|
 |
| | | Author: | Rakali  | | Posted: | Aug 30, 2020 06:00 | | Subject: | Auto-select is awful. A different solution. | | Viewed: | 145 times | | Topic: | Suggestions | | Status: | Open | | Vote: | [Yes|No] | |
|
| Long post incoming.
As someone who does like to put together the occasional MOC from purchased instructions
as well as sell, I am having a hard time understanding how anyone manages to
buy parts these days.
Here are a few of my results from using the auto-select feature on a wanted list
of ~2500 parts in 374 lots.
UK ONLY, straight auto-select:
369/374 lots, 16 stores, £290.88
UK ONLY, manually filling from my own store first, then hitting auto-select:
369/374 lots, 14 stores, £270.19
I also tried the same list from a different account with the same results, so
it isn't just excluding my own store from my own search results.
I understand it's a very difficult programming task especially for such a
large wanted list, however with just one store manually selected, I was able
to reduce the number of stores by 2, and therefore shipping costs, as well as
the parts price by 7%, so why was my store not recommended to me in the first
place?
Let's try something else, and expand our search to EU stores.
EU ONLY, straight auto-select:
369/369 lots, 4 stores, £319.71
To keep it fair I removed the 5 pieces I could not get from UK sellers from the
query.
Now we are at 10 less stores than my optimized UK result, but for £50 more. UK
shipping is not going to be anywhere close to £5 per store, so we have given
the finder more choice, and somehow came away with a worse result.
This isn't even accounting for the fact that most EU stores charge a 5% Paypal
fee that the auto-buy can't see, and also the increased shipping cost for
international transactions.
Let's try and find those hard to find pieces.
UK ONLY, straight auto-select, then EU for the missing 5 parts:
374/374 lots, 20 stores, £332.33
EU ONLY, straight auto-select:
374/374 lots, 4 stores, £438.67
Now of course with 16 less stores, I am saving a considerable amount of shipping,
however we are not accounting for the additional 5% Paypal fees from most EU
stores. So average £2 shipping across the UK stores, add 5% to the EU store prices,
and I am still saving £100 by giving the auto-select LESS choice.
Maybe it is just not designed for such a large number of lots, but my list was
just the parts for two medium sized Star Wars models from Brickvault. That seems
like it would be a popular use case.
So what is the solution?
Work backwards. Start with the stores, then find the parts.
With so many stores having a whole myriad of different fees and limits these
days, it makes more sense to start with stores you are comfortable buying from,
and work from there.
I have attached an image of a piece of software that was written by BobDeQuatre
in 2013. It would load your wanted list, then you could give it stores you wanted
to buy from and it would cross reference the prices of all the parts between
the stores. You could then select which parts to buy from each and create the
relevant wanted lists. (if this was built in to BL you could of course just create
carts).
Sadly the software soon broke due to changes on BLs end before it ever had time
to gain traction, and the author did not care to work on the project any longer.
I used this to put together this MOC:
https://rebrickable.com/mocs/MOC-6703/thire5/ucs-arc-170-starfighter/#comments
It was by far the easiest buying experience I have ever had as well as being
the largest MOC I have put together with parts from BL.
You can use the current features to find some stores with a lot of parts you
need with fair prices and terms that an auto finder cannot account for, put them
into this software and cross reference.
Need a rare part? Find a store manually that has it, put that into your table.
That store also had a few of your other parts cheaper than your other chosen
stores? Accounted for.
Unfortunately I am not a programmer, nor do I have the funds that it would take
to hire one, but I think this solution is worth something. I can offer up a token
£100 to anyone that would like a bit of a coding project to dabble with that
could be very beneficial for buyers and sellers alike.
Sorry for the ramble. Any thoughts?
|
|
|
|
| | | | | |
| | | | | Author: | yorbrick | | Posted: | Aug 30, 2020 08:38 | | Subject: | Re: Auto-select is awful. A different solution. | | Viewed: | 54 times | | Topic: | Suggestions | |
|
| It has long been a problem, and nobody (users) really knows what the algorithm
is to select stores. Is it IC, size of store, number of lots they have (even
if this includes duplicates), location, feedback, ...
Also it does not know about shipping costs from stores without IC, or anything
like 5% fees for PayPal and so on.
