Statistics about the amount of saved food

discussion
feedback

#1

@tiltec is working on implementing statistics for the amounts of food saved. So far it is only per store but group-wide statistics will follow as soon as we clarify how the statistics should work internally.

The statistics are dependent on the feedback given, of course, but we’re thinking of making them a little smarter. We could, for example, add in a way for the software to estimate the amount of pickups where no feedback was given, based on the average of the pickups before and after. What do you think?

Another thing is the potential confusion about the amount that should be entered: Is it the weight of everything the store gave out or only the weight of the stuff I personally took with me? And if it is the weight of the full pickup (as we actually intended it to be), what happens then if two collectors give feedback? Shall we calculate that out? What if those feedbacks given are really different, which one to take into account then?

Lots of questions, as you can see, and you’re invited to raise more if you see some… Would you be so kind to share your perspective? :slight_smile:


#2

This sounds like a good idea. I feel that a very little amound of people wants to insert what they picked up, also because it’s not entirely easy to guess. This function should then maybe be connected to the day of the pick up, because mondays for example tend to give a greater amound of food to save.

This could simply be defined as a headline, I’d say the total amound of edible food would be interesting. If two people give feedback, this could just e averaged.

At the moment the feeback function doesn’t seem to be very interesting, at least at FS Östersund the members hardly give feedback. I can understand this, because we only have two stores and the amount stays morfe or less the same.


#3

Currently, we only calculate the sum of feedback weight. If two people gave weight for the same pickup, it is being averaged. The idea from @Teddy to calculate average weight per weekday sounds interesting, I’ll look into it when we have a bit more feedback gathered on karrot.world.

I’d love to make good estimates about the amount of food picked up, but I fear that it’s quite hard. There’s so many things that karrot usually doesn’t know about, for example holiday seasons, big events or the employees having a bad day…


#4

I think the feedback function is a very important feature, as it is essential to account the amount of food that was saved.It is already very nice to have the number of pickups, number of feedbacks and amount of saved food. With this data it is also possible to calculate some stats (e.g. monthly) manually …

What is a problem here in Taiwan is that many people seem to have problems to estimate the amount of food they have picked up. Like sometimes I see pictures with huge amounts of food and when I check the feedback it says 500 g… and people have gotten back to me and asked for help. Some people even asked me to provide scales at the stores, but I think we will try to maybe share some example photos to give them ideas of the weight.


#5

I know it’s a very old topic but I’m interested in stats and this odd fancy has brought me here. I have a question regarding this comment:

If two people gave weight for the same pickup, it is being averaged.

What is the logic behind this? In Warsaw when multiple people give feedback in Karrot, they put weight of what each of them picked up. So their weights shouldn’t be averaged but summed up. Sometimes it happens that only one person gives feedback with weight and then it represents total weight. In the light of the above, can we switch from average to sum to have better stats in Karrot?

I personally don’t think Karrot should do any estimates in terms of weights for pickups without feedback. No feedback may mean 0 kg or 100 kg. How Karrot is supposed to know this? Averages per weekday are just averages based on the history. They might be very wrong and stats should be based on facts not estimates.

There is a simple solution to have proper stats in Karrot. Request all Karrot users to add weights to their feedback. In Warsaw almost 100% of pickups have feedbacks in Karrot including weights. This is a commonly accepted rule to add feedback after the pickup and nobody questions the need to do that. I believe this discipline is much easier to achieve than creating a forecasting algorithm inside Karrot to predict weights of historical pickups.


#6

Yeah, that’s also something that gives me headache sometimes. I search through our history. I implemented it in this Pull Request, but the decision was already made before when the feedback feature was added. From this comment:

Should the weight be about the whole pick-up or just the amount that the foodsaver took home?
It becomes relevant when calculating statistics.

Reasons for asking for the weight …
(1) of the whole pick-up:

  • maybe some people have better overview/are better at estimating
  • if multiple users give feedback for the same pick-up event, we could calculate the average

(2) only of the amount the foodsaver took home

  • seems easier to estimate smaller amounts, which might give more accurate statistics
  • conflicting feedback about the total amount are avoided

As @bruno summarized in this comment, we seemed to favor (1) at that time. Back then, we wanted to implement estimates for the average weight per pickup and extrapolate.

The feedback weight input says “Amount of food all of you were able to save (in kg)”. But it seems it’s difficult to tell them otherwise, probably because it’s too difficult for them to estimate the weight of the whole pickup.

I would consider it, although not retroactively. There are many people who entered the numbers the way Karrot told them to, and I wouldn’t want to break their feedback.

Fair point, and I think our planning came from a time when we expected only a small number of users to enter feedback. Now that it’s different and we should reconsider.

Advantage of not doing extrapolation:

  • users have more reason to give feedback; might give more accurate numbers
  • it is essentially the status quo
  • less complexity; easier to understand

Potential problems of having a direct connection between feedback weight and weight statistics:

  • users might be tempted to set higher weight to make their pickup statistics look nicer (user statistics do not exist yet, but well…)
  • variance of estimation bias between different people doesn’t cancel out

Overall, I think it would be a good choice to “say no” to more cleverness in software, it might get in the way of humans. Rather accept human error and try to nudge them in a good direction (e.g. by showing hints how to estimate weight).


#7

You are right. I’m using Polish version of Karrot and this sentence has been translated to something like: “Amount of food you saved (in kg)” which doesn’t indicate it should be a total for the whole pickup. The translation has introduced confusion here, however, I still favour the option to put weight of the food each person saved not a total amount :slight_smile:

Makes sense. I agree.

I agree 100%!


#8

I totally agree and now I actually don’t know what we should do in Warsaw. Since always we have been underestimating what we’re saving. Well, bad, but cannot be changed. What should we do not? Should we tell everyone that it works differently than we thought??? Or rather just wait until it works as we always thought it is? Going back and forth will be difficult - any change takes a lot of time to get used to it. What do you think? It of course depends on how long this would be.


#9

Hard to say. Maybe really just wait?

I think rather software should change than people. So we just need more people to change the software :wink: