There are a few basic features that Grant Achatz and Nick Kokonas should build into their online ticket sales. First, you want a good system to generate the initial allocation of tickets for a given date, second you want an efficient system for re-allocating tickets as the date approaches. Finally, you want to balance revenue maximization against the good vibe that comes from getting a ticket at a non-exorbitatnt price.
- Just like with the usual reservation system, you would open up ticket sales for, say August 1, 3 months in advance on May 1. It is important that the mechanism be transparent, but at the same time understated so that the business of selling tickets doesn’t draw attention away from the main attractions: the restuarant and the bar. The simple solution is to use a sealed bid N+1st price auction. Anyone wishing to buy a ticket for August 1 submits a bid. Only the restaurant sees the bid. The top 100 bidders get tickets and they pay a price equal to the 101st highest bid. Each bidder is informed whether he won or not and the final price. With this mechanism it is a dominant strategy to bid your true maximal willingness to pay so the auction is transparent, and all of the action takes place behind the scenes so the auction won’t be a spectacle distracting from the overall reputation of the restaurant.
- Next probably wants to allow patrons to buy at lower prices than what an auction would yield. That makes people feel better about the restaurant than if it was always trying to extract every last drop of consumer’s surplus. Its easy to work that into the mechanism. Decide that 50 out of 100 seats will be sold to people at a fixed price and the remainder will be sold by auction. The 50 lucky people will be chosen randomly from all of those whose bid was at least the fixed price. The division between fixed-price and auction quantities could easily be adjusted over time, for different days of the week, etc.
- The most interesting design issue is to manage re-allocation of tickets. This is potentially a big deal for a restaurant like Next because many people will be coming from out of town to eat there. Last-minute changes of plans could mean that rapid re-allocation of tickets will have a big impact on efficiency. More generally, a resale market raises the value of a ticket because it turns the ticket into an option. This increases the amount people are willing to bid for it. So Next should design an online resale market that maximizes the efficiency of the allocation mechanism because those efficiency gains not only benefit the patrons but they also pay off in terms of initial ticket sales.
- But again you want to minimize the spectacle. You don’t want Craigslist. Here is a simple transparent system that is again discreet. After the original allocation of tickets by auction, anyone who wishes to purchase a ticket for August 1 submits their bid to the system. In addition, anyone currently holding a ticket for August 1 has the option of submitting a resale price to the system. These bids are all kept secret internally in the system. At any moment in which the second highest bid exceeds the second lowest resale price offered, a transaction occurs. In that transaction the highest bidder buys the ticket and pays the second-highest bid. The seller who offered the lowest price sells his ticket and receives the second lowest price.
- That pricing rule has two effects. First, it makes it a dominant strategy for buyers to submit bids equal to their true willingness to pay and for sellers to set their true reserve prices. Second, it ensures that Next earns a positive profit from every sale equal to the difference between the second-highest bid and the second-lowest resale price. In fact it can be shown that this is the system that maximizes the efficiency of the market subject to the constraint the market is transparent (i.e. dominant strategies) and that Next does not lose money from the resale market.
- The system can easily be fine-tuned to give Next an even larger cut of the transactions gains, but a basic lesson of this kind of market design is that Next should avoid any intervention of that sort. Any profits earned through brokering resale only reduces the efficiency of the resale market. If Next is taking a cut then a trade will only occur if the gains outweigh Next’s cut. Fewer trades means a less efficient resale market and that means that a ticket is a less flexible asset. The final result is that whatever profits are being squeezed out of the resale market are offset by reduced revenues from the original ticket auction.
- The one exception to the latter point is the people who managed to buy at the fixed price. If the goal was to give those people the gift of being able to eat at Next for an affordable price and not to give them the gift of being able to resell to high rollers, then you would offer them only the option to sell back their ticket at the original price (with Next either selling it again at the fixed price or at the auction price, pocketing the spread.) This removes the incentive for “scalpers” to flood the ticket queue, something that is likely to be a big problem for the system currently being used.
- A huge benefit of a system like this is that it makes maximal use of information about patrons’ willingness to pay and with minimal effort. Compare this to a system where Next tries to gauge buyer demand over time and set the market clearing price. First of all, setting prices is guesswork. An auction figures out the price for you. Second, when you set prices you learn very little about demand. You learn only that so many people were willing to pay more than the price. You never find out how much more than that price people would have been willing to pay. A sealed bid auction immediately gives you data on everybody’s willingness to pay. And at every moment in time. That’s very valuable information.
15 comments
Comments feed for this article
April 11, 2011 at 5:38 pm
Rick Wash
Interesting design. My main worry would be the two-tier system. With that design, some people who are more flexible on date will bid the minimum to get the fixed-price tickets, rather than compete for the upper-half guaranteed allocation. Will it produce obvious class differences in the patrons? “Oh, them over there. They won the ticket lottery and can only afford to pay the low fixed price.”
April 11, 2011 at 8:56 pm
k
you could keep the percentage of people who get “fixed price seats” variable
but this only increases the complexity of what already seems quite a complex pricing scheme – a scheme trying to raise more revenue appears to carry significant costs
so maybe it is okay to let some money lie on the table
April 12, 2011 at 2:30 am
conchis
1. I think you implicitly assume this, but presumably the restaurant would want to set a floor price (at whatever the fixed price is). Otherwise they could end up losing money if the N+1th bid is below the floor.
2. What is N? I’m assuming it needs to be min(# of bidders-1,# seats)? Transparency seems to rely on there always being at least one bidder that misses out, but that will sometimes require you to leave seats unnecessarily empty if # bidders <= # seats.
