Hello 👋

Martin here. Welcome to another edition of Founders’ Hustle.

Today I’m sharing the startup story of JollyGut, a UK company on a mission to disrupt change the way we buy food.

I sat down with their CEO, Denis Smyslov, to find out how.

Highlights:

• How preparing for a 73km ultramarathon gave Denis the idea for JollyGut. 🏃

• The new skillsets Denis prioritized developing after transitioning across from very different industries (finance, engineering, and oil & gas). 🔧

• Why Denis holds a contrarian position to popular ‘startup wisdom’. 💡

To receive more newsletters sharing startup stories, subscribe below.👇

JollyGut is a relatively new B2C app helping folks eat more healthily and safely according to their personal needs and budget.

So far it’s had “thousands” of downloads and right now it’s mainly used by folks who are allergic or unable to eat certain ingredients like meat, gluten, or nuts.

For example, if you are gluten intolerant KRAVE cereal is a big no-no.

JollyGut will bring this to your attention with a simple scan of the packaging—instead of having to squint at size 6 font for the words “contains gluten” on the back of the box. 👇

The app also helps customers make better nutritionally informed choices generally.👇

Prior to founding JollyGut, Denis had no experience with B2C consumer apps (his background was in finance, engineering, and oil & gas).

Having done it a couple of times myself, I’m always curious as to the frameworks other founders employ when jumping into new industries.

In our Q&A Denis shared how he navigated taking JollyGut to market from scratch.

Plus, what he considers important versus popular ‘startup wisdom’. There are some pretty bold statements inside.

Here we go. 👇

🔥 Q&A

Why did you start JollyGut?

I came up with the idea of JollyGut while training for another extreme Transvulcania trail leading across the island of La Palma.

I had a coach at that time who told me to start taking supplements (magnesium, zinc, etc) 2 months before the race.

I thought: "Our bodies developed during millions of years of evolution to obtain all required nutrients from natural foods, not from chemical plants which churn out chemicals erroneously called supplements".

I used a US food composition database as my starting point to look for foods that contain the nutrients prescribed by my coach; then I browsed online grocers for available products and duly recorded everything in rows on a spreadsheet.

I spent 2 full days doing that, just to create a shopping list for only the first 3 days of my 2-months training period.

At that time I understood that eating healthily is a very difficult task, which requires an algorithmic approach to choosing and ordering.

What is the mission of the company and what 'unmet need' are you pursuing to fulfill?

If you care about your long-term health and the health of your kids, you have to solve this task on a daily basis. The mission of JollyGut is to give people the tool to choose healthy food in any part of the world.

Governments have been unable to stop the wave of obesity, cancer, diabetes and other nutrition-related diseases which make billions of people suffer from chronic diseases.

Only consumers themselves empowered by AI can make food companies produce healthy and nutrient-rich food as originally intended by evolutionary mechanisms.

That's where JollyGut fits in: by launching our product we've made the first step towards fulfilling the "unmet need" of people to break through the fog of false ad claims which we see in the grocery aisles; we need to stop becoming victims of food companies who treat us, consumers, as a way to convert cheap junk food into hard cash at the prices dictated by the industry.

There is an "unmet need" to reverse the market: consumers should become the kings while food companies should beg them to buy their products, which meet each consumer's personalized nutrition and taste needs at prices not fixed but negotiated.

The food market should become an algorithmic trading mechanism in which purchase orders placed by consumers' algorithms are met by food companies' sales orders in the same way as financial markets operate.

There will be no place for advertising since algorithms will know perfectly well what their consumer masters need and want.

After the Transvulcania trail 'moment of realization', how did you go about making your idea into a reality?

It's been a long journey, since this idea obviously is not from the "test fast, fail fast" world. It required serious thinking through, assembling many bits and pieces of the puzzle.

We had to create a diverse IT team covering Natural Language Processing, Computer Vision, Optimization, Data Science and other algorithms as well as the backend and frontend engineers. We had to do it in a very competitive talent market lacking the deep pockets of the established players in the machine learning market.

We operated from the very beginning remotely — even before the onset of COVID pandemia. This helped us cut a lot of overhead and not to be geographically limited in our talent search.

We've managed to tap into a talent pool of experienced software engineers living in different parts of the world who are interested not in astronomical salaries but in challenging tasks, seeing their efforts materialize in real products needed by people and not wasting their energy in layers of bureaucracy.

