Order delivery pattern/methodology seems sub optimal?


It seems illogical to send out large shipments first.

I would have thought that it would be better for all concerned to follow the classic programmers pattern of lazy loading. Which is to say that data/content is sent to clients on an as needed basis.

In the below link Soylet is shipping 32 weeks supply to only two users (trent,AMasterOfNone).

If a lazy load approach had been followed Soylent could ship 4 boxes to 8 customers.

As a result Soylent would have 8 happy customers (rather than 2) and another 2-4 weeks to optimize the workflow to get the product out to those eight.

With regard to the argument that it’s more efficient to send 32 boxes to two people because then Soylent doesn’t need to worry about them for 16 weeks. That’s not a good argument, because one way or another there is only a certain amount of Soylent to go round.

It would be easy to calculate a lazy load, round robin, algorithm that would maximize the number of customers that soylent could be sent to right now (in batches of 4) while also keeping up with production to keep them fulfilled in an ongoing basis.

If Soylent had used a lazy load approach they could have satisfied 2-4 times more customers. And would therefore have 2-4 more times the network effect from influencers such as @vanclute and myself.


BTW, The algorithm to do this, is this…

Number of customers we should send soylant to right now = number of 1 month supply boxes we expect to create by the end of this month.

This assumes that monthly production will continually be on the rise.

Note: As the months roll by the larger orders would become fulfilled, and those members would also get the benefit of fresher soylent in later months.

So, for example, this month we can send Soylent to 1000 customers
Next month 1500
Following Month 2000
And so on

ie The rolling customer base is horizontally larger from the get go and keeps increasing month to month.


Except they wouldn’t be able to do reorders in a timely fashion.

These people aren’t dumb. They’re programmers, too. They’ve thought through this.


@web_kunoichi yes they would. As explain above.

I understand you want to defend them, but when removing all bias, the above pattern is a more logical distribution methodology.

Programmers often make mistakes and create sub optimal code in the first place and then re-factor it later. It doesn’t make them dumb, all (sensible) programers would agree that refactoring is a major part of programming.


Because penalizing the larger orders is “optimal”?


@web_​kunoichi The larger orders wouldn’t become penalized. Assuming the order was for one user, they would have exactly the same amount of Soylent for daily use over the 16, or 32 weeks whether they got it all upfront, or whether they got it sent one month at a time.

Actually, it would be a benefit for large orders as they would have fresher Soylent on a month to month basis.

(so, more people get to use Soylent, more people happy)


ok, so lets say they produce 100 weeks of soylent per week, just as an arbitrary number to work with, and can increase productivity by 25 weeks per week per week (what kind of units am I toying with here?)
lets say the largest order is 25 weeks.
4 25s, 12 10s, 30 5s, 87 2s, 200 1s, for a total of 744 weeks

tactic 1: ship to largest orders first
week 1 (100 produced): we ship 4 shipments of 25.
week 2 (125 produced): we ship 12 shipments of 10
week 3 (150 produced): we ship 30 shipments of 5
week 4 (175 produced): we ship 87 shipments of 2
week 5 (200 produced): we ship 200 shipments of 1
week 6 (225 produced): fall short of 287 reorders

tactic 2: ship to smallest orders first
week 1 (100 produced): we ship 100 units of 1
week 2 (125 produced): we ship 25 new orders, and resupply everyone else.
week 3 (150 produced): we ship 25 new orders, and resupply everyone else
week 4 (175 produced): we ship 25 new orders, and resupply everyone else
week 5 (200 produced) we ship 25 new orders, and resupply everyone else.

so, in both of these scenarios, we satisfied our 200 single week orders at the end of week 5. the larger orders also supplied everyone else in the meantime. Sure, the smallest orders first gives you the most customers suppied initially, but the burden of resupplying them constantly severely hampers meeting all of the demands in a timely manner. This is assuming that everyone reorders 1 week too; if those initial 100 had reordered a months supply, then you have 400 weeks of demand to fulfill resupplies on week 2, and you can’t even handle your current load, much less more consumers.

Yes, this is a specific set of numbers that may have no bearing on the actual numbers, but I didn’t have to sit here and push numbers around to get this to work, this was the first set of numbers I tried. However, it does show that there does exist a set of numbers by which your premise is incorrect, and so the actual numbers could also be incorrect for your argument.

edit: oh, I misunderstood what you were proposing.

your algorithm assumes that everyone will consume soylent at the same rate. If someone ordered 2 months of soylent for 2 people, then getting it in over the course of 2 months is too slow; they need 2 months every month. Or if someone had a large order with the intent of sharing it with others, you have now cut off their excess.

Crowdfunding backers screwed

There is also the age old concept of the “back order”. There is no real reason you’d need to ship the entire amount ordered at once, other than to reduce per unit shipping costs. To bootstrap the process, a partial order of 1 month for larger orders might have been a good compromise.


@Mystify No a week is not a good iteration. I was suggesting monthly iterations. You also need to consider that people are fickle and a fair percentage (maybe 25%-50%) will not reorder after the first month, therefore making room for new people to join the party.


Their entire shipping estimate was only a month or two, and seemed to think they could get through entire order categories in a week or two. Yes, I didn’t factor in non-reorders, but it doesn’t change the basic dynamic that you had re-orders clogging up your supply chain before you get production ramped up.


These two things come to mind when I think about this…

  1. Do Things that Don’t Scale - http://paulgraham.com/ds.html

In other words, don’t worry about scaling in the early days, just worry about making people happy.

