Once, a long, long time ago, I came across a custom 404 error page using mod_speling. It was a photography page and I was looking for the embedded jpg. The website was using a number of javascript and CSS tricks to beautify the page layout, and in the process disabled to right button mouse click over an image. Instead I viewed the source to get the image URL. When I copied it into the address bar, however, I misspelled it. The handy custom 404 error message that popped up suggested the correct spelling with a hyperlink to the image.

Being a tech geek, and dyslexic, I instantly wanted to add spell check functionality to my error pages. Alas, my hosting company did not turn mod_spelling on for shared hosting.

That didn’t mean my 404 pages couldn’t be smart, or at least smarter. I implemented a referrer check, to see if the 404 page was generated by an internal linking error, and to e-mail me so I could correct the problem. I also added the ability to break a URL up into a query string so someone could search for the intended page. This was all back in 2009, hence the links from bluecentauri (the precursor to SarahKTyler.com apps, the precursor to Datayze.com.)

For the past seven years my error pages have remained largely unchanged. I did modify the email code section to email me regardless of referrer. Referrer links is how I survive. If someone is posting a link to my apps, there’s a good bet it’s in a online forum, they like the app, and they think others in the forum will like the app as well. Those are valuable eyeballs and I want to make sure they get where they’re intending to go! About 1 time out of 5 they’ll be a typo in the URL link. I’m usually on it within 2 hours, thanks to the email alert. That’s still two hours of visitors encountering 404 messages, and possibly giving up before reaching the intended app.

It dawned on me this morning that I don’t need mod_speling. I could implement it’s functionality in PHP!

The basic functionality is really simple. Start with the page URL ($url = $_SERVER[‘REQUEST_URI’]), iterate through a list of my app pages (foreach($applist as $app_url)), and check if one is a substring of the other (if substr($url, $app_url) !== false || substr($app_url, $url) !== false). The extra equal sign is important, since substr can return a number, and zero evaluates to false. This catches the case where the URL copied by the poster is not the full URL, or there’s a trailing space or punctuation afterwards like an extra quotation mark. Next I checked the edit distance using the levenshtein function to catch misspellings, which calculates the number of keystrokes needed to convert one string into another.

So I did! And I wrote up a nice how to tutorial for Datayze.

I went back over my error logs and tested the new script on every externally linked URL that caused a 404 error. It corrected every single one.

Too bad I didn’t think of this much sooner.

Today I opened the https property view of datayze.com and discovered this.

Datayze.com is getting more search result clicks than SarahKTyler.com ever received, and the slope of the line is increasing faster than the slopes in graphs generated for SarahKTyler.com. Whether that’s due to the more professional name, the paramaterized site map, use of secure protocol, or just plain luck is unclear. What is clear is a very nice trend line with a very nice slope that shows no sign of slowing down.

That’s a huge positive signal that I should continue my business. Search engine result clicks nearly doubled in a two week period.

Another neat observation? For the first time there’s a clear weekly pattern.

The number of clicks I was receiving before this point has been small enough that minor fluctuations were clearly visible. As such, there was sometimes dips around the weekends, but dips would also randomly occur during weekdays, as would spikes on weekends. The more clicks received on a daily basis, the less influence normal variations will have, and the trend line will become smoother and smoother. The fact that you can look at the above graph and easily predict the weekends is yet another sign of how much the number of search engine result clicks has increased.

I find this trend line particularly encouraging because search clicks tend to be more stable than referral traffic, which tends to spike when a link in shared and die within a day or two. I do expect the steady increasing to slow down at some point, perhaps when the entirety of the paramaterized URLs are indexed by google for example, but I don’t expect the trend to reverse. I think I’m reaching the next level.

Grow little business, Grow.

It took about two and a half months, but things are finally starting to return to the steady state from before the name switch. In some ways, my metrics have improved.

This past month I earned $36.51, my third most profitable month! Users are up by 47%, bounce rate is still relatively low at 2.4%, and I’m starting to see some traction on my lesser known apps. I had my first stretch of $2/day (lasting 8 days), and my highest day yet ($4.03). That may not seem like a lot, but it wasn’t that long ago that I made my first non trivial stretch of $1/day.

The bump in audience came is a continuation of a larger trend that came mostly at the second half of the month, and appears to correlate to an increase in search traffic rather than referral bumps. I’m continuing to see an increase in traffic for the paramaterized URLs, so it looks like I have the addition of the site map to thank for the increased traffic. Google has now indexed just under 50% of my site, and my site has doubled in size since last month! Bing webmaster is still lagging way behind. It seems there was an issue with my sitemap index. I’ll have to look into it, but since Bing makes up such a small percentage of my referral traffic (likely because of the failure to index the site) that it’s a low priority.

Part of the reason my income is up is because the value of a click seems to have increased. If I factor in as many clicks as I got in May, assuming the same rate per click, I would have earned closer to $50. That may seem like a good thing, but the increase could be a result of the time of year (it’s almost the shopping season!) and may be something I can’t rely on.

There’s certainly potential to have some very good months if Google indexes more of my site, and if the number of clicks returns to normal.

This month I released two new apps:

Word Analyzer which provides meta information about a given word, such as syllable count and audience familiarity, and can get you insight into how use of the word may affect readability metrics. It currently tells you if a word is misspelled, but doesn’t suggest corrected spellings. That’ll be the next step. I also intend for it to offer more readable/less readable alternatives but that might take a little while since I’m currently lacking data. I went ahead and released this app before it had full functionality to give Google a chance to index it.

Time to Conception Calculator which calculates the probability of conceiving within a given time frame for someone of a certain age. I know, I’m working backwards. I started with the Labor Probability Calculator, then moved on to the Miscarriage Reassurer, and am just now looking into conception. It would have been wiser to have gone the other way! That way I could keep the audience I’d acquire as they moved though each stage of their pregnancy. Now I have to hope interested parties eventually find the app on their own.