The Challenge of Seedbox Auctions – Discussion of perspectives

The Seedbox auctions were both a tremendous success – and a tremendous challenge. Lately we have been thinking hard about the successes and challenges, and solving the challenges to potentially bring it back.

The community received tremendous amounts of value. Auction seedboxes had immense and truly extraordinary value. We would like to continue pushing the boundaries.

Feedback and challenges

Let’s look at some of the common feedback, the challenges of the auctions. Could we potentially do something about them? If we forgot something, we will try to address it on future blog post – or in the comments.

Unresponsive support / No Support

This was probably by far the biggest feedback. Auctions originally were offered on the premise of low priority / no support. Many ordered approving the special terms, but not really. They either did not read them, or accept them but ordered regardless.

Some people simply claimed no support at all even for critical server issues, we noticed many cases where people jumped to forums etc. claiming no support but the poster had never even made a single ticket to begin with. Privacy is important, we do not respond anything relating to any single person publicly ever.

Huge flood of tickets

Number of tickets made certain that sometimes we just could not respond swiftly. This caused delays some times, and as is typical there are always people with unrealistic turnaround times. We might be one of the fastest in industry generally, but we too are just humans and sometimes can be swamped, or worse, sleep and rest!

Even more flood of tickets was because a lot of people were new to the services, so they had all sorts of basic usage questions. Often easily replied with knowledge base or wiki link. This caused even more delays to important tickets.

More and more tickets, auction users needed roughly 10x the amount of support than regular users. 20/20 hindsight: Of course, new user and some did not even check what they are getting just getting a service as swiftly as possible. However, there was also much much more tickets than typically for new users.

Our intention was extra savings to make auctions more financially viable. Lowering support with special terms etc., to the bare minimum/best effort only – but this simply is not how people act. You cannot use terms in a service like this to limit something like this out. Lesson learned.

No Service/Terminated Service/Cancelled Service

Some people would exaggerate, and sometimes post outright untruths (see point about we are not in vacuum below). Occasionally there was kernel of truth however. The terms for auctions were that we could cancel your auction service at any time for any reason. This would allow us to provide the service on average for less cost, and from our perspective with less hassle.

Infact businesses in general can choose their customers, any private company offer is not a right, it is a privilege. Alas, once again this is not what people think, act or want – despite being the letter and spirit of the law.

Shall we take a peak “behind the curtains”? This is how exactly, and why some people got their auctions cancelled.

Abuse, fair and simple. There was a lot of people who abused the system, for example common abuse was to “buy” multiple auctions, and in the end only pay the lowest cost one – blocking others from getting in at that price. They were swiftly terminated and refunded.

More abuse, there is a way to order “off the book” from the billing system – we removed one auction, but forgot to remove updating availability and price – so the price sunk to the minimum possible as it was not visibly orderable. These people found order links to them by essentially guesswork/bruteforce. These users were terminated and refunded.

Load balancing, (by far the biggest group!) sometimes when you add a lot of users at once our algorithm has to do guesswork, and multiple heavy users may wind up on the same server, causing regular seedbox service users grief. From our perspective, the situation is very simple: User A through F all pay a premium service and has been with us for a long time, User G is a auction slot paying fraction and signed up short while ago and is using the server to the utmost maximum. Should we go through huge pains of manually migrating user G to his own private server, or the users A through F – knowing that either action we take will take 1 year to recoup the labor cost involved? No, for the greater good we load balance cancelled those auction slots. Typically at end of billing period, not giving the user chance to renew and letting them know well upfront this was going to happen. In extreme cases we did immediately, or nearly immediately, terminate a few auction slots. If their term got cut short a refund was provided.
It may sound really harsh, but ultimately it comes down to finances. The Users A through F could be paying say 100€ a month total combined, while User G (auction) only 3€ a month. Sadly user G being an auction slot would be load balanced out, instead of sacrificing long term users A thru F

Slow Service/Bad service or variations

This was to be expected. There are always those users who had different expectations, or the service did not fit their purpose. More users the faster you take in – the more of these there will be.

This is true for any and all services out there. One example could be that you live in malaysia, your round-trip latency to our server is 450ms and you want to use the server for 4k hdr livestream from the server – that is a very tough situation even for high end server, let alone a lowend specced low cost service! (Btw, it is totally doable from even our lowliest servers if your RTT latency is minimal and the server is having nominal loads, not high! Some buffering may be required)

General rule of thumb is that the higher the monthly cost of a service, that much more you can expect. It is quite true, for someone paying 25€ a month we are much more happy to keep server in low utilization for their needs, but when user pays 4€ a month we have to put a lot of users on the server to even break even on cost.

Fortunately, we generally use quite small sized servers at only 4 drives, so even for auction seedboxes it’s not that many users, but it definitely always is more users than higher cost services. Good example of this principle is the Dragon series – Those servers have always plenty of resource reserves and fairly low average usage – partially this is because we can keep the server sizes smaller as well and get same revenue from 4 users on a Dragon server than 20 on a Value250 server.

