Skip to content

Precision AgronomyHummingbird Tech

As part of a cross-functional product team in Hummingbird, I designed and delivered a new platform which uses high-res drone imagery to analyse the state of crops, and enables farmers to create precise application plans. 

The product space

Agri-tech SaaS, focus on enabling precision agronomy.

The brief

Develop a new SaaS platform and deliver MVP in 3 months.

The work

Planning, research, UX and UI designs, dev hand-off and QA. 

The problem space

Agriculture is one of the biggest industries in the world worth trillions of dollars. However, in terms of technological advancement it is still trailing behind other parts of the food supply chain. Agronomy develops at a healthy pace but the application of better practices is dependent on having software and hardware which can analyse massive areas of land and zoom in on single-plant level. When done manually, all crops are sprayed a flat rate of chemical input (nitrogen) which incurs costs for the producers and has an impact on the quality of the fresh produce. 

Farmers need to be able to have an easy overview of key aspects of each crop head – size, quantity, health – and understand the patterns in the crop development to be able to create custom and detailed application plans. This would allow them to save costs on nitrogen input, it will reduce crop wastage, it will produce fresher, untreated food and it will have a positive impact on the soil.

The challenge

This new product, internally called “Fresh Produce”, was meant to be an improved and optimised spin-off of the original Hummingbird SaaS platform. Working with a legacy product can provide an advantage, e.g. existing insights and research, and the opportunity to address problems in completely different ways. However, reusing functionality and components comes with tech debt and limitations in the back-end. 

Additionally, the timeline to launch a whole new MVP was extremely tight and non-negotiable as the end users (farmers) are always restricted by seasonality – missing the time window means a year’s delay.

Kick-off and insights

The team featured myself, 2 product managers, 2 front-end developers, 2/3 back-end engineers and  a data scientist.

We set to achieve the following goals:

  • Determine the functionalities of the current platform which are:
    • most accurate and efficient
    • deliver most impact to the customer
  • Reduce impact of historical tech debt
  • Streamline business offering and USPs
  • Align product with business objectives for the next 12 months


The first phase required a lot of combined effort to go through all the research, findings and background knowledge the team had – especially people who had been customer-facing. I went on 3 farm trips where I followed customers around and conducted in-depth interviews. Upon completing the research phase, I compiled and analysed the insights before playing them back to the team.  I mapped out various possible flows and facilitated workshops with the SMEs in the squad until we arrived at a final, refined proposal. I was leading the next steps of quick prototyping, testing and iterating.

Final user journeys (MVP)

Yet another challenge

Testing in the B2B world is always a major obstacle because the a formulaic approach such as “3 rounds of 5 test subjects cover 85% of major issues” is already ambitious. Users to test with are hard to find in a niche market where a “customer” stands for a massive corporation or a global food producer. Test subjects are hard to find, get hold of and in the agri world people are not particularly open to casual conversation – everyone is very protective. 

This problem was there when we started this product and it persisted throughout the full cycle, even after deploying the MVP. 

How did I tackle it? By getting to know the customers and categorising them. Developing a whole new persona for each one would have been useless because they all fit the target audience. However, each customer had a different background, especially in terms of tech savvy. This way I knew what to expect from every interview. One customer would be the gold star – they were very technologically advanced which helped push the ambitions of the team and scope the long-term roadmap. Another customer was years behind in their tech adoption so thy were a perfect indicator of how simple and friendly the MVP was. 

As long as you know where the feedback is coming from and you can put it into context, you don’t have to be confused and fall into the trap of convoluted and widely diverse user feedback.

💡 As long as you know where the feedback is coming from and you can put it into context, you don’t have to be confused and fall into the trap of convoluted and widely diverse user feedback.


The team effort resulted in simplified user journeys and a whole new IA which focused on three main functionalities – displaying analysis, booking flights and uploading imagery. The user journey was stripped to its very core and any processes which were not yet automated would be done manually backstage – ensuring the user has a seamless experience with as little friction or waiting time.

MVP information structure


In order to avoid overthinking the new concept and getting lost in the desire for perfection, I moved quickly from proposed flows and IA to rapid prototyping. Timings didn’t allow for a fully-fledged round of low-fi wireframe testing with users – this was only sufficient for internal feedback and sign-off. I took on the task to develop a new UI and apply that in the hi-fi prototypes. Killing two birds with one stone – testing usability while the front-end devs already had new UI components to start working on. It’s all about stacking up work load.

I’ve recorded a walkthrough of the MVP prototype – have a look below. 

Execution and delivery

Making sure that everything is well communicated, organised and easy to access in the future, I put together a new design system based on the new aesthetic. 

Button styles


The 3 month deadline was met which was a great collaborative effort with a lot of last-minute crises and tweaks. All the testing I had done with users had provided the team with a list of ideas for features which we had assessed based on feasibility, resource availability and business goal relevance.  I continued to iterate on the prototypes and did regular testing to help the product owner evaluate and prioritise upcoming work. 


nitrogen input saved
0 %
per ha added value
0 £
new customer trials started in Q1

Like what you see? Curious to find out more? I’m always happy to dive deeper and give more details on my recent work so just get in touch…