Congratulations to Josh on the del.ico.us acquisition. Yahoo will make a great partner for the bookmarking service.

Now a little part of me is cringing as I write this. Having founded a bookmarking company in 1999 with pretty much the exact same vision as the new crop of services, I’ve got to feel, well, a little stupid. (or angry, or depressed, or whatever). Maybe writing about it will make me feel better and maybe even help me make a point or two about product development.

When we founded Blink.com (no link love, it’s a crappy search site now) the founders and I imagined a self-reinforcing product cycle:

1. Consumers needed portable bookmarks so they wouldn’t lose them, would be able to access them from any computer, and could share them with friends or coworkers;

2. As part of the process of bookmarking sites and organizing them into "folders" users would be indicating a measure of quality and connectedness among the URLs;

3. Profit!

OK, step 3 was a little more complicated. But the essence was that we would use the personal-backup product attributes to create a public search engine and "discovery engine" (I believe the marketing folks wanted to use that phrase!) based on user bookmarks.

This really shouldn’t sound too different from what del.ico.us was able to do, and we had something like $13 million to play with to make it happen. Not to mention that there were others with the same idea. Remember Backflip? So (besides the money), why did we fail and del.ico.us and the other Web 2.0 companies succeed?

I don’t think it was that we were "too early" or that we got killed when the bubble burst. I believe it all came down to product design, and to some very slight differences in approach.

To start, we launched Blink with a bevy of marketing dollars and a message very much focused on the individual storage benefits. We were very successful at attracting users (at its height Blink has 1.5 million members, del.ico.us currently has 300,000) and getting them to import their bookmarks into our system.

Mistake: Defaults Matter

Members were given the ability to make their folders "public", which would then allow those folders to be seen on a member page (here’s mine). Folders were private by default, so users had to choose to make folders public one at a time. This severely limited the proportion of our bookmarks that were made public. In retrospect I think we didn’t fully buy in to our vision of a shared bookmark database, and as a result we designed the product with an over-emphasis on the private bookmarking aspect, instead of the public. We were also sure that users would have vast troves of porn bookmarks.

Mistake: Folders Suck

Our first iteration on using bookmarks to create a shared information library was an extension of the "public folder" concept. We believed that users would not only make their folders public, but also would categorize those folders into a directory structure. We called this the "Public Library" and created a Yahoo-like node structure on which users could post. This could have made sense since categorizing folders would be less work than categorizing individual bookmarks – after all, the folders were already "categories" of a sort.

There were several severe problems with this folder-based approach. First, people are very bad and inconsistent at organizing things. One day etrade.com will go into the "finance" folder and another day it will go into the "favorite links" folder. We were taking this fundamental flaw and squaring it – asking users to use graph their existing categorization onto a second arbitrary structure within the public library. Does my "finance" folder go into the "Business" directory or the "Personal" directory?

Then there was the issue of how deep to go when categorizing folders. If I’ve got a folder of "online brokerages" do I put it in the directory at the level of "Finance" since my folder is in a sense a sub-category of finance, or do I put it within the pre-existing "Finance -> Brokerage" directory? Users were confused, and with good reason.

Mistake: Make it Instantly Useful

Even with the severe ontological problems presented by our flawed public folder concept, we still had thousands of new folder postings per week. Users liked the concept of sharing bookmarks and were anxious to give it a try. But how useful was all the data being posted?

Once again, the folder concept doomed us. We only displayed what was posted on the site, which were folders. So if you were to search or navigate to the "Finance" section of our directory you would see…folders…hundreds of them. All of which had been given the helpful name "Finance" by their owners. Or maybe the more helpful "Financial sites" or "Finance bookmarks" or "Finance and banking". You get the point.

Why didn’t we just list out the top 10 bookmarks within each category? Well by building a folder based system we had made the job of calculating the union of all this data an order of magnitude more complex. The query to find the top sites that existed in hundreds of folders would have brought down our (poorly designed) database in a heartbeat. Eventually we did build something like this, but by that time it was too late.

Mistake: Don’t Let Technology Decide

We understood that actively searching for new bookmarks would be a fairly rare activity among our users. We needed to build a way to let them discover new sites relating to their own interests in a more passive way.

The engineering team went to work, building a complex algorithm for evaluating the groupings of sites within folders (damn folders!) and finding other sites that had been grouped similarly. It was pretty sophisticated stuff. Sophisticated enough that it couldn’t run in real-time and often had a several day processing backlog to overcome.

We added a "find similar" button to the Blink interface, and the results were often quite good. The problem was that we were once again asking the user to go out and do things. The vision was that the similar sites would just be there, the same way Amazon presents you with the related products. But the servers couldn’t handle it. They could barely handle it when the users actually clicked to see results, let alone on every pageview.

I suggested a compromise. Instead of the algorithm, how about just using the text of the folder name as the key. Show the top 10 sites in all folders with the same name, across all users. I was basically suggesting a rudimentary tagging system, using folder names as tags. I was voted down – too simplistic, too hacky. And anyway we had this sophisticated algorithm, which might work one day …

The Bottom Line: Getting It Right

This post wasn’t meant as a defense of Blink or my own decisions while I was there. My intent was to show that product design matters. We had more money, more users, a five year head start, and some really, really smart people working on bookmarking in 1999. The bottom line is that we simply didn’t get it right. Some simple innovations like using tags instead of folders, making public the default, building better discovery features, etc made the difference between being an also-ran and a hot acquisition target.