[Proposal] Long term strategy on fees: % based tx fees

[Permie]

Make it so market makers get paid, and market takers pay %. This will create liquidity like crazy.

  If possible
Liquidity is what brings in the most customers, right?

[santaclause102]

Having different fees for different types of tx and % based fees must not be a contradiction. What is your specific argument against % based fees? 

You can't justify them .. the amount is represented as an integer with 64 bits ... all those bits have to be stored on the blockchain .. independent of the ACTUAL amount ...
I can see that there is an economical reason to do so though .. there's just no technical one ..

I see your argument. It is part of this post: https://bitsharestalk.org/index.php/topic,17721.msg226010.html#msg226010
But there is another way to look at this (second approach, see the post).
We might agree if the conclusion of the second approach is that the pain for the customer is too high to justify high total fees that result from % based fees. But who knows that it is so? And who knows that the equation I suggested is not super positive (high return and low customer pain).

[xeroc]

Having different fees for different types of tx and % based fees must not be a contradiction. What is your specific argument against % based fees? 

You can't justify them .. the amount is represented as an integer with 64 bits ... all those bits have to be stored on the blockchain .. independent of the ACTUAL amount ...
I can see that there is an economical reason to do so though .. there's just no technical one ..

[santaclause102]

By what other means than tx fees can the DAC make money?

- market orders and trades
- bonds
- different fees for public and blinded transactions

in the end .. everything you do in bitshares is a transaction .. but you can tie different fees to each of them .. and most are independent of the AMOUNT, some are dependent of the SIZE .. none are in % ... all can be defined by 'delegates' in bts2

Having different fees for different types of tx and % based fees must not be a contradiction. What is your specific argument against % based fees? 

[xeroc]

By what other means than tx fees can the DAC make money?

- market orders and trades
- bonds
- different fees for public and blinded transactions

in the end .. everything you do in bitshares is a transaction .. but you can tie different fees to each of them .. and most are independent of the AMOUNT, some are dependent of the SIZE .. none are in % ... all can be defined by 'delegates' in bts2

[Frodo]

Okay my above comparison isn't working 100%. I'll try to look at this differently:

On a logical level a transaction fee should consist of two parts: raw tx processing cost + platform development cost. As xeroc said development cost is independent of nearly everything. It is a constant. So we have to distribute these costs arbitrarily with the goal that the user base perceives it as fair. To define fairness is always difficult but here are my thoughts. Development cost part of a tx fee should be dependent of:
   - the importance of our service to a customer
   - the importance of the customer relative to all customers (i.e. big company vs individual)

By setting a fixed tx fee customers pay linearly for development with the number of transactions they make. That is imo a very bad cost distribution regarding the two points made above. (Why should a customer who makes frequent small transactions pay much more for development of the platform than a customer who depends on less frequent but big transactions? Our service is equally important to both and thus both would be willing to pay the same amount of fees.) A more accurate picture would be total transaction volume, as it accounts in some way both points I listed. Hence I would argue that tx fees should be in some way dependent of volume.

[santaclause102]

A fixed % and a min fee would produce very high (too high to compete) total fees for high volume tx.

I still dont see the point why it should technically cost more to transfer $1M than it costs to transfer $1 .. both are just unsigned integers and have the same technical cost in terms of processing and storage .. You can't justify this except for "making a bigger profit" for the DAC .. and I think there are better ways to may more profit than that .. IMHO

By what other means than tx fees can the DAC make money?

[xeroc]

I disagree. What you are paying for is not only the raw costs of computing power to make the transaction but much rather the development cost behind the system.

And that cost is still independent of the amount a user wants to transfer, isn't it?

I believe it is the same with BitShares. Perceived value of transferring $1M is higher than for $1. So why shouldn't you pay different amounts for development of the platform regardless of raw tx costs?

sure ... include development time and all that stuff .. but using the system should cost all participants the same .. independent of HOW they use it ..
I can use my computer to play video games or to establish a million-$ company like google .. the computer still cost me the same .. independent of how I use it

[Frodo]

A fixed % and a min fee would produce very high (too high to compete) total fees for high volume tx.

I still dont see the point why it should technically cost more to transfer $1M than it costs to transfer $1 .. both are just unsigned integers and have the same technical cost in terms of processing and storage .. You can't justify this except for "making a bigger profit" for the DAC .. and I think there are better ways to may more profit than that .. IMHO

I disagree. What you are paying for is not only the raw costs of computing power to make the transaction but much rather the development cost behind the system.