Churn is a challenge

For auctions churn was more of a challenge than typical service – but different kind of churn you’d first expect to.

During auctions we did peak at higher than usual churn, nearly double. That was expected however, a lot of people just wanted to try it out and test it out.

The challenging churn was normal service, or even premium service users changing to a auction slot, and then expecting, or even demanding, exactly the same service level (ties into support above!) than on their premium service. While sometimes cutting as much as 40€ from their monthly cost. It is not a problem when just a couple people do that, but when that happens en masse – that is an serious problem.

Imagine a situation of, let’s say 10 000 users. They have a service A, but opt to go for service B which cuts into the basic service specs, and saving each 5€ a month. This is 50 000€ a month of change in revenue – THAT IS A LOT of money! Then these same people start demanding they still get everything in the service A – not allowing the service provider to leverage any cost savings, or some even actively trying to harm the service provider if the same service is not provided.

The expected churn of new signups was a non-issue, we knew it would happen and a lot of people just giving it a try, nothing else. Not a big deal at all.

We are not working in vacuum

This is not a vacuum, and there are other players in the field. We got into their radars in a really bad way, and ever since starting the auctions it’s been constant attacks in one form or another, from annoying (spam) to very detrimental to our users (DDOS) and actually attacking our users more than us as a business, causing grievances to our users. Every time we released something new, a new attack would begin. Every time we publish something, a new attack would begin.

These attacks continue to this very day

There was also the individuals looking to game or abuse the system.

Lower the cost of service, harder it is.

Lower the cost of the service that much harder it is to provide service for the users. This is an very old lesson, but once again got reinforced. If user has very low budget, it is much harder to cater to them. Usually very low budget user demands more, those euros are much much more valuable to him than to many others. This is generalization and not everyone is like that.

Lessons learned

There was a lot of lessons learned from this, very important lessons

The good

First of all people were reeaaaaally excited about this, there was even commentary that it was “marketing genius”. People made scripts to monitor the auctions and discussed them.

Many would not signup for anything above 3€, 4€ or 5€ – so this was their only way to get a service.

Many of those who took a auction seedbox, would go on and upgrade also. Relates to the churn challenge, it was a two way street!

Mindshare and tons of it for our services. Even premium services like Dragon sold better at the time.

Churn was overal much less than expected. This speaks volumes about our overall service quality, for very large part people are really happy with their services. Auction or not.

Learnings from challenges

Special terms: Special Learnings

However the special terms of the auctions was mis-guided and outright wrong for many parts. The limited support, and service cancellations were not good at all. Those should not have existed. Some special terms were acceptable (SLA level, no bonus disk quota, abuse, marketing) however.

Whenever we leveraged those special terms, there was outrage. Sometimes however fake outrage (ties to point not in vacuum).

Outscaling ourselves: Too many users at once

We took in way too many users at once. We shouldn’t have sold that many in that short duration. Pacing it more would have tremendously helped with server load and helpdesk load.

We took in too many users at once caused server load balancing issues, these should have been done using migrations rather than service cancellations (see point special terms!)

Too low price point meant that there was obvious lack of resources for support, 3-4€ a month just is not enough to provide customer support, top notch servers and network, cover all the overheads and leave any profit on the table at the same time. 3€ is very swiftly eaten up by support alone (TL;DR; Finland has super high taxes, human time is super expensive). Read the challenge point about lower the cost, harder it is.

Regular service similarity

Similar characteristics of auction services than regular caused churn on regular and premium services – Auctions should be clearly distinctive to avoid this negative outcome. For example, service lines which are now considered deprecated – to keep the servers full for one extra year would be a true Win-Win situation: Users get low cost service, we get just a little bit more lifetime out of an old server.

Conclusions and discussion. New auctions?

We hope to bring auctions back quite soon. These are the changes we are thinking about:

Regular service, regular terms

New auctions would be completely regular service with (mostly) regular terms. You could opt for any billing term (sans long term discounts). You would have normal support at the same priority as every other user. Only special terms would be related to abuse, marketing, SLA and bonus disk quota (No bonus disk quota).

Limited types of service

Auctions would only be something where we have true win-win situation, not from the regular service server pools. For example, Value250 and M1000 has been superceded, so the new auctions would be from those series.

We would not auction from the newest series of services, for example M10G, new MDS servers.

Number and price limitations

We should not allow much below 4€ a month price point, and we should concentrate on a little bit higher price point on average. This would allow us to put more towards human resources, AKA support.

Number of auction slots should be more restricted to not cause a spike in helpdesk load, and to provide every one (auction or not) prompt, swift and polite customer care.

Last words

Just the few things above will solve a lot of the challenges with the auctions. What do you think? Did we miss or forget something? Comment below or contact helpdesk. Every comment will be read by our staff.