On Wednesday the 15th of June 2011 at 21:22:30 GMT+2 a lunar eclipse took place over the African sky. This eclipse led over 15 thousand unique visitors to click through to the website www.lunareclipse.co.za in just 4 days, and got the site mentioned on the SABC3 News and the front page of the Cape Argus.

How it all happened?

It all started with the video below, which got me thinking about the idea of selecting high volume, once-off or infrequent events, in order to identify low cost adwords which I could use to send users to a useful site that contained relevant information on the event, as well as some advertising for my company’s product.

Converse Domaination from Ross Martin on Vimeo.

So I registered the lunareclipse.co.za domain name and rolled out a default WordPress install. With the help of my boss, we put together some interesting content and images that helped to explain the lunar eclipse to the layman. Lastly, I added some social sharing buttons (Facebook, Twitter, and G+) and switched out the default WordPress comments in favour of the Facebook comments plugin.

With the eclipse taking place on Wednesday the 15th of June, we began the week with just over a thousand unique visitors on Monday. On Tuesday, this more than doubled to just under four thousand visitors.

On the day of the eclipse, Wednesday, the event started to get a huge amount of media attention and visitors on the site shot up. By the end of Wednesday, we had just under 10 000 unique visitors! Unfortunately my server took a dive a few hours before the eclipse, as the national television news channel, SABC3, used an image from our website with a reference to our URL!

Interestingly, Adwords reported that we received over 13 000 clicks on Wednesday. I suspect that around 3000 of the clicks that we missed were as a result of the server being down (eeek!). Also of note is the high click through rate and the resulting low Average CPC. For international visitors, 1USD = about 7 rand. So avg. CPC was sitting around 7 dollar cents!

What I learnt?

1. Social
I think one of the biggest learnings for me was the impact of “social media”, especially when used in the context of a trending topic. Lets look at the numbers.

• Facebook: 473 likes & 43 comments.
• Twitter: 44 tweets sent from the site & a number of mentions of the URL.
• Google+: 8

The real power of social media, however, came from the comments section of the website. A number of readers added their views and thoughts to the site and this is ultimately the content that was quoted in the Cape Argus newspaper report. The press wanted to know how the average person felt about the eclipse and the social aspect of the site could provide that to them.

2. Relevance
While I can’t release the exact numbers on how well the PropIQ adverts performed on the website, I can say that it wasn’t as successful as I had hoped it would be. We believed that a number of people visiting the lunar eclipse website would not know that one could buy a property valuation and that they would indirectly find out about it while browsing the site.

Unfortunately this didn’t happen, teaching us that relevance is hugely important!

3. Scalability
The lunareclipse.co.za website was setup on my entry level linode VPS, using a LAMP stack. To begin with, I wasn’t expecting much traffic to the website so didn’t do any planning on how I would scale the website if we were to get a sudden influx. On the Tuesday afternoon, after having received over 2000 visits on the day, I turned caching on after having noticed the server was taking some strain.

By Wednesday afternoon, however, the server was getting absolutely hammered, with Apache not able to handle all the concurrent connections. I reduced my KeepAliveTimout on apache to 2 seconds and this helped to free up some of the apache processes to serve more visitors.

Ultimately, my lack of scalability planning meant that the site was inaccessible for almost two hours on the day of the lunar eclipse – doh!

Next time: Nginx, enable MySQL Query Cache, and possibly (especially if it’s just for a short once-off event) host on AWS with an elastic load balancer and Amazon’s RDS for the database.

Conclusion

I had loads of fun with this little experiment and learnt a huge amount about social, relevance, and scalability.

The overall take-away for me, however, was the importance of being able to be nimble when it counts. I work in an environment where development is planned many months in advance and new projects go through rigorous testing before they’re released. This is crucial when you’re releasing to paying customers who rely on your service being stable and reliable.

When you’re trying to latch on to  a hot trend or talking point, however, it’s imperative that you are able to build fast and release soon – even if it means you run into a couple of hiccups when things start to take-off.