TUBR: Taking on “small data” problems — pt.I
Dash Tabor, Cofounder and CEO, shares her founder journey and the TUBR time-series machine learning algorithm.
TUBR is building better, faster and more frequent predictions with less data. Co-founder and CEO, Dash Tabor, and team wanted a solution for tackling the frustrations born from busy commutes but realised that in order to solve the problem with data, technology would need to be advanced to predicting with smaller data sets.
Their time-series machine learning algorithm can process only a fraction of the data normally required making near real time predictions possible across dynamic rapidly changing environments, like the London Underground. The team has designed a London based-app helping people find the optimum time to travel in order to have the best journey experience. Their mission is to improve how we utilise our spaces to improve every day experiences and solve those “small data” problems.
As part of Ada Lovelace Week here at Aerospace Xelerated, I had the chance to sit down with Dash to hear her journey in tech to date, how TUBR first started and where their product is heading.
Dash, it’s great to have you with us for Ada Lovelace Week. To start, how did your career lead you to starting TUBR? Where were you before?
I have a Master’s in City Planning and Sustainable Development, and I received my undergraduate degree in Public Communications, PR and Mass Communications. I went into technology mainly because that’s what everybody seemed to be doing at the time. I started in the support team at Experian and then worked my way into a data building team where we built the data updates. This was back when automation was such a hot topic and everybody was thinking “You can automate this” and “This technology is super cool” so I ended up automating myself out of a job and into a Product Manager role, work I’ve been doing for the last decade.
Working as a PM, I was involved in multiple different stages of businesses. I helped a friend start a FinTech mobile app three years ago and that was all the way from creation, so the very early days, through to when we got accepted into Techstars in Denver, Colorado. Since then I’ve worked for large companies, SME’s and also a Series A company. I spread my skills across multiple different views of companies and then thought that I was ready to take everything that I’ve learned and try and apply it and make it work.
“There’s all this data just floating around us, we’re just walking through it really, and why aren’t we utilising it to use every minute of travel.”
How did the idea for TUBR come come about and what is your artificial intelligence application? You mentioned your degree in city planning, has your background inspired you in some way too?
Yes, absolutely, TUBR is a bit of a combination of background. The city planning degree gave me a bit of understanding of how things work, how planning works, how to get proposals through, and how the government sets priorities. However, I would say the real background element was the data. I was building all these different data propositions and I was so tired of having an armpit in my face on the London Underground.
I realised that if I had left my house at 8:17, I would get through the barriers early and I could probably get an aisle spot, maybe even a seat. But if I left 10 minutes later, I’d be locked outside the barriers. In that case, I would go get a coffee and come back and walk straight onto a train with no issues. I started to realise that we were all crowding to the tube at the same time but then the time that I would get to work would only be maybe 10–15 minutes different. Those earlier trains were actually taking longer than waiting a few minutes.
That’s how the idea came about. I thought, “this is crazy, data can solve this problem for us.” There’s all this data just floating around us, we’re just walking through it really, and why aren’t we utilising it to use every minute of travel. I realised quite quickly that we didn’t have access to the data that we needed. When we went out looking for the data, we realised that what we needed didn’t exist, at least not at that point, and the access to the data that we did have was so sparse that we couldn’t use machine learning at all at that stage. My co-founder and I decided that we have to solve this machine learning problem in order to make the data that we have valuable for this use case. I was lucky enough to find a brilliant technical co-founder and he created a machine learning algorithm that processes time series data with fewer data points. That’s when we said that we’re actually kind of cracking this bigger world of small data problems and the time series space, and that’s how we ended up doing it.
That’s interesting what you’re saying about data, something we talked about recently on our Demystifying AI panel. It’s a bit of a chicken and egg problem. You need data to get started, but where do you find the data?
Often, there’s all this data but it’s not usable, either because it hasn’t been processed correctly or we’ve waited too long to process it. Then, it has all these complexities that it’s almost too scary to do anything with. Additionally, there’s the element of it being scary, because what are people’s perception going to be of using this data? How can we create an educational piece around the fact that data isn’t scary, and the majority of it isn’t personal? And you can solve a lot of problems with that. But there’s a lot of misconceptions, misconceptions in that space. Right now. We don’t do anything with personal data. We don’t know who anybody is, we don’t track any particular information, but we aggregate how people are moving and where people are moving around in order to solve the problem.