Whenever I want to buy from a wants list, I always go for the expensive parts
first, find a (typically) UK store that has them at a reasonable price then check
if they have other parts at a reasonable cost. Manually takes time, but is often
way better than any solution BL comes up with.
|
|
| | | | | | | | | |
| | | | | | | Author: | SylvainLS | | Posted: | Aug 30, 2020 10:04 | | Subject: | Re: Auto-select is awful. A different solution. | | Viewed: | 52 times | | Topic: | Suggestions | |
|
| In Suggestions, yorbrick writes:
| | It has long been a problem, and nobody (users) really knows what the algorithm
is to select stores.
|
I’m not sure non-users know either  
| | Is it IC, size of store, number of lots they have (even
if this includes duplicates), location, feedback, ...
|
It would be conforting to know but would it really help predict or be happier
with the results?
With a multi-criterion valuation, especially when some criteria are “soft” (not
originally a number), you’ll always want to tinker with the weights.
| | Also it does not know about shipping costs from stores without IC, or anything
like 5% fees for PayPal and so on.
|
AKAIK, it doesn’t directly know about any shipping cost, not even IC.
It’s all estimation, an “average” of the “recent” shipping costs between seller’s
address/country (how precise?) to buyer’s.
| | Whenever I want to buy from a wants list, I always go for the expensive parts
first, find a (typically) UK store that has them at a reasonable price then check
if they have other parts at a reasonable cost. Manually takes time, but is often
way better than any solution BL comes up with.
|
I’d say you’re right but are you sure or is it just a feeling? I mean, do you
spend even more time checking every time or do you just always go for the manual
option and never bother with Auto-select?
Or is it not really possible to check because you ended up buying things that
were not on you WL (so, do you now count them as wanted?).
And how much do you value your time?
Auto-select results vary from simple to double ( https://www.bricklink.com/message.asp?ID=1218711
) but, if you try multiple times and add a bit of manual optimization, in the
end, is not Auto-select “good-enough” for many people?
|
|
|
| | | | | |
| | | | | Author: | DanialR | | Posted: | Aug 31, 2020 22:14 | | Subject: | Re: Auto-select is awful. A different solution. | | Viewed: | 52 times | | Topic: | Suggestions | |
|
| | | I have attached an image of a piece of software that was written by BobDeQuatre
in 2013. It would load your wanted list, then you could give it stores you wanted
to buy from and it would cross reference the prices of all the parts between
the stores. You could then select which parts to buy from each and create the
relevant wanted lists. (if this was built in to BL you could of course just create
carts).
Sadly the software soon broke due to changes on BLs end before it ever had time
to gain traction, and the author did not care to work on the project any longer.
|
| | Unfortunately I am not a programmer, nor do I have the funds that it would take
to hire one, but I think this solution is worth something. I can offer up a token
£100 to anyone that would like a bit of a coding project to dabble with that
could be very beneficial for buyers and sellers alike.
|
This might fit in with anther pet software project of mine. How much info do
you have on the software and whether BobDeQuatre is willing to let others run
with something he spent time building?
|
|
|
| | | | | |
| | | | | Author: | chuckondrums | | Posted: | Oct 9, 2020 13:01 | | Subject: | Re: Auto-select is awful. A different solution. | | Viewed: | 25 times | | Topic: | Suggestions | |
|
| In Suggestions, Rakali writes:
| | Long post incoming.
As someone who does like to put together the occasional MOC from purchased instructions
as well as sell, I am having a hard time understanding how anyone manages to
buy parts these days.
Here are a few of my results from using the auto-select feature on a wanted list
of ~2500 parts in 374 lots.
UK ONLY, straight auto-select:
369/374 lots, 16 stores, £290.88
UK ONLY, manually filling from my own store first, then hitting auto-select:
369/374 lots, 14 stores, £270.19
I also tried the same list from a different account with the same results, so
it isn't just excluding my own store from my own search results.
I understand it's a very difficult programming task especially for such a
large wanted list, however with just one store manually selected, I was able
to reduce the number of stores by 2, and therefore shipping costs, as well as
the parts price by 7%, so why was my store not recommended to me in the first
place?
Let's try something else, and expand our search to EU stores.
