Optimal Auctions Blog

Designing a Better Nortel Patent Auction

Posted by Mike on July 24th, 2012
Designing a Better Nortel Patent Auction
Note (8/14/2012) - This blog posting was written before the Canadian courts released the previously sealed details of the auction last week. This article was written based on leaks and educated guesses about the auction that were available previously.

It's been about one year since the Nortel Bankruptcy Patent Auction finished, and during that time the value of Intellectual Property has skyrocketed, especially technology-related IP. This ground-breaking auction set the precedence for how much these patents are worth, and which companies would be controlling the future of these patents. But, when we take a look at the auction itself and how it was structured, we can see many critical mistakes - mistakes that could be mispricing all patents currently being traded around the world.

There were many mistakes by the auction designers of the Nortel Patent Auction and these mistakes likely cost Nortel and its shareholders hundreds of millions of dollars.

The goal of every auction is easy - to get the highest sales price for the seller. A perfectly designed auction is able to do that for a seller. The problems start when the auction designers begin to stray from the rules and best-practices of a well-designed auction. Every mistake by the auction designer destroys value for the seller by creating a less efficient auction process.

Background

Nortel Networks, which at one time was the biggest company by market cap in Canada, had filed for bankruptcy and was going through a liquidation. After selling all their liquid assets, one of the last assets to sell off in the bankruptcy proceedings was also their most valuable - their technology related intellectual property, 6000 patents in all. Initial estimates pegged this at about $1B, but it wasn't entirely clear how much the portfolio would be worth since a patent portfolio this large and valuable doesn't come up for sale often. Past precedence on pricing didn't exist.

The auction was set to kick off with 5 bidders - Apple, Intel, Google, a group consisting of Ericsson, RIM, Microsoft, Sony and EMC, and a group called RPX (a defensive patent company with dozens of stakeholders).

The Auction

  • Google set a starting price of $900M. If no one bid during the auction, Google would have won at that price.
  • On the first day, Intel made a first bid of $1.5B. Immediately RPX dropped out of the auction.
  • The 2nd day, the Ericsson group dropped out and joined up with Apple to form a larger group.
  • The 3rd day, Intel dropped out, and joined up with Google to form a new group.
  • On the 4th day, the two remaining groups of bidders bid back and forth in $100M increments until the Google group dropped out at $4B.
  • The Apple-led group won the auction, and all 6000 patents, at $4.5B.
  • Within the Apple-led group, the group agreed to break down the large portfolio into smaller groups.

    Apple would pay $2B and claim ownership of all 4G related patents and any patent they felt they could use offensively against Android.

    RIM and Ericsson would pay $550M each to license the entire portfolio.

    Microsoft and Sony would pay $500M each to license the entire portfolio.

    Finally, EMC would pay $400M to claim ownership of a small group of patents that were storage related in nature.

The Problems in This Auction

Public Disclosure of Bidders - This is one of the most common mistakes auctions make, and it is potentially one of the biggest and costliest mistakes as well. By announcing the participants in an auction before the auction, participating firms can get a good estimate on the maximum amount each bidder will bid during an auction (e.g. a firm like Apple with $100B in the bank will be able to bid more than RIM, who is losing money every day). It also allows firms to "self-remove" themselves from the bidding process - a company would seemingly say "there's no way I can beat Apple, I won't even bother to enter the auction".

The effects of this should be obvious - the auction is transforming the demand curve from its free-market equivalent before the auction even starts! By encouraging small bidders to drop out, and by giving bidders estimates on other bidders' maximum prices, the biggest advantage of an auction (defining a market price in a market lacking a clear demand curve) is removed.

Allowing Google to Place a Starting Bid - Besides the mistake of publicly disclosing Google as a participant in the auction (see above), this allows Google to send a signal to other bidders about their bidding strategy and the maximum amount they would be willing to bid in the auction.

Apple vs. Google

In fact, placing the initial bid was a HUGE DISADVANTAGE to Google in this auction. Google's bid was only $900M on their starting bid, and in retrospect with a final price of $4.5B, it showed that Google was in a weak position in this auction. They were trying to keep the price low with their starting bid, and other bidders could take advantage of this weakness during the bidding process.

Conversely, look at what would have happened if Google was in a position of strength in this auction. Suppose Google placed an opening bid of $3.8B. This would have sent a signal to all other bidders in the auction that Google would go to any lengths to win the patents. Most likely in this scenario there would have been no other bidders, and Google would have won the portfolio at $3.8B ($700M less than the final price).

Allowing Bidders to Enter their Own Prices - One of the most-discussed results of this auction was Google's "humorous" bids during the auction. They famously entered bids like "$3,141,593" during the auction (the digits of pi). This mistake allows bidders to send "signals" to other bidders during an auction. Even if all the bidders were anonymous and kept secret, a bid of "$3,141,596" would give a huge hint that Google is partaking in the auction.

