Posted by Matt Mazur (@mhmazur)
On Monday, January 16, 2012 I launched Lean Domain Search by posting a link to it on HackerNews, making today it's six month anniversary (woohoo!). In this post I'll go over the major changes to the site since its launch, review some of its key metrics, and end with a glimpse of where its headed next.
Interface and features: then and now
To get an appreciation for just how far the site has come in six months, you need look no further than the search results page.
Here’s what it looked like at launch vs. what it looks like today.
Click to view full size:
A few things to look out for:
- The color scheme will be the first major change that jumps out at you. When it launched, I was trying to differentiate the site from all the light-themed domain search tools out there. Quite a few people complained that the shade of green it used was too intense, so I went through several iterations trying to adjust the shade of green before deciding to move away from the dark design and over to the blue design that’s in use today.
- The original version lacked sorting and filtering options. I didn't think these weren’t necessary to launch the initial version (which is probably true), but they were so highly requested right after the launch that I wound up adding rudimentary versions within hours of launching.
- There weren't any share buttons on the original version. While not critical in terms of whether or not the product works, I probably missed out on a lot of traffic because the site was not easier for the initial wave of visitors to share. For future apps, having social sharing buttons built in on launch day is a must.
- The original version lacked a search history. This was also not necessary to launch, so I held off on implementing until a few weeks later.
- On the right in the new version you can see the "Popularity Index", which is a measure of how popular your keyword is based on how often it appears in other registered domain names. I ran an A/B test which showed that users tended to search for more results when they were presented with this tidbit.
- Also, in the original launch the results were not sorted at all — not just that you didn’t have the ability to sort them, but the order of the search results was more or less random. I eventually figured out a way to gauge the popularity of the prefixes and suffixes used in the results and order them that way by default, which is why the search results these days seem much better.
- One other obvious change is that the search results are broken up into sections to make them easier to browse without losing your place.
- Also not shown here, but when you double-check that a domain name is still available by clicking on it in the search results, Lean Domain Search will automatically check whether the Twitter name is still available and whether or not there is an existing trademark that it might infringe upon.
I’m happy with these changes because almost every one contributed to making the core product better. It's so easy to start tacking on features when you should be improving the features that are the essence of your product.
Metrics, metrics, and more metrics
Number of searches per week
The spike on April 2 was a blog post where I announced that I had been adding 100 search results to Lean Domain Search per week since the launch and it now checked 2,000 domain names per search.
The spike in mid-April was due to another blog post, this one about calculating the top domain name prefixes and suffixes, and a mention several days later on a The Next Web article about naming your startup.
Finally, the spike in mid-June was due to a mention by Cyrus Shepard in an article he wrote for SEOmoz about picking a domain name, where he listed Lean Domain Search as the top domain name search tool.
For the curious, if you add up each month the total number of searches is just over 200,000.
Average number of searches per user
When I launched Lean Domain Search, I thought the average number of searches per user would be an important metric for gauging how useful the tool was. After all, if the average number of searches per user is low then it likely means that folks perform their first search don't like what they see and leave never to come back. The problem with this line of thinking is that if the search results are really, really good then you'd also expect to see this number low because it means folks are finding domain names that they like and are done searching. I include this only to emphasize that you can't just blindly look at metrics; you have to understand how to correctly interpret them.
Average number of available domain names returned per search
When Lean Domain Search launched, it checked the availability of 1,000 domain names per search. Almost immediately, I set out to add 100 new search results per week — that's why the average number of search results rises so steadily in the beginning. As you'll see in the next graph, these extra search results came at a cost in terms of time though, so after hitting 2,000 I put the breaks on and started looking for ways to improve the search speed.
As I made improvements in the speed, I continued adding search results; it now checks 2,500 domain names per search.
The funny thing is that even though it can now handle more in terms of capacity, I can't do it without degrading the quality of the results because I've already added almost all of the "good" prefixes and suffixes.
Average amount of time it takes to return the search results
As noted above, as I added more search results, the amount of time it took to perform those searches steadily rose as well. I eventually started caching the search results so that if someone searches for a term that has already been searched for, it looks those results up in a database instead of performing the search the time-consuming algorithmic way.
I also spent a lot of time trying to devise new ways of performing the searches that would decrease the amount of time it took for non-cached queries, the results of which you can see starting at the end of June.
Percentage of searches that resulted in an error
For a variety of reasons, sometimes a user's search would not finish successfully. For example, some queries took upwards of 30 seconds to complete, causing the search to fail. At it's peak, about 4% of searches or 1 in 25 resulted in an error. Over time, I developed ways of reducing the number of errors — today, the error rate stands at about 0.3%, or 1 in 333.
Percentage of double-checked domain names that are actually registered
As noted above, the search results are calculated using Verisign's .com zone file, which supposedly contains all of the registered .com domain names but doesn't. As a result, some of the domain names that show up as available actually aren't. Also, as people register domain names (both Lean Domain Search users and everyone else), some of the domain names that were available no longer will be.
When you click on an available search result in Lean Domain Search, it automatically double-checks that that domain name is still available. The graph above shows what percentage of domain names that are double-checked are actually registered. As you can see, early on almost 3 in 10 "available" domains were no longer available. These days, it's less than 1 in 10 (and since Lean Domain Search tracks which are registered and no longer shows them as available, this number will keep on improving over time).
Lean Domain Search is my fourth and most successful web app in terms of both revenue and popularity.
(One good decision I made early on was to immediately add nice things folks were saying about Lean Domain Search to its testimonials page. Since launching, I've collected 160 unique testimonials, a testament to how much people seem to enjoy it.)
Looking forward, I see three primary areas to focus on:
- Continue improving the search interface to make it easier to find keyword-based domain names
- Explore methods for generating quality generic domain names
- Invest more time marketing the site in order to educate people about its existence
The last point is especially important because no matter how good or how fast the site is, it doesn't matter one bit to the 99.9% of folks struggling to find a domain name because they don't know Lean Domain Search exists.
Here's to the next six months and beyond —