Algorithmic Risk

Currency risk is a well known form of risk that arises in businesses with operations, assets, or expenses in more than once currency.

For example, let's pretend I have a business selling socks online (what's that? oh right, I actually do). If my sock business, which is based here in the USA, had an office in the UK (which we actually don't) also selling socks, we would be conducting business in both US dollars and British pounds.

My sock company is now exposed to currency risk.

Here's how: Let's say on average the UK office sells £100 worth of socks per month. My accountant tells me that the UK office made $160 USD worth of sales last month because the conversion rate at the time was £1.00 = $1.60. But if this month dollar appreciates against the pound, and the conversion rate becomes £1.00 = $1.50, and the UK branch sells the exact same amount then I really only made $150. I just lost $10 to currency risk.

Currency risk is well understood and companies have a variety of tools to reduce or hedge that risk. For example, my sock company could simply balance expenses and revenues in th UK. If the UK office also spent £100 per month, then when the currency fluctuated this month, those expenses would go from $160 USD to only $150 USD, keeping expenses in line with revenue and neatly eliminating currency risk. So although companies may not have control over currency fluctuations, there are ways to minimize risk exposure.

This is true of most types of risk - credit risk has credit default swaps, equity risk has all kinds of hedging strategies, interest rate risk has interest rate swaps, and on and on.

But there's a new type of risk that I suspect is on the books of many high-flying tech startups that I don't think has a name. So I'll offer one - Algorithmic Risk.

This article from TechCrunch reveals that online ticketing and event registration company Eventbrite sells on average $2.52 worth of tickets every time a user shares a link to a paid event on Facebook. The post's author, Erick Schonfeld, notes that Facebook sharing is superior to typical email link sharing "because it broadcasts to all your friends". But does it, really?

Last I checked the default setting on the Facebook news feed is "Top News" - a curated news feed based on some black-box algorithms, not a live feed of all your friends' activity.

So what if, all of a sudden, Facebook algorithms decided that Eventbrite's activity items were not interesting? Eventbrite items would stop showing up in feeds, and that $2.52 per share figure would plummet. Facebook as a sales channel would go dark for Eventbrite. That's algorithmic risk.

This post by Fred Wilson speaks to the same thing. Fred notes that as Gmail's new Priority Inbox feature gains traction, it will be more and more difficult for unimportant emails to get the attention of recipients. Though the random community newsletter you don't recall ever signing up for isn't quite spam, it's also not the same as an email from your college buddies or business partners. Priority Inbox can make the distinction and relegate newsletters like this to the backwaters of your inbox. Suddenly, lots of direct marketing companies that rely on generating ROI from email marketing will find themselves in a lot of trouble because Google changed their filtering algorithm. That's algorithmic risk.

These two examples are relatively new, but one area that's always been subject to algorithmic risk is search engine rankings. Small changes by Google to its ranking algorithm, or Adword bidding algorithm can wreak havoc with businesses that count on search results to drive revenue. As a result, the SEO and SEM industries have sprung up, ready to snap into action to remedy any slide in search result rankings.

But this is a really inefficient way to deal with the problem. Look at the financial markets - when a company is exposed to currency risk, do they hire a company that can frantically trade currencies in an attempt to unwind the damage done by fluctuating prices? No! They create mathematically sound hedges. Same for other types of financial risk.

If algorithmic risk is just another type of financial risk, how long before we see PageRank default swaps?

On Nice URLs

URLs don’t have to be tidy, but they have to be usable. If I go to google and search for hello, I can figure out how that part of the querystring works. That’s not the case here because of the double encoding. It’s a non-standard URL parameter and it’s unnatural.

Twitter has great URLs. Super intuitive and clean.

Case in point - can you guess where these URLs lead?
http://twitter.com/cclark
http://twitter.com/cclark/lists
http://twitter.com/cclark/favorites
http://twitter.com/cclark/following

eBay has terrible URLs:
http://feedback.ebay.com/ws/eBayISAPI.dll?ViewFeedback2&userid=cclark&ftab=AllFeedback

If the Twitter folks were running the show, it would look like:
http://ebay.com/cclark/feedback

URLs are just as important as any other part of the application that faces users. I've have got burned on all kinds of weird things over the years by having bad or poorly thought-through URLs. Think of URLs as APIs, think of them as functionality. That's easy for anyone who writes REST services on a regular basis, but can prove a bit more of a leap for developers who don't normally focus on web services.

URLs are not just a place to dump data. They should be guessable, usable, and (of course) bookmarkable.

Closing the Loop on Disrupting Education

Starting at the 3:50 mark in this brief video, my favorite blogger Fred Wilson talks about the recent proliferation of reputation-based web services like Stack Overflow and Quora. I think it will ultimately be reputation based systems like this that close the loop on platforms that have been aiming to disrupt education. There's a lot of talk about web technologies disrupting classic educational institutions, and many of the services (like Sal Kahn's Academy about which you recently blogged) are providing quality content. Along with universities like MIT posting their lectures and class materials online, there doesn't seem to be a lack of quality educational materials.

But what has been missing, for better or for worse, is a way to certify that you have actually learned something. Regardless of the knowledge you may have, if you can't document it and put it on a resume, hiring managers and HR departments won't care. Your education and qualifications must be documented somewhere.

In the past, many have attempted to demonstrate their expertise by blogging about the subject. If you can create a blog with quality content and attract a readership, that's pretty good documentation that you know what you're talking about. However the problem with that route it...you have to know how to attract a readership. Marketing your blog and driving traffic is really a separate skill-set and one that someone trying to prove their expertise in, say, project management, may have relatively little interest in.

Enter reputation-based social services. Now, if you see on a resume "Attended Sal Kahn's Academy and became the #1 rated question-answerer on topic xyz on Quora", that's worth looking into. Someone soon will come up with a way to not just provide educational materials, but also provide an ecosystem where someone's credentials can be established and documented. Delivering both those pieces as a unified platform is the key.

Regardless of the true quality of the education, a name-brand university diploma will always be more valuable than web-based learning until the web-based systems' certification carries as much weight as the name-brand reputation.