  1. The number one reason startups fail is because they scale prematurely (based on a study of 3,200 high growth tech startups)


@Mystify A 25%-50% churn rate weighs in pretty heavily! :wink:

Also note, to deal with the issue you bring up you could simply do this…

Number of customers we should send soylant to right now © = number of 1 month supply boxes we expect to create by the end of this month (s) … divided by reducing factor ®

(c = s) / r

Doesn’t really matter what you make r, the point is you can roll out to more customers than by following a strict rollout of only larger orders first.


OMG U GUYZZ! Taht won’t wurk bcuz… geeks!


I kid, I kid. I love that you’re all trying to algorithmically optimize their shipping department. Go crowdsourced solutions!


They should have just gone in order of orders received. That way nobody could complain about others getting their Soylent first. And they also should have made a one month order the maximum amount. This approach would have led to the most happy customers.

Why didn’t they follow the above approach? Well a certain percentage of customers are going to be disappointed in the final product for one reason or another. Either the taste, digestibility, texture, or lifestyle just won’t work for them. The Soylent team knows this. So they naturally accepted and even encouraged very large orders. There is nothing wrong with any of this of course, this is a business, and getting someone to buy 4 months of your product without even trying it first is quite an accomplishment! More power to them. Note, looking forward to my Soylent!


a 50% churn rate doesn’t overcome the drastic difference between the strategies.

and you will strand a lot of them without enough soylent, which is something they are prioritizing against. If you start soylent, you should be able to keep using it without a hitch.

Also, most of the customers are <1 month of shipments. Most of the soylent is in <1 month orders. This strategy does nothing to help distribute to 1 month orders, as you ship once and they have their initial order met already.
The actual gains from a round robin system are fairly minor in the grand scheme of shipping soylent, the main thing it accomplishes would be adding complexity to the management of it, which will lead to errors, which will lead to waste, which will counteract some of it.


Makes me wonder if there might be at least a few people who ordered multiple months’ worth, don’t care for it, and will want to unload it. Seems like ebay ought to start seeing Soylent by the box any time now if this is indeed the case!


I can guarantee you there is a more optimal way to distribute based on all the criteria that needs to be met than to just ship largest orders first. You’re going to make me model this in a spreadsheet to prove this, aren’t you. :wink:


This will happen, no doubt! And it might not even be because they don’t like the Soylent, it could be a way to make some money for their next order Soylent. Just put one month worth of single bags on ebay, and use the other three months yourself.


Yeah, I think the problem with this approach, while seemingly brilliant on the surface, is that it is not sustainable for the first round of pre-orders, and would make it harder to Rosa to move into the next phase of shipping (re-orders and subscriptions).

Let’s say in the first shipping period you ship the first month of multiple-month orders to X number of people. You’ve got enough Soylent left to ship the entire orders of Y number of people. Let’s assume that Y is smaller than X, given supply constraints.

Then, during the next shipping period (or the one after that, maybe), you now have X’s second month orders to deal with, plus another Y group. Each month, X gets smaller and Y gets larger, until everyone has their Soylent.

The benefit is that more people get thier Soylent in the first month, but that’s assuming that the constraint is the amount of Soylent on hand. Perhaps the bottleneck is actually the number of shipments (logisitcs, management, etc.). In fact, it might be nearly as fast to ship 4 boxes to one customer as it would be to ship just 1. In that case, your plan provides literally no benefit, and instead create a non-zero percentage of additional overhead, due to the larger number of shipments.

But we can’t be sure of that, you may be right. But, either way, there are multiple obvious downsides:

  • Considerably larger number of overall shipments, at a greater total cost.

  • While a number of smaller orders would ship sooner, the total pre-order shipping period would not be much shorter (if any shorter), and due to the larger number of shipments, the final pre-orders could actually go out even later than planned.

  • The early smaller orders (the first Y group) are now starting their re-orders sooner than before, which means even more shipments during the pre-order shipping period, meaning both greater risk of shipping delays, and considerably less tolerance for shipping delays.

  • Multiple-month orders would also be starting their subscriptions immediately, which again, introduces more risk earlier in the process as well as less tolerance for failure.

  • Many multiple-month orders are actually intended for multiple people in a single household. These people would be negatively impacted. Some would wait for their second shipment to begin using their Soylent, which would defeat the purpose and in effect put them at the back of the line but without giving up their place in line to anyone else.

  • The pre-order shipping period would overlap entirely with the on-going subscription shipping stage, robbing the Soylent team of any opportunity to re-group or apply lessons learned.

Now, the longer the pre-order shipping period, the greater each these problems become. However, assuming the best case scenario that there are only one or two “Y” groups, and that all of the pre-orders are shipped by the end of May, before anyone needs to reorder (fingers crossed), then these issues would be reduced but not eliminated, but… any practical benefit of your proposal would also be reduced. The shorter it takes to ship out the first orders, the less value this idea provides, while on the order hand, the longer it takes to ship out the first orders, the worse this idea becomes. It’s a lose-lose.

Ironically, your post about the pitfalls of scaling too early apply directly to your plan, which requires Rosa Labs to scale into production mode immediately. Fulfilling pre-orders by size allows them to focus on making their first customers (as a total group) happy in the simplest way possible with the lowest amount of overhead, before pivoting into sustainable production.


‘As needed?’

But everyone “needs” it all at once, and parceling out smaller shipments to a wider audience is LESS efficient when you’re talking about spatial manipulation of matter instead of computational power.

Programming algorithms work well in theory, but do not translate well into the physical world, where economies of scale are very real.