EU ONLY, straight auto-select:
369/369 lots, 4 stores, £319.71
To keep it fair I removed the 5 pieces I could not get from UK sellers from the
query.
Now we are at 10 less stores than my optimized UK result, but for £50 more. UK
shipping is not going to be anywhere close to £5 per store, so we have given
the finder more choice, and somehow came away with a worse result.
This isn't even accounting for the fact that most EU stores charge a 5% Paypal
fee that the auto-buy can't see, and also the increased shipping cost for
international transactions.
Let's try and find those hard to find pieces.
UK ONLY, straight auto-select, then EU for the missing 5 parts:
374/374 lots, 20 stores, £332.33
EU ONLY, straight auto-select:
374/374 lots, 4 stores, £438.67
Now of course with 16 less stores, I am saving a considerable amount of shipping,
however we are not accounting for the additional 5% Paypal fees from most EU
stores. So average £2 shipping across the UK stores, add 5% to the EU store prices,
and I am still saving £100 by giving the auto-select LESS choice.
Maybe it is just not designed for such a large number of lots, but my list was
just the parts for two medium sized Star Wars models from Brickvault. That seems
like it would be a popular use case.
So what is the solution?
Work backwards. Start with the stores, then find the parts.
With so many stores having a whole myriad of different fees and limits these
days, it makes more sense to start with stores you are comfortable buying from,
and work from there.
I have attached an image of a piece of software that was written by BobDeQuatre
in 2013. It would load your wanted list, then you could give it stores you wanted
to buy from and it would cross reference the prices of all the parts between
the stores. You could then select which parts to buy from each and create the
relevant wanted lists. (if this was built in to BL you could of course just create
carts).
Sadly the software soon broke due to changes on BLs end before it ever had time
to gain traction, and the author did not care to work on the project any longer.
I used this to put together this MOC:
https://rebrickable.com/mocs/MOC-6703/thire5/ucs-arc-170-starfighter/#comments
It was by far the easiest buying experience I have ever had as well as being
the largest MOC I have put together with parts from BL.
You can use the current features to find some stores with a lot of parts you
need with fair prices and terms that an auto finder cannot account for, put them
into this software and cross reference.
Need a rare part? Find a store manually that has it, put that into your table.
That store also had a few of your other parts cheaper than your other chosen
stores? Accounted for.
Unfortunately I am not a programmer, nor do I have the funds that it would take
to hire one, but I think this solution is worth something. I can offer up a token
£100 to anyone that would like a bit of a coding project to dabble with that
could be very beneficial for buyers and sellers alike.
Sorry for the ramble. Any thoughts?
|
It seems like it would be in the wheelhouse of Bricklink to use the algorithm
to find parts at a user specified price-per-part average. You may end up with
very many stores, but if the price is good enough, the shipping and fees could
still be under the total costs the algorithm is putting out now. It would give
users a better handle on the true costs of what they're buying. Plus it
would demotivate higher pricing for larger inventories.
|
|
|
| | | | | | | | | |
| | | | | | | Author: | SylvainLS | | Posted: | Oct 9, 2020 13:23 | | Subject: | Re: Auto-select is awful. A different solution. | | Viewed: | 39 times | | Topic: | Suggestions | |
|
| In Suggestions, chuckondrums writes:
| | […]
It seems like it would be in the wheelhouse of Bricklink to use the algorithm
to find parts at a user specified price-per-part average.
|
You mean a total price?
Because you can already set a Max Price to all your items and tell Auto-select
to only consider lots under that Max Price.
Okay, “max” is not “average” but to calculate an average, you need a total, so
if you set a price for each item, it has to be a max.
| | You may end up with very many stores,
|
And many parts not found…
| | but if the price is good enough, the shipping and fees could
still be under the total costs the algorithm is putting out now.
|
How so? 
The current algorithm tries to minimize the total price (estimated shipping included).
You want it to find items at the user’s specified prices (“on average”).
If these user-set prices are lower than what it could find now, it won’t find
them.
And if these user-set prices are higher than what it could find now, how could
the total be lower than what it finds now?
| | It would give
users a better handle on the true costs of what they're buying. Plus it
would demotivate higher pricing for larger inventories.
|
|
|
|
| | | | | |
| | | | | Author: | Scharnvirk | | Posted: | Oct 16, 2020 11:28 | | Subject: | Re: Auto-select is awful. A different solution. | | Viewed: | 52 times | | Topic: | Suggestions | |
|
| In Suggestions, Rakali writes:
| | Long post incoming.