“Is this a problem that people want to be solved? Are people actually going to change their behaviour in order for me to make a difference in this space?”
TUBR started in 2020, and the pandemic struck shortly after that. We’ve talked about the data issue. What were some of the other challenges that you faced?
This is something I’ve been wanting to build for years. And I decided early in 2019 that 2020 is my year and I’m starting a business. One of the first things that I wanted to figure out was: Is this a problem that people want to be solved? Are people actually going to change their behaviour in order for me to make a difference in this space? I started my market research before I even registered the company so in January 2020 before we even knew about the pandemic. I ended up registering the company a few months later in April.
Looking back, that’s the one thing that we’ve looked at the entire time, how are people’s perceptions changing from pre-pandemic to during the pandemic. I think one thing is definitely clear — those uncomfortable moments on public transport are something that people didn’t like then, and they put up with but they’re probably not going to put up with now. We’ve spoken to commuters and 34% of them say that they will actually get off the train if it gets too busy and they may even turn around and work from home. People are massively willing to change their behaviour now more so than they were previously, although there was always an appetite for this so experiences become a lot more important. What’s interesting is that at the beginning of the pandemic, there’s a lot of fear and anxiety that hasn’t completely gone away, but people are starting to push themselves out of their comfort zones a little bit more. When we polled the public recently, about 1/3 of the people are saying that they’re not going to get back on public transport.
Some of the challenges that we faced through the pandemic were firstly, it was really hard to test. I’m a product manager so I like to fail fast and pivot. That means testing everything and getting that validation, which was really hard to do during the pandemic. We also learned quite quickly, once we were able to get the app out there, that the market research that we had done early on wasn’t the way people were travelling anymore. Right now, we’re actually in the process of completely redesigning our app and hopefully, that will be coming out at the end of the month. A lot of the market research you conducted pre-pandemic and during the pandemic just isn’t valid anymore. And it’s something that nobody could predict so that made it a little bit tricky. Also, being virtual is wonderful in a lot of ways because you can accomplish so much more in a day, but you don’t build those face-to-face relationships as easily. This has definitely impacted fundraising and building our advisory board so from a business perspective, that’s been a little bit tricky.
You’re talking about how things changed during the pandemic and how you’re currently going through the redesign process. How did the pandemic impact your initial value proposition?
Originally, we were just thinking about crowding and being able to recommend those minutes in travel. That’s one thing that we realised, as soon as we released the app, that there are a lot of other things people are interested in. We were originally thinking about crowding whereas now commuters are thinking about space. We’ve done all this research but people really don’t care about how crowded it is, they care about how much space they’re going to have. We’ve had to rethink how we show space and we’re still working on this, but space is a big deal to all of us now. For most of us, we’ve now spent over a year in our homes and, whether they’re small or spacious, we definitely haven’t had an armpit in our face. Right now, carriages are crowded but that jam packed, sardine, experience isn’t an everyday occurrence on the tube, at least not yet, but we could get back to that point quite quickly.
For us, being able to make predictions and use our AI to make minute-by-minute predictions for every hour, up to a week in advance, but obviously, the hour is most accurate, because that’s what’s happening now. Then it takes events into consideration, so if it’s raining, if there’s traffic, if it’s a holiday, if there’s an event going on, all of that impacts how people behave and what happens. Those triggers will come in and say “it started pouring”. That’s going to change your predictions for the next 15 minutes, 20 minutes, an hour. We’re working really hard to make sure that it’s accurate, and that we’re taking into account what’s happening around us so that we’re always feeding out the right information.
Thank you Dash for sharing more about your mission at TUBR and approach to solving those “small data” problems.
Stay tuned for part II where Dash talks about how her previous experience as a Product Manager impacted her journey, how TUBR approaches bias in artificial intelligence and ethics, and her best startup advice.
We hope you can join us for Ada Lovelace Week — use #AdaLovelace and #WomenInTech on socials to shout your support and celebrate the women you respect. You can follow the content we’re publishing this week via the #AdaLovelace tag here on Medium.
Aerospace Xelerated is a 3-month funded programme for exceptional autonomy and AI startups to accelerate the growth of the aerospace industry. Learn more about our work in our FAQ or book an Office Hours call to chat with the programme team.
Stay up-to-date with our latest updates via Twitter and LinkedIn and subscribe to our mailing list!