I like to compare this to microprocessor market. What you pay for is mainly R&D and not chip manufacturing. And sometimes you will be sold the same piece of hardware with some locked features for less money than the unlocked version. That might seem completely decoupled from reality as well, but it works. And it actually reflects perceived value.

I believe it is the same with BitShares. Perceived value of transferring $1M is higher than for $1. So why shouldn't you pay different amounts for development of the platform regardless of raw tx costs?

[xeroc]

A fixed % and a min fee would produce very high (too high to compete) total fees for high volume tx.

I still dont see the point why it should technically cost more to transfer $1M than it costs to transfer $1 .. both are just unsigned integers and have the same technical cost in terms of processing and storage .. You can't justify this except for "making a bigger profit" for the DAC .. and I think there are better ways to may more profit than that .. IMHO

[santaclause102]

Resource-wise, a $2 transaction costs the same as a $1 transaction. So the current fee model is the most accurate.

What does "accurate" mean? There are two ways to approach this:
1. Tie tx fees to the network's actual costs.
2. Tie tx fees to perceived value of making the respective tx.
You have to ask why you would choose which approach.

Also, I don't want to pay 100x the fees for sending $100 as opposed to $1.

This makes sense from a customer's perspective. It doesn't make much sense form a business perspective. Any business, also Bitshares has to be as profitable as possible and you do that most efficiently by making the following ratio is as positive as possible: money made per customer interaction / degree to which the fee/price bothers the customer. The latter also has to include how likely it is that the respective customer will use some competing business, so at best you are as expensive as possible but as cheap as necessary to not drive people away from your business respectively to give them enough of an incentive to change from a competitor to you. The outcome of this equation might be that fees don't actually rise much if you transfer a lot of value because competitors out there (other crypto currencies) offer lower fees and high volume customers are valuable for your (exchange) business otherwise. But looking at it this way makes a lot of sense in general if you want to survive as a (decentralized) company. Money made from optimizing this equation then can be invested into your infrastructure (bitshares' workers for example).

Yeah, you're right.. it's about competition and perception. And it may work quite well for market orders.

But the sliding % scale sounds confusing.. wouldn't it be better to have a fixed % and a min fee? Or a tiered cost structure.

I still think for basic transfers, competition will soon drive the price very close to the true cost.
Also, in these cases it will not be possible:
 - UIA's of unknown value (so it would only work for liquid, tradeable assets)
 - Confidential tx's which hide the amount

Good point with the exceptions!

A fixed % and a min fee would produce very high (too high to compete) total fees for high volume tx.

[roadscape]

Resource-wise, a $2 transaction costs the same as a $1 transaction. So the current fee model is the most accurate.

What does "accurate" mean? There are two ways to approach this:
1. Tie tx fees to the network's actual costs.
2. Tie tx fees to perceived value of making the respective tx.
You have to ask why you would choose which approach.

Also, I don't want to pay 100x the fees for sending $100 as opposed to $1.

This makes sense from a customer's perspective. It doesn't make much sense form a business perspective. Any business, also Bitshares has to be as profitable as possible and you do that most efficiently by making the following ratio is as positive as possible: money made per customer interaction / degree to which the fee/price bothers the customer. The latter also has to include how likely it is that the respective customer will use some competing business, so at best you are as expensive as possible but as cheap as necessary to not drive people away from your business respectively to give them enough of an incentive to change from a competitor to you. The outcome of this equation might be that fees don't actually rise much if you transfer a lot of value because competitors out there (other crypto currencies) offer lower fees and high volume customers are valuable for your (exchange) business otherwise. But looking at it this way makes a lot of sense in general if you want to survive as a (decentralized) company. Money made from optimizing this equation then can be invested into your infrastructure (bitshares' workers for example).

Yeah, you're right.. it's about competition and perception. And it may work quite well for market orders.

But the sliding % scale sounds confusing.. wouldn't it be better to have a fixed % and a min fee? Or a tiered cost structure.

I still think for basic transfers, competition will soon drive the price very close to the true cost.
Also, in these cases it will not be possible:
 - UIA's of unknown value (so it would only work for liquid, tradeable assets)
 - Confidential tx's which hide the amount

[bitmeat]

Make it so market makers get paid, and market takers pay %. This will create liquidity like crazy.

[santaclause102]

Resource-wise, a $2 transaction costs the same as a $1 transaction. So the current fee model is the most accurate.

