Published Last Week

Just Eat Takeaway & Online Food Delivery

The food delivery business provides an interesting case study on the threat of vertical integration. In the early 2000s, the first online food delivery marketplaces were launched to connect restaurants with hungry customers. Platforms like Just Eat (UK) and Takeaway.com (NL) facilitated orders which restaurants delivered to customers. It was a capital and complexity-light model. 

Platforms would acquire customers and route orders to local restaurants for a 10-15% fee. Restaurants would receive incremental orders for zero incremental capital spend. Given small restaurant owners typically already had drivers, in many cases often their children, incremental orders drove higher profitability. It was a win-win relationship. 

Just Eat and Takeaway’s original marketplace earned ~40%+ operating margins and >60% ROE. Such returns attracted vertically-integrated entrants like Deliveroo, UberEATs, and Wolt offering to acquire orders and deliver the food for restaurants. Like Amazon moved further down the value chain than eBay, Deliveroo and Wolt moved into delivery to add more value to restaurants. This led the original capital-light marketplaces to launch delivery services to protect its marketplace profit pool. On the surface, this has destroyed the profitability of companies like Just Eat Takeaway.

But food delivery is hyper-local. The competitive game is played within towns, not countries. The vertically-integrated players focused on cities first. Each company sourced top restaurant chains that typically didn’t offer delivery. Deliveroo has an exclusive relationship with Nando’s, for example. Such large chains provided instant scale to acquire customers and build a brand. The original marketplaces reacted by employing or partnering with gig couriers and competing for large chains to protect its city market share. 

While this highly competitive game within the cities continues to unfold, the original marketplaces still dominate small towns and suburbs. Deliveroo or Wolt are not targeting indie restaurants in towns of 50,000 people. Yet. 

This highlights one question we’re exploring around TKWY: how may the competitive pressure of Deliveroo and Wolt in cities like London and Berlin expand to smaller towns where TKWY earns its profit?

This interview with a Former TKWY Senior Executive explores how the delivery models and economics compare in UK and Europe. For example, last year, Just Eat recently moved back from an employed to courier model in the UK. Courier models are cheaper and potentially better align incentives between the driver, platform, restaurant, and customer.

purely looking at a wage basis, as far as you can compare, there is usually an industry generic cost difference between gig and employed models. How much depends on the legislative regime, and what’s offered in such a gig model. Sometimes the differences are minor, but it can also differ 20-30%. There's also a similar difference related to operational aspects, such as planning around peak moments and how to deal with hold times at restaurants. As the impact of those varies between employed and gig models. It's also about no-shows and weather patterns. If you have an employment model and it starts raining, how do you ensure you can source more couriers on the spot? - Former Senior Executive at Just Eat Takeaway

The shift to the courier model has led to Just Eat Takeaway's UK EBIT doubling to £140m in 2023. Given TKWY has ~3bn EUR enterprise value, and has recently sold Grubhub, this has piqued our interest. 

Just Eat Takeaway UK P&L, 2022/23
Just Eat Takeaway UK P&L, 2022/23

We plan to cover TKWY within the UK and Europe over the next month.  

Google DeepMind: AlphaFold

This interview with a structural biologist at AstraZeneca explores how large pharma companies use Google’s AlphaFold to predict the structure of proteins. The more accurately one can predict protein structure, the more effective biologic R&D. This interview also has readthrough into Danaher and Sartorius’ bioprocessing division:

AlphaFold reduces this workload by predicting protein structures computationally, sometimes eliminating the need for lab bench work. These predictions, based on AlphaFold's algorithm and transformer-based architecture, are quite accurate compared to real-life results. In academia, I typically use AlphaFold at the beginning of a project to predict potential structures based on available data from the PDB, where publicly available protein structures are stored. As AlphaFold has evolved, it can now predict multi-protein complexes, including protein-RNA and protein-DNA complexes, whereas it was initially limited to single monomers. - Senior Structural Biologist at AstraZeneca

Experian: Growth beyond Credit Scores

Experian is one of the highest-quality UK listed companies. Its the largest credit bureau globally and aims to leverage its unique database to move into adjacent markets. The company regularly prints 25%+ EBIT margins which almost fully converts into operating cash flow. This interview explores its growth opportunities beyond credit scoring and the unique attributes of its current model:

In what industry do you get most of your raw material for free, aggregate it, clean it up, and then sell it back to those who provided the data? It's highly cash flow generative. The bureaus operate as oligopolies, with typically only two to three big bureaus in any market, except India, which is unique with six competitors. Each has had to define its market. The non-traditional bureau companies, like Credit Karma, leverage the big bureaus' data. They don't maintain their own data sources. The moats for the bureaus are deep, with centralized data management, cleansing, and maintenance. I believe it's sustainable for the next five years. We were concerned about data providers holding onto their data or charging for it. We do pay to collect court and rental data, but it hasn't become a major issue. - Former Senior Executive at Experian Global Technology.

IP Fieldwork: the Art of Primary Research 

IP Fieldwork is a free podcast that explores the pitfalls, challenges and opportunities inherent in conducting primary research - work that we consider is an art and a science.

You can follow IP Fieldwork on Spotify and Apple or add our feed to any podcast player here.

Upcoming episodes on Primary Research Methods

  1. How to structure project requests to Expert Networks
  2. Strategies for sourcing quality former executives 
  3. How we vet executives
  4. Common mistakes by Expert Networks
  5. How to prepare for expert interviews
  6. Structuring interview questions 
  7. Building trust
  8. Techniques for asking questions 
  9. Questions vs line of questioning 
  10. Common mistakes in asking questions 
  11. Active listening
  12. Post-call interview review process 
  13. Our process of curating primary research