Nothing. The answer was nothing.

In December I told the first part of a true* story about a Christmas Miracle that had occurred some years before on a UK mobile network. Two weeks before Xmas, the quality of the network suddenly jumped upwards. A significant improvement in quality was maintained throughout December, but as the clocks chimed midnight to call out the New Year, network quality started to fall back down again. At the end of this post I ran a poll, asking: "What was it that caused network quality to improve?"

"Activate additional capacity" was the most popular answer with 47% of people selected this option. It was wrong. Or rather, it was a plausible answer - but not the reason why quality improved in this case. In fact, none of the options given were the reason that the Christmas Miracle occurred. I'm sorry for posing a "Trick" question, but if it makes you feel better, the poll options are the first things that came to my mind when I saw the original KPI chart, many years ago.

So what was it that caused the Christmas Miracle? Nothing. Nothing was the correct answer. Now, before you get fed up, I'd better explain ... and quickly. 'Nothing' means that no changes were made to the network: no new hardware was deployed, no new software was rolled out, no new features were switched on ... and under no circumstances were any of the thousands of different network parameters optimized!!! What actually happened was a "Data Freeze" - for a 4 week period, no-one was allowed to make any changes to the network (with a few rare exceptions). The result was that the network was able to stabilize itself, and quality improved without hundreds of (well-meaning) humans making trade-offs to solve isolated issues and trying to find new ways to boost performance. After the Data Freeze ended, a backlog of pent-up change requests soon dragged network quality back down towards it's original level.

So now you know the answer - but is this story just a quirky anecdote from the past? And what relevance does it have to the networks of today - and tomorrow? Of course, if you've read any of my previous articles you'll know by now that 3 x key takeaways are on the way:

1. Sometimes less is more: in this case, fewer humans "touching" the network led to better, not worse, quality. That's not meant to be a judgement on the skills of highly trained and vastly experienced telecoms engineers - but an assertion that networks today are simply too complex for humans to fully understand all of the dependencies of the system (as a whole). We know well enough how to fix a specific problem - but we can't always comprehend the potential butterfly effects.
2. If we cannot rely on humans alone, then the case for AI in telecom networks today is irrefutable. But when we apply AI to network and service operations, we should think carefully about the right benchmarks to use to judge success. Humans aren't always perfect. And data is rarely perfect. So why do we expect AI to be right, 100% of the time? That's probably a topic for another article, but in the meantime I'd encourage you all to make a New Year's resolution to stop being so unfair to the robots! :)
3. Statement of the obvious: there are some really big issues that we need to fix if we are ever going to realize the vision of Autonomous Networks. And while AI can help us master complexity and handle many of these problems, I also believe that human creativity will be essential. We will need humans to design and build the networks of the future - and to put the necessary guardrails and control mechanisms in place.

Happy New Year!

* It really is a true story, although it is possible that I've forgotten some things - and I have taken a few creative liberties to (hopefully) make the story easier to understand and more compelling.

https://www.dhirubhai.net/posts/andrewburrell_true-story-many-years-ago-a-christmas-miracle-activity-7275145657798082560-4hD1?utm_source=share&utm_medium=member_desktop