We have spent two and a half years building the team and the product.

Where is the business today and what key milestones have you achieved?

We launched 2 products at the end of last year: an iOS app and a browser extension. We are launching an Android app in June this year.

We have limited ourselves to the UK market, not to spread our marketing budget too thin. But, we have rolled out all the infrastructure in the US and are ready to launch there when we decide the moment is ready.

We are not "in the money" yet. We went through three rounds of financing raising a total of USD 2.6m, which covers our runway for the next 15 months.

Introduce me to your founding team and CTO. How did you go about hiring them?

The other two co-founders are Erwin Parviz and Eric Personne. I met Erwin ages ago — we had the same mountain guide in Chamonix and went through quite a few exciting experiences together. Erwin has an applied math background and spent most of his career in the banking industry engineering algorithmic products.

Eric met Erwin many years ago — they used to work for the same bank early in their careers. Eric majored in physics and is an endurance runner who has accomplished such pinnacles of super marathons as Ultra Trail de Mont Blanc (160 km horizontal, 10,000 meters vertical), Le Diagonal des Foux (the Madmen's Diagonal) on the Isle of Reunion.

Our CTO is Stuart Reed, he made a successful exit as a CTO of a startup that also dealt with gathering and analyzing a lot of data. I met Stuart in a co-working space, and we had spent a lot of time together — both in the office and in pubs exchanging ideas and discussing various topics before realizing that Stuart is a great fit for our start-up.

Your background is exceptionally varied (finance, engineering, and oil & gas, etc). What new skillsets did you prioritize developing for launching a consumer app business? What existing skillsets were helpful?

I obviously have a lot of experience in launching new ideas and putting them to work.

The new skillsets I needed were:

• Learning to code and understanding machine learning tools

• Knowing how our data gathering and processing pipeline is construed and works from the inside.

• What ML algorithms we use, how and on which data we train our language and computer vision models.

This is very close to the way one sets up an engineering design firm, (which was my previous business), but in a new domain.

A completely new thing to me, which you don't normally see in the B2B market, is analyzing user feedback and data in order to generate new product features and fine-tune or completely drop previous features or functions which failed to work.

I'm totally excited that to generate completely new products and functions you need mainly brain power and only months of time.

Electricity, access to powerful cloud infrastructure, and open access software libraries are other inputs that are cheap and readily available. Consumers are readily available at a touch of the computer mouse.

In the industrial engineering world one needs millions of dollars of investment, many years of design, testing, and demoing before one gets a chance to put a new product into operation — if you are lucky enough to find an industrial partner who agrees to use your new product.

Where is the JollyGut app today? Functionality, downloads, etc.

We didn't launch that long ago, we do not have millions to spend on marketing. We are in the thousands, not yet hundred of thousands of downloads.

But our user base is growing fast. I think we'll reach the first milestone of a hundred thousand users by the end of the year. This is in the UK, with a population 5 times smaller than in the U.S, and iOS only.

We've developed very differentiated product recognition technology, which recognizes products on grocery shelves not by bar code scanning but using advanced computer vision algorithms.

We've optimized for very low bandwidths: in many stores which have thick walls, the network speed can be as low as 3G, even though the world outside may be installing 5G towers.

Our next target is multi-product recognition which will be well suited for the fast-developing smart glasses technology.

We have also built and keep improving our knowledge graph of food products. We can find the same and similar products across multiple grocery stores. This is the basis of our smart shopping basket optimization tool, which we'll start rolling out this year.

We have also built the base for many other new features which we are developing or testing.

Technology. What's required to build an app like JollyGut?

A lot of data science (data gathering, data processing), natural language processing (converting unstructured data we gather on the Internet into structured datasets which can be used for model training, stored in our databases, etc.); computer vision (mainly product recognition, but also product data input, etc.); machine-learning operations); IT infrastructure engineering; front-end design and engineering.

Did you build an MVP to test your idea first? If so, how and what did that look like?

We did have a very simple demo model which we used to attract our first investment. It was based on a real backend engine though. We could not have a classic MVP to test our ideas, because we had to gather a lot of data first, and that required almost a year and a half of effort.

I would say that our MVP was like many products in the same domain as us with millions of users but based on very simple tech ideas.

Such an approach can get you a lot of users. But, you can't give those users advanced functionality unless you completely rework the tech basis of your app or hire engineers with totally different skill sets, etc.

