We ran a test of Alert Carolina recently and have some real world SMS delivery numbers that I think are worth a post. We sent a little over 24,000 messages in slightly more than 7 minutes. A key metric in that is that we sent almost 80% of the messages by the second minute and many had already begun to arrive on user’s phones. By 5 minutes, 97% of our registered users had messages on the way to them and a large portion of our campus had likely already received the message.
In a real world campus emergency situation, these are excellent numbers. I have long said we don’t need to reach everyone in a classroom or on campus, just enough people so that they can notify others. We already know that students are going to tell other students what is going on much quicker than we can. If we can start that process with good information we are more able to do an effective job of keeping our campus informed.
From a technical standpoint, it is interesting to dive into some of the numbers in order to get a better understanding of the way SMS messaging, in particular large scale messaging, is handled. The first pass at delivery to all of our registered users took just under 2 minutes. As the aggregators sent back information on those delivery attempts, our vendor’s system began to retry delivery or use alternative delivery mechanisms (for example SMTP versus the initial SMPP push) and continued that until we had sent all of the alerts. Almost all of our messages were sent via SMPP, probably because smaller 3rd tier carriers are not very common among our registered users and therefore we have SMPP pipes available to us.
I think this outlines the complexity involved in getting good SMS delivery. After all, we are using / abusing a technology that wasn’t really designed to do what we are asking it to do and still managing to get really good rates of delivery. The key in this is working with a messaging vendor who deeply understands the issues and the limitations and has invested in appropriate redundancies, pathways and carrier relationships.
This is far different than the post Virginia Tech sales opportunity extravaganza that took place as small companies with untested and unknown systems infrastructure made quick sales to lots of different Universities and other organizations. Despite its limitations, SMS alerting is going to be around for some time and we have to treat it like an enterprise application and invest in the right kinds of partnerships.