April 12, 2011 at 11:23 am
jeff
indeed, you may want to set a reserve price and then you have an answer to your point 2. the price is the larger of the N+1st price and the reserve price. And so this is just saying that if there is not enough demand you don’t fill the restaurant.
not likely to be an issue for Next any time soon 🙂
April 13, 2011 at 12:36 pm
conchis
Yeah, I guess I just find it interesting that here “not enough demand to fill the restaurant” actually means “not enough demand to very-slightly overfill the restaurant”, and that if you have only precisely enough demand to fill the restaurant, you have to have to leave potential gains from trade ‘on the table’ to maintain the premium to the floor price that the bidding system is supposed to generate.
In case it’s not clear what my thinking is here, details below. (In case it already is, apologies.)
Assume for a given auction bids=capacity, and that this is common knowledge.
Setting N=bids=seats would mean that the optimal bid is to bid the floor price, and profit = capacity x (floor – cost/seat).
Setting N=bids-1=capacity-1 maintains transparency, which will generate a premium to the floor price, at the cost of leaving the potential profit from the final seat ‘on the table’. Profit = (capacity – 1) x (floor + premium – cost/seat). It’s likely that this is greater than profit in the first case (provided [capacity-1] x premium > floor – cost/seat). But, as I said, I still find it interesting that profit maximising in this case requires you to not make a trade that would be beneficial for both parties. (I guess in the same way that this is generally true for a profit maximising monopolist who is unable to price discriminate.)
Obviously, things are a little more complicated if individuals don’t know how many bids there will be, but I think the basic trade-off is still there.
April 12, 2011 at 7:57 am
nick kokonas
I don’t disagree with the theory on any of this design… but now explain that to customers in 50 words or less. 🙂
April 15, 2011 at 2:55 pm
Danny
@Nick,
Aren’t you and @Gachatz all about blazing new trails? The ticketing system was a new trail, and now this is an even newer trail!
April 12, 2011 at 8:49 am
Sandeep Baliga
Nick: If people can use eBay, they can understand point 1 above and use a simple auction. The eBay proxy bidding feature is basically like point 1.
That alone is better in terms of revenue that what you are doing now. The rest is icing on the cake to distribute seats more cheaply and make the resale market more efficient. But either the rest does not have to be implemented or can be done without diners knowing the nitty-gritty behind the scene details.
As we said, let us know if you want help. This would be fun for us and profitable for you. Jeff and Sandeep
April 12, 2011 at 5:46 pm
Next Restaurant: pricing and ticketing innovation redux « Knowledge Problem
[…] My Kellogg colleague Sandeep Baliga wondered the same thing yesterday at Cheap Talk, and then Jeff Ely picked it up and ran with it later in the day at Cheap Talk (note also that Nick Kokonas commented on both posts, love it). Here’s Jeff’s […]
April 14, 2011 at 6:15 pm
Ben Brockert
But people don’t understand eBay. Have you ever talked to buyers? One of the most common refrains is “if I’d known it was only going to go for $mybid+bid_increment, I would have bid more.” The idea of actually bidding the maximum they’re willing to spend is alien to them. This is why eBay sniping services and programs exist and work well.
In essence it’s like suggesting to a haggler that you state your actual maximum price, asking for their minimum price, and setting at the average. In theory it would be an efficient way to do it, but in reality no one trusts the other party that much. Even when there’s an impartial third party in the transaction, like eBay.
April 15, 2011 at 8:22 am
Michael Griebe
I may be mistaken, but I think the efficiency of the auction is also based on limiting individuals to a single bid. That is not as easy to do in practice as one would assume in theory.
Also, and here is the real kicker, do we really think people know how much they are willing to pay for something? I may be willing to pay £195 for a camera, provided I am unable to pay less for it. If I find that I may have payed less, then I am suddenly made worse off and feel like a dupe.
Economic Agents may not be, but people are social creatures.
April 19, 2011 at 11:26 pm
Maximizing Consumer Surplus Part III: Rationing The Top « Cheap Talk
[…] are not eliminated altogether because of the rationing at the top. This can be dealt with by controlling the resale market. Indeed here is one clear message that comes out of all of this. Whatever motivation the […]
April 20, 2011 at 6:03 am
iPad2: Product Rationing and Allocation Schemes « The Operations Room
[…] Why not sell the tickets on-line the night before like Next restaurant, or even follow the auctioning scheme proposed by my colleague and wine-lover Jeff Ely. (Admittedly, higher and dynamic pricing may not be aligned with Apple’s current tag of […]
May 16, 2011 at 9:08 am
Next Restaurant Pricing Scheme Continues To Fascinate « Cheap Talk
[…] partners in Next and the person responsible for the innovative pricing scheme, is reluctant to use an auction to capture the surplus Next is generating for scalpers. He is worried about price-gauging. We have suggested one […]
February 17, 2012 at 11:54 am
Next Restaurant Adopts Auction Scheme « Cheap Talk
[…] We blogged many times about the Next restaurant’s innovative ticket scheme. Potential restaurant goers had to sign up to try to acquire seats at a fixed price for a set meal. Good for the restaurant in terms of knowing what inventory to hold, predictable revenue etc. The scheme turned out to be extremely successful with resale prices of the tickets running into thousands. Why not just auction off the seats – that is what very economist would say? It turns out that Nick Kekonas and Grant Achatz do not want to make too much money and want everyone to be able to afford to come. In response to my blog post Nick commented: Since we have universally high demand right now, the question is why don’t we flatten the pricing towards the top of what the market will pay? There are a few reasons for this, mostly having to do with customer service and the hospitality industry. Simply, we never want to invert the value proposition so that customers are paying a premium that is disproportionate to the amount of food / quality of service they receive. Right now we have it as a great bargain for those who can buy tickets. Ideally, we keep it a great value and stay full. […]