6 ms·

> 4. You can't A/B test a farm (yet) This is done all the time. When evaluating a new seed variety, a different seed spacing, a new nitrogen application, farme

by coin 4y ago

> 4. You can't A/B test a farm (yet) This is done all the time. When evaluating a new seed variety, a different seed spacing, a new nitrogen application, farmers will do a test plot next to the control. If things looks good then they'll expand the new change to more acreage.

MonkeyMalarky 4y ago
This is really funny to me after living near a place that has done exactly that for ~100 years: https://en.wikipedia.org/wiki/Central_Experimental_Farm https://en.wikipedia.org/wiki/Central_Experimental_Farm
- monsieurbanana 4y ago
  Or the Incas, experimenting with micro climates 500 years ago https://en.m.wikipedia.org/wiki/Moray_(Inca_ruin) https://en.m.wikipedia.org/wiki/Moray_(Inca_ruin)
- fastaguy88 4y ago
  Or almost 150. https://en.m.wikipedia.org/wiki/Morrow_Plots https://en.m.wikipedia.org/wiki/Morrow_Plots The US Dept of Agriculture funds comparative research and agricultural extension agents help put it into practice. They actually know quite a bit about technology adoption in agriculture.
kcb 4y ago
That is a pretty wild statement. I would assume A/B testing in farming has happened for centuries. Farms are made up of a multitude of fields. And so are a perfect environment for A/B testing. I did A in one field but stuck with method B in the other and saw that the former crops had higher yield, etc. etc.
jvm___ 4y ago
There's farms near me that have a field of corn, every 5 rows has a different sign up with some sort of code and a logo for a seed company. I'd always assumed thats what they were testing.
- robbiep 4y ago
  Those are field trials of different varieties. It’s not at scale
  - Turing_Machine 4y ago
    Could also be producing hybrid seed.
    - drewcoo 4y ago
      This. The big clue is signs for a seed company. As a teen I detasseled corn for a couple of summers - pulling the male tassels off the female rows of corn to create hybrids.
    - jcampbell1 4y ago
      I have worked on a corn seed farm detassling odd rows of corn. It is every other row for seed hybridization. The rows were marked, but it was pretty obvious because the two heirloom varieties were distinct. If there were more that two varieties, it is not seed production.
      - hinkley 4y ago
        I used to live in a town in farm country and saw a lot of it from the back of a bicycle. There’s not much to look at when the corn is high except the corn. I don’t know if things have changed in the last couple of decades but it was pretty common for detassling to remove the pollen from four or five rows at a go, rather than odd/even pairings.
bagels 4y ago
I came in to make the same point. You can reserve a small part of your crop for the experiment. This is exactly akin to how social media companies do growth hacking. Try something new out on x% of the users/crops, measure the results. You don't have to risk your whole farm on it, unless there is a huge capital expense on equipment.
Turing_Machine 4y ago
Some of the newer farm equipment can actually apply different amounts of fertilizer (or whatever) to different parts of the same field, based on GPS and a grid of soil samples.
- whalesalad 4y ago
  This is precision ag and it’s been around for quite some time.
  - hinkley 4y ago
    Someone wa pitching a project for this when I was at NCSA (which was less than two miles from corn fields) in the 90’s, as Netscape was taking over browser development and they needed something else to do (NSF is not fond of competing with commercial interests.) That would have been an excellent match but I was barely interested and others were less enthusiastic. They started a collaborative software project instead. Which was cool but CRDT wouldn’t be invented for another five to ten years so I wonder what happened with that project.
HPsquared 4y ago
It's hard to have a truly controlled experiment though. The result will be valid for that farm, on that year - but next year the conditions will be different, etc.
- Kamq 4y ago
  That's true of all A/B testing for real world things isn't it? You could have measured increased traffic because something was trending on twitter/HN instead of the UI changes you just deployed. Might be completely different tomorrow.
- jeltz 4y ago
  That is true for A/B tests too in tech.
hinkley 4y ago
I’m in the permaculture community and most of these guys have stories about their neighbors looking at them like they’re crazy for years until there’s a catastrophic event and they see the land across their fence is doing better than theirs and only then do they start asking for advice, which they then take very slowly. For the people who think this is where we need to go to fight soil loss and global warming, this sort of learning by osmosis is very, very frustrating. People blame these guys for making their money off of speaking engagements and consulting, but it’s going to take a lot to get to the point where your neighbors are on board. If you want to teach someone you have to seek them out, not just work locally.
vharuck 4y ago
If I picked up a random statistics textbook printed before 1990, I'd be very surprised to not see numerous word problems about crop treatments.