Not Dividing the Portfolio Up - the auction entailed 6000 patents in one portfolio. Essentially one single product was auctioned. However, judging by what we saw from the results, some groups of patents were more valuable to some firms than others, and some weren't worth anything to some firms. The fact that EMC was able to purchase storage-related patents with no interest from the other participants indicates that those patents didn't belong in the portfolio with the other patents. The auction process would have been much cleaner if those storage patents were sold as a separate product - then not only could EMC bid on it, but firms like SanDisk, IBM, etc. could bid as well, without the constraint of having to also bid on patents they weren't interested in.

Allowing Bidders to Form Groups During the Auction - The final mistake made in the auction design was allowing bidders to drop out of the auction individually, but continue in the auction as a group, or by joining with another individual bidder to form a group. This mistake is the culmination of all the other mistakes made in the auction design - it's the result of public disclosure of bidders and not dividing the portfolio of patents up into smaller portfolios.

The fall-out from this mistake may not be obvious - you might even think it leads to higher prices, since more companies means more money involved. However, ultimately it likely led to less overall revenue for Nortel. Consider this case - if all participants were anonymous, would Google be willing to license the patent portfolio for $500M like other winners? Absolutely! However, due to the business hostilities between Google and Apple, they would never partner publicly in that way. So, Nortel loses out on revenue because of a business relationship out of its control.

How to Fix This Auction

The mistakes in this auction could have been prevented with some patent auction experience and a better auction design.

Don't Publicly Disclose the Bidders! - This is easy to do, and very important. The number of bidders and the identify of the bidders should be kept top secret at all costs. This ensures that all interested bidders will take part, and that a true demand curve will be created for the auction.

The Starting Price Should Be Researched in Private! - It is up to the auction manager to do the research necessary to determine the starting price in the auction. Just a little bit of research on this matter would have allowed the auction administrators to see that a starting price of around $500M - $800M would have been appropriate. A starting price too low also has no negatives besides making the auction last a few hours longer.

All Prices Should Come From the Auction Admins! - Never ever allow bidders to enter their own prices. Besides using them for signaling, it can slow the pace of the auction and allow collusion. All prices should be set by the auction administrators using a publicly disclosed algorithm (e.g. the price will go up 5% each round).

Divide Up the Products Appropriately! - The auction administrators should have done more research into the patents to determine their specific technological fields, and then grouped them appropriately into smaller products. (See section below).

Don't Allow Bidders to Form Groups During the Auction! - Of course, this wouldn't be possible since all the bidders are top secret and private already. (See section below).

The Optimal Auctions Solution for the Nortel Auction

How do we allow bidders to own only the patents they want to own, license only the patents they want to license, and do nothing with the patents they have no interest in?

Obviously, the Nortel Auction design did a very poor job of this. Apple ended up owning patents it didn't have much interest in, Microsoft has licenses in patents it doesn't have much interest in, and Google didn't get to license patents it wanted to license.

Product Design - As you see in the auction design above, we've divided out the products to be more logical groupings of the patents. We have a product we called "4G" that companies like Google, Apple, and AT&T would be interested in, but companies like EMC and Sony would have little to no interest in. Conversely, we have a product we called "Storage" that EMC would be very interested in, and Apple would have little interest in. We've created as many products as we felt were natural (our auction experience shows us about 10-20 products is ideal).

Bidding Design - In order to meet the needs of every bidder in the auction, we have created 3 possible bids on each product. In any round, a bidder can choose to "Own" the product, to "License" the product, or to place "No Bid" on the product.

In every round, if more than 1 bidder bids to "Own" the product, a winner would be chosen by random numbers, and the price would go up by 5%. Bidding to "License" the product allows an unlimited number of bidders, but the price to license the product would be set at a percentage of the "Own" price. A bidder can also choose to "No Bid" a product in a round. Finally, and this is important, bidders must choose one of those 3 options every round. This ensures a timely completion of the auction with no bid sniping, ensuring an efficient auction process.

End of Auction - the auction would end when there's 1 (or none) bid to "Own" each product. At that point, each product would have an owner at the final price they bid to "Own" the product, and each product would also have an unlimited number of licensers, at a percentage price of the "Own" price.

How the Nortel Auction Should  Have Happened

Using Optimal Auctions' design, how would the Nortel Auction really happened?

4G

Take the "4G" product as an example. During the auction, Apple and Google would have bid every round to "Own" the patents in this product. Microsoft and Intel would have bid every round to "License" the patents in the product.

In round 6, the price for "4G" would have been $3B and Apple and Google would have both bid this amount, with Apple being the winner by a random tie-breaker. Microsoft and Intel would both have bid "License" on the "4G" product, at a price of 10% of the sale price ($300M). RPX would have been "No Bid" on this product, and would have no access to the "4G" product.

In round 12, with the price at $4B, Apple would bid again to "Own" the "4G" product. Google would determine that the price is too high and instead switch their bid from "Own" to "Licenese" at $400M. Microsoft and Intel would also bid to "License" the "4G" product at $400M.

After the round 12 results are calculated, the auction would have ended, since Apple was the only bidder on the "Own". Apple would own the patents in the "4G" product at $4B, with Microsoft, Intel, and Google all paying Apple $400M (10%) to license the patents.