projecteverest

Proposal adopted

The Magic of Measurement and Metrics

by
Gabriel Raubenheimer
Gabriel Raubenheimer | Jan 8, 2018 | in Knowledge Base

The build - measure - learn cycle is key to the lean startup model, and FarmEd Field Development has identified facilitating this as a priority. This will allow us to continuously iterate, affirm (or refute) the validity of our ideas, and provide stakeholders with a concrete PoC. Ultimately, this iteration is a key way to maximise the chance of FarmEd's long-term success.

 

As we dig deep into the build stage with a functional test farm coming to fruition, we've been asking ourselves which metrics would be most effective in achieving this. Three factors are of key importance in our minds;

  • Growth rate of the plant - we want customers to be able to get blueprints up and running quickly, and replace plants that die whilst minimising lost productivity.
  • Crop yield (amount and quality) - this will be the key financial metric, and increasing this is our ultimate goal.
  • Plant health and longevity - again, we want healthy plants that provide long-term benefits with minimal intervention.

 

The attached image shows the measurements we intend to take. This will be put into a database, because even those numbers that mean very little to us now (ie. because we have no control bed or past iterations) may well be very useful in the future.

 

Equally important is iteration of the metrics themselves - they will only be increasingly important, and refining which numbers are most is important is invaluable - data we collect may even be useful for expanding chatbot capabilities.

 

Comments/refinements/opinions? Please drop them below.

edited on 8th January 2018, 11:01 by Gabriel Raubenheimer

Roya Ghodsi Jan 10, 2018

Hey Gabriel! This measurement template is fantastic, and definitely something the Malawi team can adopt. How would you go about measuring the first few factors (dry and wet weight, root mass etc) without damaging the plant? Or are these measured once the crops have matured and been harvested?

Reply 2

Gabriel Raubenheimer Jan 11, 2018

Thank you for your comment Roya, I'm very happy it's helped.

Very good point. If you look at column two, these are definitely destructive measurements. Our use of it will be at full maturity, and we'll have to sacrifice a plant to do so. The benefit is that it will give us a good idea of the end-state health of the plant.

Please feel free to contact me to discuss further, and let me know if you make any improvements.

Reply 2

Roya Ghodsi Jan 10, 2018

Team - check this out.

Reply 3

Vivienne Smith Jan 10, 2018

Hi Gabriel, great article :) have you checked out FarmEd Fiji's database? This google sheet is a placeholder for our data while we build a new database system, which corresponds with surveys we've been conducting: https://docs.google.com/spreadsheets/d/1gT1F_...c-TM/edit#gid=0

We are working to streamline the data we've collected so far (over 80 farms) and we've been discussing with our Agricultural consultants Zoe and Lisa for the metrics they think are important (I've tagged them below). We envision that all the data from each country for FarmEd will be in the one database.

Furthermore, just wondering about collecting data for plant growth – is it expected that trekkers will return to individual farms to remeasure the exact same plant that was initially measured? This may be time-consuming.

Otherwise great job guys! :) glad you posted this here so we can discuss!

Reply 1

Liv Hendy Jan 10, 2018

Hi Vivienne,
This is specifically for the test plots located in malawi, cambodia and fiji in order to establish clear metrics for growth rather than using anecdotal evidence, further streamlining the ability of teams to collect crop specific data. Just a thought, can be used in conjuction with existing databases instead of replacing them :)

Reply 5

Cec Cameron Jan 11, 2018

Hi Gabriel,

Thank-you for posting this. It has give the Malawi team a really good insight into your measurements and therefore what we may consider for our demo farm. I have a few thoughts and questions surrounding your data collection.

1. What is the purpose of measuring (i.e. what does that information tell us about the plant's performance?): wet weight, dry weight, plant height and number of flowers?
2. What are the specific steps involved in measuring dry weight and the surface area of leaves?
3. How will you calculate disease and weed prevalence?
4. How do you define plant longevity and how is this measured?

Here is a link to our measurement spreadsheet at the moment (https://docs.google.com/spreadsheets/d/1YmHjY...afE/edit#gid=0). It is still very rudimentary, with large knowledge gaps and your insights would be greatly appreciated.

Love your work!
Cheers,
Cec

Reply 1

Ciaran Hoare Jan 12, 2018

Hi Cec,
Wet weight, dry weight, plant height, and number of flowers are all indicators of plant health. Compared with the control bed this can indicate success (or lack thereof) of the blueprint implemented.
(Twins or Gabe tell me if I'm wrong)

Users tagged:

Reply 2

Cec Cameron Jan 12, 2018

Thanks Ciaran! Unfortunately my question was unclear, I am familiar that these measurements are an indicator of plant health. I rather meant more specifically how the will be used to indicate plant health? Will the measurements collected just be compared to an average for that variety, collected and observed how they change according with variables changing? Or is there a plant health issue that wants to be measured (e.g. spring is contracting and early flowering tends to be better). What results and analysis will these measurements contribute to?

I feel collecting these measurements is the important part, and the rest can be worked out later.

Reply 1

Lisa Paisley Jan 21, 2018

I think getting DW/WW is important, however, it would be cool and useful to flesh this out. For example, measure the vegetative material and the reproductive material separately. Cec I want your thoughts, but this way we can observe where/how the plant is allocating resources. i.e. if the nutrients are going to the flowers or leaves of the plants. This could help us understand the plants' response to stress. BUT for an accurate representation of this idea, you would need to collect this regularly, which wouldn't be feasible on a small scale, for a destructive measurement.

But as is mentioned below, we would need a standard/previously existing data to compare these results from.

