Project Everest

Work Update

What can Campbell's Soup teach us about farming?

Cec Cameron
Cec Cameron | Jan 9, 2018 | in Agriculture Assessment

There can be a tendency to want to be perfect. To create the perfect product to fulfil our business’ value proposition. Howard Moskowitz, a researcher, food consultant and psychophysicist, researched and demonstrated in the 1980s that creating the perfect spaghetti sauce for Campbell’s Soup was not the most profitable option. Rather, segmenting their product line and creating a new product of extra chunky sauce (for those consumers who liked it chunky ’n’ funky) was best. This generated over $600 million in profits over the following 10 years. 

Just like consumer’s tastes in spaghetti sauce, every farm is different. For Malawi’s Agricultural Assessment team applying this theory of horizontal market segmentation is leading us to broaden  our product options. This is leading to a more tailored service for farmers. Originally the minimum viable product that our team was ideating to sell was a ‘blueprint’ of a diverse farm that optimised yield, nutritional diversity and had multiple harvests throughout the year. An issue with this idea is that  there are multiple factors that influence what grows well on a farm and how plants are managed. Therefore, how is one blanket solution, in the form of a blueprint, meant to be a viable option to maximise land productivity for every single farmer that might desire our product?

“Different arrangements - or tastes - will work for different people” 

To address this, we are suggesting designing several blueprints. To effectively (and efficiently) evaluate which of the X blueprints will be suitable for our customers, we will use digital technologies to recommend the optimal blueprint, given certain inputs. Variables that are likely to act as inputs include: 

1. Soil 

    • Type (e.g. clay content)
    • Fertility (Nutrient composition)
    • pH 
    • Salinity / Sodicity
    • Soil Organic Matter

2. Climate

  • Rainfall
  • Irrigation
  • Temperature 

3. Land

  • Size
  • Topography 

4. Management

  • Capital available (e.g. ability to buy livestock)
  • Paddock history (crops, fertiliser, pests, disease)
  • Objectives (e.g. yield maximisation, profit maximisation)

 These inputs will be entered as a score, for example out of 10, and each input given a weighting (which will be derived after extensive research). Additional, there will be ‘rules’ whereby if there are limits hit (e.g. toxic quantities of nutrients, extremely low / high pH or a severe pest/pathogen issue) the recommendation will be ‘flagged’ and therefore require greater human review. Each recommended blueprint will address specific issues. For instance there will be a blueprint that is more drought resistant, one that is for an irrigated farm and one for severely degraded soils. To ensure the blueprint is being implemented and managed appropriately, there will be a consultancy service. The reason this is more desirable than working with 1 blueprint and then iterating it for each farmer is due to the increased scaleability, as well as increased personalisation.

Refining our approach to be tech based is a way to increase the scaleability of our business. A 2016 Deloitte report titled “From Agriculture to AgTech” outlined that AgTech grew at a 63% CAGR  between 2010 - 2015. The combination of the size of the agricultural industry, the importance of it to society, and the scalability of technology is why Agricultural technology is such an exhilarating space to be working in.

To manage this technology the aforementioned inputs need to be measured and managed whilst we are developing these blueprints. 

It is unlikely that the first product that is sold to customers will be a digital technology as outlined above. Instead, it is likely that the customer roll out will begin with a personal (face-to-face) service and as these customers come on board more data can be collected as experiments can be conducted at these farms. This data can be used to develop and refine the algorithm outlined above. 

What are your thoughts? Comments on the product ideation welcome!

2016 Deloitte report: 

edited on Jan 10, 2018 by Cec Cameron

Cec Cameron Jan 10, 2018

Brady, I couldn't tag you in the post as the number was capped, but I would love to hear your thoughts (along with the rest of The Company's) about where the January team is taking the product. Continuity of projects between months is so vital for projects to progress forward so please let us know your concerns, questions and factors you think we may need to give more consideration!

Users tagged:


Lucy Noble Jan 12, 2018

Awesome crowdicity content Cec!
Could there be the potential to have a singular blueprint design with a common value proposition, however within the design, have the flexibility to manage the varying aforementioned inputs and data collection points (soil, climate, land and management variables). Rather than having X number of blueprints that could be recommended to a farmer, it would look more like a customised singular product.
This does not mean that you cannot have practices that address drought resistance or are tailored for irrigated properties, however, it would make the design of the blueprint/s more efficient and during rollout and testing of the product - results could be more accurately measured and compared. I don't see the reason for any less personalisation or potential scalability with this approach - but please do elaborate.
Also, whilst moving forward with this approach, make use of the Nsambudzi property and test the input scoring system. As you refine this scoring system it will be really interesting to see what recommendations you could make on the plot of land as a result of the input analysis.


Rachel Chan Jan 15, 2018

Hey Jan team! It is great to see some ideation on the blueprint solution. Even better that it involves technology to increase its scalability.

I agree with Lucy - rather than having several blueprints that each address specific issues, could you design one blueprint that can be adapted according to the different inputs and issues identified? I would think that most farms require a more holistic solution than one targeted at specific issues such as resistance to droughts. These issues could instead be factored in as considerations in the blueprint design.

What kind of technology and expertise will be required to develop these algorithms? For example, how will you know what variables to measure and how to draw useful conclusions from analysing them? Perhaps it would be useful to see what tech solutions already exist (e.g. CropIn) and if anything can be learned from them. I am interested to know how you came up with the weighting system and the rules that require human review.

The project seems to be moving in a good direction and I’m excited to see where it goes!

P.S: Love the reference to Campbell’s Soups. :)

Users tagged:


Kate O'Donnell Jan 28, 2018

Really interesting, this is the same impression I got from the December handover docs. In the first stages of farmers coming on board, would this input system be used? Those inputs would need to be tested to see what sort of variance is present in the region, and also how these variances affect the blueprint.
Really excited to expand on this idea more in the next month.


Amber Johnston Jun 30, 2018

Status label added: Work Update