Your browser is not supported. Please update it.

29 November 2018

Is Akamai getting ahead of the game in analytics?

CDN pioneer Akamai appears to have made a substantial step forward in the analytics game, demonstrating during a company webinar this week how the industry is getting closer to closing the frustrating gap between the lines of monitoring a problem within a network and actually getting around to solving said problem as quickly as possible.

Following a major update to Akamai’s mPulse What-If Analysis feature in October, the company is now offering customers the ability to predict key business metrics by modifying or removing components and viewing what the end result on their website, app or stream looks like, in a virtual environment. Faultline Online Reporter has seen a lot of analytics dashboards this year and none have impressed us more than What-If v3.0.

The technology is a potential game changer, not purely for service providers, but technology vendors too, as Akamai can show a service provider exactly how a third-party asset is influencing quality of experience. If a particular asset is found to be slowing down page load times, for example, What-If can suggest either pulling the plug or making smaller modifications and therefore see the end user results, before actually going ahead and applying it to the real end user experience.

The webinar, called ‘Preparing for Glory in 2019: are you peak ready?’, centered around live sports with specific references to the World Cup this summer, which hit a peak record of 23 Tbps on Akamai’s CDN infrastructure. This demand naturally caused congestion and induced masses of complaints about poor live streaming experiences. After snippets of What-If’s features were showcased, we asked the Akamai panel if any third-party assets were standout culprits in contributing to poor streaming performance during the World Cup?

“It would be impolite to highlight any specific companies but the most challenges were caused by personalization software,” replied principal software engineer Simon Hearne, in what was one of the key takeaways from the webinar. Faultline Online Reporter is guilty at times of criticizing certain video service providers for not building in more personalized features and powerful recommendation engines, without really accounting for the fundamental network issues and damaged reputations they may cause.

What-If can then break analytics down to another dimension which is page groups, including time on page, pages per session and session duration, where one small change to an area page can bump up key business metrics. This feature is also available in native apps as well as mobile browser pages, supported with an SDK rather than a Javascript tag, and Hearne said Akamai is seeing a big focus on improving performance on Android devices due to easier control compared to iOS.

In summary, Akamai says What-If can spin up possibilities such as how making minute changes to metrics such as time to interactive can substantially improve bounce rate – leading to hundreds of thousands of extra conversions. It can also project revenues from different scenarios in its virtual environment, and filter resources by load duration and let customers play around with speeds, for example changing to a 10ms delay and assessing if the ROI is worth it.

The screenshots provided gave the impression of as close as you can get right now to an all-encompassing bit of kit, with the ability to name and shame specific third-party vendors potentially raising real concern for some companies – particularly in the ruthless live sports space. If a broadcaster, for example, identifies a particular third-party asset wreaking havoc on speeds in its virtual environment, then generally the action taken by customers is to either show the data to the partner directly or take it to their in-house development teams, according to Akamai. How long before What-If has blood on its hands as a direct result of a switch out job?

Reputation was the webinar’s central, somewhat scaremongering selling point, with many operators, broadcasters and streaming companies apparently approaching Akamai with woes of mounting social media criticism about the quality of their video services.

Another newly updated Akamai product, CloudTest, claims to take performance testing to the next stage, by building and generating at the scale customers want using realistic tests from the cloud (boasting over 6 million cloud servers). Gareth Lynn, General Manager, gave a brief account of an anonymous European broadcasting customer delivering large scale, subscription-based live streaming events. It planned to enter new markets but first wanted to address the negative press on social media about its poor UX. Akamai assigned an experienced test engineer, who prepared the script, conducted tests, then provided an extensive 20 to 30-page result summary comprising areas of concern and recommendations, using CloudTest technology. It apparently identified a number of configuration and technical issues which required fixes, although Lynn neglected to say if the story had a happy ending.

Machine learning was not mentioned at all in the webinar, coming as a small surprise given the buzz around adding ML capabilities to analytics products, of which we have seen an abundance of demos of this year – not necessarily to be seen as a negative of course.

“The scary thing about third party assets as a website or app owner is feeling out of control, i.e. on the device itself, with our data showing a direct correlation between CPU performance and third-party components, therefore delaying analytics and leading to either removal or modification,” said Hearne.

One thing is for certain, if Akamai released a list of customers using What-If, it would probably radically change our perspective of what influences decision making in the industry.