What does "accurate" mean? There are two ways to approach this:
1. Tie tx fees to the network's actual costs.
2. Tie tx fees to perceived value of making the respective tx.
You have to ask why you would choose which approach.

Also, I don't want to pay 100x the fees for sending $100 as opposed to $1.

This makes sense from a customer's perspective. It doesn't make much sense form a business perspective. Any business, also Bitshares has to be as profitable as possible and you do that most efficiently by making the following ratio is as positive as possible: money made per customer interaction / degree to which the fee/price bothers the customer. The latter also has to include how likely it is that the respective customer will use some competing business, so at best you are as expensive as possible but as cheap as necessary to not drive people away from your business respectively to give them enough of an incentive to change from a competitor to you. The outcome of this equation might be that fees don't actually rise much if you transfer a lot of value because competitors out there (other crypto currencies) offer lower fees and high volume customers are valuable for your (exchange) business otherwise. But looking at it this way makes a lot of sense in general if you want to survive as a (decentralized) company. Money made from optimizing this equation then can be invested into your infrastructure (bitshares' workers for example).

[giant middle finger]

Answer to the microtransaction dilemma :

create a "super poweruser" account

[monsterer]

Percentage based fees are great because they enable micro-transactions and automatically levy a higher tax on those who are able to afford to pay more.

[fav]

I don't see a need. power users will upgrade their account, so fees are no issue.

[xeroc]

Resource-wise, a $2 transaction costs the same as a $1 transaction. So the current fee model is the most accurate.

Also, I don't want to pay 100x the fees for sending $100 as opposed to $1.

I could see marketing benefits, but the % model is decoupled from reality.

^^ This!

[roadscape]

Resource-wise, a $2 transaction costs the same as a $1 transaction. So the current fee model is the most accurate.

Also, I don't want to pay 100x the fees for sending $100 as opposed to $1.

I could see marketing benefits, but the % model is decoupled from reality.

[xiahui135]

% based fees allow spamming with tiny transactions.  So there would have to be a minimum fee.

the blockchain is not designed for some applications, such as pegging.

Because it cost to changes orders, to react to the price change.

[Ander]

% based fees allow spamming with tiny transactions.  So there would have to be a minimum fee.

[santaclause102]

I think it is safe to say the fee should never be more than 10% of the value transferred, but must be at least a min amount and at most a max amount. 

$10 for max fee.  $0.01 for min fee and 0.5% otherwise.

That would make it $0.50 to transfer $100. 


Sent from my iPhone using Tapatalk

How do you come to the conclusion that it needs an upper limit?
I would probably also suggest an upper limit but I would just go by the rule: As expensive as possible while still being cheaper then centralized exchanges / competitors today. BM said that something like that too on mumble a while ago. I wonder why % based fees where not part of the solution.

[Frodo]

The question that I have is how we can determine the value of a transaction? For example a rarely traded UIA can't be easily measured in USD or BTS.

[sittingduck]

I think it is safe to say the fee should never be more than 10% of the value transferred, but must be at least a min amount and at most a max amount. 

$10 for max fee.  $0.01 for min fee and 0.5% otherwise.

That would make it $0.50 to transfer $100. 


Sent from my iPhone using Tapatalk

[santaclause102]

I dont agree with the percentages, I think the % fee should decrease more drastically with higher volume as very high volume traders have extremely low fees.

But overall I really like this approach 

The actual numbers where just an arbitrary example...

[MrJeans]

I dont agree with the percentages, I think the % fee should decrease more drastically with higher volume as very high volume traders have extremely low fees.

But overall I really like this approach 

[santaclause102]

If the assumption is that (1) we want to keep fees as low as possible so to not deter anybody from using bitshares and (2) we want to maximize revenue (earn as much as possible from tx fees), I would argue that adopting a % based fee is the superior strategy.

Let's say we have 2 transactions tx where we overall want to maximize (1) and (2): Transaction (a) is transferring 1$ and transaction (b) is transferring 1 million $.

-> With a flat fee structure of let's say (simplified) 40 cents per tx we we would deter the users from transaction (a) and we would only earn 40 cents in total from the two transactions.

-> With a % based fee structure, let's say (decreasing % with increasing volume) it would be 10% (should be close to the networks actual costs the lower the volume gets) at 1 $ and 0.01 % at 1 million USD. The network would then earn 10 cents + 100 $.

Summary: A % based fee structure would help to bring the tx fees closer to the perceived value BitShares provides.