Mass Update of Pricing - Questions

pmitch

I would like confirmation on the Price History data. Am I correct that the Current Prices are what are currently for sale for the given condition (new or used); the 6-month price is the average weighted price PAID for items 6 months ago; similarly, the 1-year price is the average weighted price PAID for items 1 year ago.

Questions - For the 6-month and 1-year prices, what sales are used to get the weighted average? That is, it is obviously not the amount of sales of the item on the day 6-months ago. Is it all of the sales within the last 6 months? (And similarly, within the last 1 year?)

On a related issue, I would like to be able to do a mass update on pricing, but I would like not to be limited to the choices available on the drop-down. For example, for the choice 'Price - Price History', it asks you to choose a Price History Type. The choices range from Current Average, Current Max, Current Min, 6-month Average, 6-month Max, ... 1-year Average, ... 2-year Average, ... and finally, 2-year Minimum.

I have been setting my pricing on the 1-year average, but for 'hot' items, that sometimes drastically understates the market value, so I 'leave money on the table'. Using the Current Average is unattractive since it includes the super high price stores and/or mistakes that skew the average price and make the items priced well above the viable market value.

Ideally, I wish there was a way to use a formula approach, but I understand the limitations. At this point, I should ask if I am understanding everything correctly or if there is in fact a way to achieve a customized price scheme.

Assuming that I have it right, I would suggest that having as one of the Price History options to be "Average of Current (Sales Prices) and 6-month (SOLD prices)" as well as the "Average of Current (Sales Prices) and 1-year (SOLD prices)".

I suggest that by using one of these schemes would result in TWO positive results: (1) The prices should be closer to the true, viable market price, so sales would be higher (i.e., happier customer and much happier vendor); and, (2) The effect of skewing (overstating) prices by very high-price vendors and/or mistakes could be minimized.

Sorry for the heavy math/stats discussion, but this has been a continual issue for me. Maybe I am overthinking, as my wife says. BTW, it seems as I get older, my wife is actually correct more often. I'm in trouble.

Thanks for your thoughts.

Paul
Central Bricking

Stragus

You are definitely not overthinking this.

I don't know if most sellers bother to realize this, but setting optimal prices is actually a very complicated problem of statistics and maths, having to predict/guess missing information (future part production), hidden variables (reserved stock not yet available for sale), selling patterns (variety vs quantity), hidden correlations (part A sells better with part B in stock), market manipulations ("phantom" stock from non-selling sellers), and so on.

It's not computational fluid dynamics, but it can still keep a not-too-bad-with-maths software programmer busy for months.

I don't have a good solution to recommend in the short term, unless you are interested to start coding... (I'm hoping to release some software that'll help with that tricky pricing problem, although I think the "automated pricing" algorithms wouldn't be available with the basic free version).

pmitch

Great comments, Stragus. Insightful analysis with some impressive Economic Theory jargon (e.g., correlations - part A sells better with part B in stock --> the old hot dog and hot dog bun example). I'm an old (mainframe) programmer that knows the theory but don't have the syntax or patience to apply it to this problem. BUT, I do have time to think about it. Actually I have used this as an example (my day job is teaching high school math) of complexity in real world problems and how algebra is relevant.

You were a little more direct. You mention 'market manipulations' ("phantom stock from non-selling sellers). I suspect, although I haven't done the search for examples, that some 'clever' sellers (and I don't mean 'clever' in a nice way) choose some select items and put very high prices with large quantities, which will of course never sell (phantom) to skew the 'current price' which is used by many of the (newer?) vendors. Then, this 'clever' guy puts the same items for sale at a reasonable price - knowing that he has manipulated others to overprice. Result? New vendors have suppressed sales and struggle; the 'clever' vendor has lots of easy sales.

Maybe the BO market is large enough that this manipulation have a large affect, but I don't doubt that someone is manipulating small, select segments in this way. I have seen some cases where a seller has multiple sales of a lot at different prices. This may occur innocently by someone parting out without merging, but it is still suspicious. The red flag would be if there was a high price and quantity and a separate moderate price for the same lot (id # / color / condition).

As if Lawrence doesn't have enough to do, these instances could be culled programmatically and either excluded or used to 'throw the bums out'!

But I would suggest an easier approach that would also solve the innocent mistakes. When the average prices are calculated, exclude the outliers. This can be done with simple or very elegant approaches, but I suggest that the end result would be a significant benefit to the 'average' vendor. And, as in most good things, the improvement and larger sales will likely go unnoticed (dog bites man v. man bites dog).

Stragus, I look forward to your upcoming software to address the pricing strategies. I think our market is a fascinating expression of a free market, which of course includes sellers with different levels of knowledge, the 'clever' manipulators, and extraneous complications (international postage, VAT, paperwork requirements, etc). Our selling by unit id/color/condition from 1 to 1000+ is different than the relatively simple markets like eBay and Amazon. My window has passed, but I can imagine that a deeper analysis of BO or BL (or especially their interaction!) could provide fertile territory for a PhD dissertation in Economics or Statistics or Math (Game Theory). I'm just curious - does anyone know of any serious papers or research down this line? I would be very interest in it.

Again, sorry, but you wouldn't have read the whole post if you weren't a math nerd.

Paul

Stragus

Oh neat, an old mainframe programmer. I bow with respect.

Personally, I don't think it's Lawrence's job (or BL's job) to interfer with average prices, not even by applying filters attempting to exclude outliners... They shouldn't provide tweaked/adjusted guidelines of selling prices, but rather all the raw information from which sellers could make educated decisions.

Giving just the average selling price is very limiting. What about the standard deviation, for both sales and stock? Providing the variance seems some pretty basic information to me. This is problematic because, if the marketplaces don't provide these simple statistics, it could be tempting to generate large amount of web traffic just to compute the values... Many BL price guide pages (from which you can extract a lot of valuable information) weight over a megabyte!

In contrast, a binary blob containing the raw numbers (historical and current) for all 72003 part/minifig & color combinations could take... what, maybe 5 megabytes, not even compressed? Downloading that once per week doesn't sound so bad.

I know small scale market manipulations do occur, and this may become a more serious problem if many sellers start relying on algorithms that aren't trained/hardened against such manipulations (poor algorithms could even react to and amplify each other's moves, leading to sudden unexplained crashes, just like on the stock market!).

The dynamics of this kind of free market are indeed a fascinating problem. Don't count on me to write a research paper on the topic though.

pmitch

I agree that having the standard deviation would be nice. Without some measure of variance, the average has limited value and is sensitive to manipulation. The median helps control the effect of outliers, but...at some point, this may be more statistics than most people want.