by coin 4y ago

MonkeyMalarky 4y ago

This is really funny to me after living near a place that has done exactly that for ~100 years: https://en.wikipedia.org/wiki/Central_Experimental_Farm https://en.wikipedia.org/wiki/Central_Experimental_Farm

monsieurbanana 4y ago
Or the Incas, experimenting with micro climates 500 years ago https://en.m.wikipedia.org/wiki/Moray_(Inca_ruin) https://en.m.wikipedia.org/wiki/Moray_(Inca_ruin)
fastaguy88 4y ago
Or almost 150. https://en.m.wikipedia.org/wiki/Morrow_Plots https://en.m.wikipedia.org/wiki/Morrow_Plots The US Dept of Agriculture funds comparative research and agricultural extension agents help put it into practice. They actually know quite a bit about technology adoption in agriculture.

kcb 4y ago

That is a pretty wild statement. I would assume A/B testing in farming has happened for centuries. Farms are made up of a multitude of fields. And so are a perfect environment for A/B testing. I did A in one field but stuck with method B in the other and saw that the former crops had higher yield, etc. etc.

jvm___ 4y ago

There's farms near me that have a field of corn, every 5 rows has a different sign up with some sort of code and a logo for a seed company. I'd always assumed thats what they were testing.

robbiep 4y ago
Those are field trials of different varieties. It’s not at scale
- Turing_Machine 4y ago
  Could also be producing hybrid seed.
  - drewcoo 4y ago
    This. The big clue is signs for a seed company. As a teen I detasseled corn for a couple of summers - pulling the male tassels off the female rows of corn to create hybrids.
  - jcampbell1 4y ago
    I have worked on a corn seed farm detassling odd rows of corn. It is every other row for seed hybridization. The rows were marked, but it was pretty obvious because the two heirloom varieties were distinct. If there were more that two varieties, it is not seed production.
    - hinkley 4y ago
      I used to live in a town in farm country and saw a lot of it from the back of a bicycle. There’s not much to look at when the corn is high except the corn. I don’t know if things have changed in the last couple of decades but it was pretty common for detassling to remove the pollen from four or five rows at a go, rather than odd/even pairings.

bagels 4y ago

I came in to make the same point. You can reserve a small part of your crop for the experiment. This is exactly akin to how social media companies do growth hacking. Try something new out on x% of the users/crops, measure the results. You don't have to risk your whole farm on it, unless there is a huge capital expense on equipment.

Turing_Machine 4y ago

Some of the newer farm equipment can actually apply different amounts of fertilizer (or whatever) to different parts of the same field, based on GPS and a grid of soil samples.

whalesalad 4y ago
This is precision ag and it’s been around for quite some time.
- hinkley 4y ago
  Someone wa pitching a project for this when I was at NCSA (which was less than two miles from corn fields) in the 90’s, as Netscape was taking over browser development and they needed something else to do (NSF is not fond of competing with commercial interests.) That would have been an excellent match but I was barely interested and others were less enthusiastic. They started a collaborative software project instead. Which was cool but CRDT wouldn’t be invented for another five to ten years so I wonder what happened with that project.

HPsquared 4y ago

It's hard to have a truly controlled experiment though. The result will be valid for that farm, on that year - but next year the conditions will be different, etc.

Kamq 4y ago
That's true of all A/B testing for real world things isn't it? You could have measured increased traffic because something was trending on twitter/HN instead of the UI changes you just deployed. Might be completely different tomorrow.
jeltz 4y ago
That is true for A/B tests too in tech.

hinkley 4y ago

I’m in the permaculture community and most of these guys have stories about their neighbors looking at them like they’re crazy for years until there’s a catastrophic event and they see the land across their fence is doing better than theirs and only then do they start asking for advice, which they then take very slowly. For the people who think this is where we need to go to fight soil loss and global warming, this sort of learning by osmosis is very, very frustrating. People blame these guys for making their money off of speaking engagements and consulting, but it’s going to take a lot to get to the point where your neighbors are on board. If you want to teach someone you have to seek them out, not just work locally.

vharuck 4y ago

If I picked up a random statistics textbook printed before 1990, I'd be very surprised to not see numerous word problems about crop treatments.