From the developers’ side, they build their own tools without testing relevant features such as scalability and state-of-the-art performance. According to Gartner, 43% of supply-chain companies are planning to use AI-powered forecasting from the client’s side. Nixtla is prepared to address these significant opportunities.
(Garter: advisory company)
How do you know customers need what you’re making? How do you know people want this?
We have worked with data scientists, developers, and companies. And we’ve heard from all of them that there is a lack of transparency in how machine learning models work for time series, and it’s difficult to deploy these models to production. From this perspective, we’ve learned a general need for transparency of the models and an accessible tool to deploy the models to production. From the enterprise side, only 45% of supply-chain companies are using AI-powered forecasting.
What are the top things users want?
An open-source alternative so they can know what’s happening with the models. Instead of using black-box technologies. And they need this alternative to be fast, scalable, and highly accurate.
Why isn’t someone already doing this?
In fact, big tech companies offer forecasting services but they use a closed model that doesn’t democratize the use of machine learning, ai. We don’t really know what code is being executed, how the users data is being handled. It is not a transparent technology.
We want to democratize the use of tested machine learning models leveraged by the big open source community.
Time-series forecasting and the model’s deployment to production require deep technical knowledge in a wide range of topics, such as deep learning, cloud computing, programming, scalability. And we know we can do it because we have already done it at a lower scale.
What makes new users try you?
New users are already trying us. With the pre-launch version of nixtla, we have tested our tech has state-of-the-art performance, replicability, transparency of how models work, and scalability. And that’s what our users like the most.
Why do reluctant users hold back?
There may be eventual expenses when switching to our technology.
How do users know they need this?
Companies already need time series forecasting; there is an ample opportunity for them to change to the ai-powered forecast. Also, there is a big community contributing to open-source software.
How are you understanding customer needs?
Feedback from both users using our open-source repository and clients make better decisions through our forecasts. In the future, we will also have a monthly feedback loop with early customers and facetime during webinars.
What’s new about what you make?
Quality, performance, usability, affordability and transparency. Let me explaon. The premium forecasting software’s side (such as Amazon Forecast or Azure ML) is usually expensive, and the user doesn’t really know what code is being executed and how. And current open-source technologies lack rigor and quality. For example, take pytorch-forecasting or pytorch-ts. They claim to be implementing state-of-the-art models, but they don’t show their performance.
Nixtla is the only company addressing this problem from an open-source mentality. We want every data scientist to access premium forecasting technology and allow people to collaborate and grow a high-quality code repository that is freely accessible.
Our open-source philosophy is based on its proven success in other fields. For example, pytorch-lightning is an open-source library for building deep learning models and they have also an enterprise version named grid-aidand recently they have raised 18.6 million US dollars. And there is also hugging face they build nlp models and have their public resources and also an enterprise solution and they have raised 40 million us dollars. We want to be the pytorch-lightning and hugging face of the time-series-related tasks.
Who are your competitors?
AmazonForecast, Azure ML, lokad, ForecastPro, Relex
How many people are in your target market?
According to Gartner, the 55% of the demand-forecasting market don’t use forecasting technologies based on machine learning and 43% of the market is planning to use AI powered forecasting.
How many $B is the market?
“The estimated impact of AI in the supply chain is between $1.2T and $2T in manufacturing and supply chain planning.”
US$ 14.5 by the end of 2030.”
How fast is the market growing?
Expected growth of 16% from 2020 to 2030 (demand planning baseline).
Useful phrases
I would like to compliment on that.