As someone who does like to put together the occasional MOC from purchased instructions
as well as sell, I am having a hard time understanding how anyone manages to
buy parts these days.
Here are a few of my results from using the auto-select feature on a wanted list
of ~2500 parts in 374 lots.
UK ONLY, straight auto-select:
369/374 lots, 16 stores, £290.88
UK ONLY, manually filling from my own store first, then hitting auto-select:
369/374 lots, 14 stores, £270.19
I also tried the same list from a different account with the same results, so
it isn't just excluding my own store from my own search results.
I understand it's a very difficult programming task especially for such a
large wanted list, however with just one store manually selected, I was able
to reduce the number of stores by 2, and therefore shipping costs, as well as
the parts price by 7%, so why was my store not recommended to me in the first
place?
Let's try something else, and expand our search to EU stores.
EU ONLY, straight auto-select:
369/369 lots, 4 stores, £319.71
To keep it fair I removed the 5 pieces I could not get from UK sellers from the
query.
Now we are at 10 less stores than my optimized UK result, but for £50 more. UK
shipping is not going to be anywhere close to £5 per store, so we have given
the finder more choice, and somehow came away with a worse result.
This isn't even accounting for the fact that most EU stores charge a 5% Paypal
fee that the auto-buy can't see, and also the increased shipping cost for
international transactions.
Let's try and find those hard to find pieces.
UK ONLY, straight auto-select, then EU for the missing 5 parts:
374/374 lots, 20 stores, £332.33
EU ONLY, straight auto-select:
374/374 lots, 4 stores, £438.67
Now of course with 16 less stores, I am saving a considerable amount of shipping,
however we are not accounting for the additional 5% Paypal fees from most EU
stores. So average £2 shipping across the UK stores, add 5% to the EU store prices,
and I am still saving £100 by giving the auto-select LESS choice.
Maybe it is just not designed for such a large number of lots, but my list was
just the parts for two medium sized Star Wars models from Brickvault. That seems
like it would be a popular use case.
So what is the solution?
Work backwards. Start with the stores, then find the parts.
With so many stores having a whole myriad of different fees and limits these
days, it makes more sense to start with stores you are comfortable buying from,
and work from there.
I have attached an image of a piece of software that was written by BobDeQuatre
in 2013. It would load your wanted list, then you could give it stores you wanted
to buy from and it would cross reference the prices of all the parts between
the stores. You could then select which parts to buy from each and create the
relevant wanted lists. (if this was built in to BL you could of course just create
carts).
Sadly the software soon broke due to changes on BLs end before it ever had time
to gain traction, and the author did not care to work on the project any longer.
I used this to put together this MOC:
https://rebrickable.com/mocs/MOC-6703/thire5/ucs-arc-170-starfighter/#comments
It was by far the easiest buying experience I have ever had as well as being
the largest MOC I have put together with parts from BL.
You can use the current features to find some stores with a lot of parts you
need with fair prices and terms that an auto finder cannot account for, put them
into this software and cross reference.
Need a rare part? Find a store manually that has it, put that into your table.
That store also had a few of your other parts cheaper than your other chosen
stores? Accounted for.
Unfortunately I am not a programmer, nor do I have the funds that it would take
to hire one, but I think this solution is worth something. I can offer up a token
£100 to anyone that would like a bit of a coding project to dabble with that
could be very beneficial for buyers and sellers alike.
Sorry for the ramble. Any thoughts?
|
You may noticed that there are stores which generally have everything, but for
truly outrageous prices. Like 2$ for a standard common 1x2 plate and such. Bricklink
algorithm likes to pick pieces from as few stores as possible and so ends up
picking them from super expensive ones.
The simple and working solution is to go through list of stores and if you see
one offering pieces for absurd prices, mark it as "disliked" and then use a checkbox
for not picking disliked stores. I have about 15 stores disliked in EU and get
nice and predictable results almost every time.
|
|
|
|
|
|
United Kingdom, England
France, Nouvelle-Aquitaine
USA, Washington
Poland, w. Łódzkie