How are you thinking about monetization?

Our technology allows us to implement a lot of monetization ideas, which are inaccessible to other companies in our domain.

Small hint: since our technology is not based on bar code scanning, we are not inherently limited to single product recognition. Our technology allows us to recognize multiple products (5, 10, 20, 50) at a time; both from still images as well as from video streams.

This gives us immense opportunities to convert our app into a food product search engine with AR features, which can be used for monetization ends.

Important thing: our monetization plans do not include selling our users’ data.

Our main advantage is that we know what product a user — even if she is completely anonymous to us — is viewing at the point of making a purchase decision.

That's much more powerful than gathering data by spying on people when they visit websites, do a web search, etc.

What KPIs do you optimize for? Why?

• Daily and weekly active users (user acquisition metric).

• Returning users to new users (stickiness metric).

• Number of product photos taken per user.

• Number of product searches made per user.

• Cost per installation.

• Organic installation/paid for installations: this affects user acquisition costs.

What type of customer is most suited to using JollyGut right now? i.e. your ideal customer profile.

At this time we are targeting users who have particular nutritional/dietetic needs; for example, people allergenic to some ingredients; those with particular dietary preferences (we have implemented gluten-free, vegan, vegetarian, FODMAP, and are working to add more); users who want to eat less processed food; less sugar, salt, etc.

We constantly test the market for new functions and keep adding them.

What customer acquisition channels have been important for early adopters?

We use digital channels only, such as App Store, Instagram, Facebook, Twitter (in that order). We have a lot of potential users who have Android devices, so Google Playstore will probably become important, too.

How are you thinking about scalable customer acquisition?

Our main strategy is based on raising the number of organic installs. It's at 30% of total at the moment, we plan to raise it to 50% and higher.

Our product differentiation and unique features make such a target very feasible. That is also why we constantly work to improve our algorithms and technologies to make the products stickier and stickier.

What have been the most challenging aspects of building this business so far?

Hard to say — everything we do is quite challenging since everything we do is not according to the "market wisdom" implanted by numerous accelerators, VC funds and "how-to-books" ("play fast, fail fast", "be laser-focused on one thing", etc.).

[What we’ve acomplished] is hard to do for a startup that didn’t start with hundreds of million dollars in seed money. That is, developing several technologies in parallel.

But, we wouldn't be so differentiated from the rest if we didn't follow this path.

I get the subtle feeling you hold a contrarian view against "startup wisdom". Let’s delve deeper here. Why aren't you able to comply? Or, why don't you feel that is the best option?

VC fund managers have to cover up their asses by doing what everybody in their industry is doing.

If they fail within the standard modus operandi, they are safe. If they fail by doing something not standard, they will be eaten alive.

Only a select few can allow themselves to be mavericks. Therefore the industry operates within a standard framework, which propagates to accelerators, which makes startups conform to the bankers' expectations.

It's akin to a high-end night club: to get inside you need to pass face control. Once you are in you look around and tell yourself: "Who the f... are all these people? Something must be very wrong with me if the guy at the entrance thinks I'm one of them".

An idea which you can pitch in 2 minutes has the deepness of a pop song.

The more a startup conforms to the prevalent mental model of VC funds, accelerators, and the like, the lower the chance that it will lead to any significant breakthroughs.

What actions or initiatives by your team have delivered outsized results?

• Decision not to depend on bar code scanning for product recognition.

• Decision not to depend on third-party data providers for product data gathering.

• Decision to operate in distributed mode from the very beginning.

What are your core objectives for the rest of 2021?

Our main tasks for this period are acquiring users at a low cost and increasing the "stickiness" of our products.

We need to reach the threshold after which we can start testing our monetization ideas.

Thanks Denis!

Want to check JollyGut out? Download it on the Apple Store and Google Chrome.

Read Next:

A Better Way For UK Startups To Claim R&D Tax Relief

I made it my personal mission to find the best way to claim R&D relief from HMRC.

During my investigation, I discovered Claimer, a super startup-friendly solution. 👇

Why is Claimer better than using legacy options like consultants and accountants?

• 10X faster and easier (30 mins). ⏩

• Low cost (5% fee!) 😍

• Reliable, maximised claims (expert reviewed). 📈

Bonus: use this referral link to get 10% off your first claim.

Click Read Now below to access the full report. 👇