Reply 1

Gabriel Raubenheimer Jan 22, 2018

Thanks for your comment Lisa. Is this something that would have enough value to warrant implementing? If it is, I'm sure we can put systems in place to do so.

Are you happy with the rest of it? As it moves into database form, it'll be harder (although by no means impossible) to implement changes.

Users tagged:

Reply 0

Gabriel Raubenheimer Jan 16, 2018

Hi Cec, thanks for the insightful comment, you've raised some highly pertinent points.

Exactly as you suggested, dry weight, wet weight, height and numbers of flowers will be compared to species averages, as well as results from our own control plots. Importantly, it will also be compared to data collected from further iterations, as a way for us to measure the effectiveness of changes we make.

In regards to why these measurements are necessary despite their destructive nature, when a measurement such as yield is arguably of greater direct value, we are concerned with the holistic health of the plant. A 'healthier' plant should produce higher quality crops, and provide greater benefits to adjacent plants. These things aren't easy to measure directly, so 'plant health' as an umbrella measurement is useful for us.

Measuring the surface area of leaves can be done by tracing onto graph paper and counting the number of squares enclosed within. Not a precise methodology, but it should provide enough information.

Dry weight measurement standard procedure can be found here:
https://clu-in.org/download/ert/2034-R00.pdf
I don't know how you're positioned in Malawi, but in Cambodia we will need to improvise a little bit - probably using sunlight to dry out the plants, whilst ensuring no plant matter is lost.

Disease and pest occurrence is a categorical, qualitative measurement - simply checking which diseases and pest are present. Disease and pest prevalence is numerical and quantitative, and can be expressed as a percentage of crop samples afflicted by a certain condition.

Plant longevity is time from germination until death. Plant death, from what I understand, is a less than binary measurement. If cause of death is water deprivation, the plant will reach a permanent wilting point (ie. point of no return). If the cause is disease, a certain proportion of the plant being covered indicates death. We can develop a full list of definitions, but unless we are measuring every day, we may be able to measure qualitatively, ie. by best judgement depending on the plant. This is definitely worthy of further discussion, and I'd love to hear your thoughts.

That spreadsheet looks very good - I particularly like the focus on recording plant stages, so we can develop more concrete timelines.

I'd be very interested in collaborating to create a single, cohesive template for all of our measurements. I also have some information as to how Fiji has been approaching this, which I'll look to reconcile with our spreadsheets. And agreed re. collecting being the priority. How we use the numbers is something that can be progressively defined as we go.

Reply 1

View all replies (5)

Corey Middleton Jan 23, 2018

I have attempted to read through this thread and I don't think I seen this thought left anywhere as of yet. Please tell me if i am wrong.

1) Is there a plan in place to make sure all countries are using the same system of tracking, measuring and comparing? so that we can average results across regions, countries etc.

2) I feel another important metric would be quantity of herbicide, pesticide and soil ameliorants used to achieve the level of plant health. Because if you achieved 20% better yields with plot A, but spend 30% more on chemicals or ameliorants; was plant health really a good measure of success?

Does this make sense? Can it be done

Reply 0

Corey Middleton Jan 23, 2018

also zoe -

Users tagged:

Reply 0

Gabriel Raubenheimer Jan 23, 2018

Thanks for the comment Corey. Both excellent points.

1) No, but this is something I'm passionate about developing implementing. Without it, we are going to end up with poor quality data that is geographically limited in its usefulness. There are several things to consider when doing this, ie. time between measurements, sample sizes for each measurement, standards for what defines a 'dead' plant, etc. Definitely a powerful objective for Feb.

2) Very good point. This is definitely important to track. In Cambodia, we are experimenting with several soil treatments and will need to track this and correlate with plant metric data we receive. Whilst we (and Fiji I think) are working organically, soil ameliorants in this context will be the cause, and plant metrics will be the effect. As you say, both must be measured. Thank you for raising this, I'm working on database design and will be sure to integrate that. (If you're interested: https://projecteverest.crowdicity.com/post/481529).

Cheers!

Users tagged:

Reply 1

Zoe Cahill Jan 23, 2018

Another point that Corey brought up while we were talking, was looking into how big farming companies measure, collate and analyse their data from farms, either back in Australia or Internationally. This could be a potential avenue to alleviate pressure off creating an entirely new system, but rather adapting some strategies that would be more effective at gathering the data that we need and entering it into our database.

Reply 0

Zoe Cahill Jan 23, 2018

Another point that Corey brought up while we were talking, was looking into how big farming companies measure, collate and analyse their data from farms, either back in Australia or Internationally. This could be a potential avenue to alleviate pressure off creating an entirely new system, but rather adapting some strategies that would be more effective at gathering the data that we need and entering it into our database.

Reply 1

Gabriel Raubenheimer Jan 24, 2018

Excellent point. It may be that that provides an immediately (or quickly) accessible and highly effective system.

Keep in mind that the work needed to adapt a system to our needs, contexts, resources, goals etc. may outweigh the benefit, but this is definitely something we should explore. It will definitely reveal things we are missing too.

Reply 0

Ciaran Hoare Jan 23, 2018

RE: 1)
Liv is thinking one step back from this around what to do for field dev and farm set up. This will be posted when it's done, but system around ensuring accurate tracking across all countries is definitely something we should have.

Reply 1

View all replies (6)

Mallory Dobner 7 months ago

Status label added: Work Update

Reply 0

Gabriel Raubenheimer 7 months ago

Status label added: Proposal adopted

Reply 0

Gabriel Raubenheimer 7 months ago

Status label removed: Work Update

Reply 0

Share