Good ad tech should improve campaign performance for advertisers, media performance for publishers and user experience for consumers.

Note: I appreciate how busy we all are; therefore, if you want to skim this long article please just skip to the part at the end labeled “Summary“.

Background

In this years IAB MeasureUp conference I presented something that I call “ALT” which stands for Ad-Tech Latency Tax. It is the effects of internet latency on ad tech and how that affects media viewability. The presentation started by recapping the previous IAB events which covered these three topics:

  • Frankenteck – The idea that, on average, there are upwards of 10 x pieces of ad tech which are all used in unison to bring 1 x piece of digital advertising to a consumer.
  • The Clippet – The idea that, if you are not careful, that the majority of the media budget could be spent on ad tech and not on media.
  • Oranges vs. Mangoes – The idea to separate ad tech into two categories (a) mangoes which have a large core of technology surrounded by a layer of palpy service vs (b) oranges which have much smaller core piece of technology and you end up paying for the service.

Overview

A new topic was introduced to build on the top of the previous topics. This new topic was the concept of Ad-Tech Latency (ATL). The hypothetical question was simple; Does too much ad tech slow down websites and degrade overall media performance?

This report started by looking at the baseline of global internet speeds from the free and independent company SpeedTest.net. The report showed that average internet speed in our country were 7.8 Mbps which is much slower than most developed countries.

In my personal opinion the core of the problem is that the NBN (National Broadband Network) has one core fundamental flaw. For some countries with fast internet their version of the NBN has internet connected to the exchange but in our country we have a lackadaisical attitude that NBN should be connected all the way to our lounge rooms.

I have first hand experience of the NBN rollout in Sweden whereby houses are grouped together by the local council and each house pays circa $3K to have fibre optic cables laid underground from the exchange to the house. This money is amortized over 3 years, at 0% interest rate, which is subsidised by the government and costs each house only about $50 a month. As a result, this type of infrastructure will offer a massive 200 Mbps for the same price as average Australian’s are paying for only their paultry 7.8Mbps.

This raises the question: If the local internet is slow, then it is slowing down ad tech providers?

A good place to start with local ad tech providers is MediaScape as it contains a list of many of the leading ad tech companies in the country. I picked a random sample of 20 x common ad-tech companies for this report. This presentation was not designed to name & shame so from this point forwards in the document each ad tech vendor was anonymised and given assigned a random number from 1 to 20.

The Results

Using the free tool called PING which is part of the Windows Operating system I called each of these ad tech providers 1,000 times and took the average response time for each server. The results below were pretty interesting. It showed that (a) most ad tech providers had their servers in Singapore and (b) the average response time from a server in Singapore is 150 ms.

This report hopefully now shows two clear questions being raised;

  1. Frankenteck – There are lots of ad calls for 1 x digital creative to be seen and
  2. Ad-Tech Latency (ATL) – Each individual Ad-Tech can take to 150ms to respond

Combining these two points questions gives us the following question.

Does ad tech have a negative effect on website performance?

In order to measure this I took a look at some of the most common websites that operate in our market. To find the websites I looked at the free online tool called SimilarWeb.com. This is a great resources which works globally and can help you find interesting pieces of information about the majority of websites on the planet.

In the screenshot above I’ve added a few “ring-ins” to this list which include (a) W3C and Wikepedia as these free reference websites have massive global internet traffic and famously have no internet advertising and (b) I’ve also included one US based political news portal as it has a large volume of traffic but also has a very large amount of advertising. Therefore, these ring-ins should be the extreme points and used a references in any testing.

Each of these sites was tested with another free tool called HTTP Watch. This tool plugs into your browser and lets you see what code is running in the background whilst your browser is loading a web site. It’s a fantastic free tool and I urge everybody to download it learn more about how a browser works.

The results from HTTP Watch were very good and helped me understand the affect of digital advertising on these sites. The green bars above represent the speed to which the DOM loads on the web site whilst the RED bars are when the page is fully loaded.

Think of a website as a oil painting on a canvas. The DOM (Document Object Model) is like the frame of the painting and the Page Load is when the white canvas has been fully painted.

In the graphs above when the RED bards are high and the the green bars are load it means that you (the consumer) is seeing whitespace whilst you wait for the page to load. These tests showed that, on average, the average page load time for popular local web sites was 13 seconds.

To check if the page load time was affected by other factors instead of just ad tech then I looked at three individual tests which I will summarise very briefly below.

  • Google Speed Index – I used the free Google Page Speed tool which looks at how well a website is coded. This showed clearly that local websites in AU were trimmed for performance just as well as our UK or US counterparts.
  • W3C Validator – This is a free tool which checks if the HTML in a page is coded correctly. In our paintinganalogyit is like checking that the frame is well mounted well together and the content isn’t going to fall outside the frame. My tests showed that the HTML was well written and there were no problems there.
  • Page Load – I also tested to see if local sites were “heavier” than their counterparts in the US or UK and I found that they weren’t. Most popular news portals globally seem to have a page size of about 1.7 Mb per page.

Now, after we showed that local websites were just as good as international ones, it was time to look at the affect that ad tech has on a website.

The free tool called Evidon (nee Ghostry) was used to look at different sites. Firstly a popular US based site was checked and we can see a number of interesting things. Firstly we can see that there is allot of technology behind the page call. (For instance, in the screen shot above you can see that there are 238 ad-tech calls form 29 unique companies with 1 x page load). However, secondly we can see that the page responds relatively quickly (this can be seen by the green donut in the top left corner which shows that all the tech is responding quickly).

However, it’s a totally different picture when we look at a local AU website.

In this screen shot above we can see (a) far fewer ad tech calls (only 137 calls from 20 vendors) but we can see a much higher overall latency (as seen by the red donut with much higher latency).

From this it could be seen that US based sites respond faster because (a) the consumer is in the US, (b) the web site is in the US, and (c) the ad tech is in the US. Whilst in our case we have all our ad tech in Singapore which causes much more latency.

To bring all this research together it was important to see if this extra latency from having so much ad-tech in Singapore really affected performance of media. In order to do this we turned to over 8 billion rows of Viewability data provided by Moat. This data looked at the 100% On Screen definition of Viewability and this was plotted against the free HTTP Watch load time and analysed in Microsoft Excel. The results were as follows.

The Summary

Viewability load time

The horizontal axis on the chart above is the time it takes for the page to load and the vertical axis is the average viewability. The line on the chart is a linear trendline created by Microsoft Excel. The angle of the trendline shows that there is a roughly linear relationship between Viewability and Page Load time.

Every 1 second improvement in page load time will improve viewability by circa 1%

The Epilogue

Whilst this result didn’t surprise me, it was pleasing to see that it could be proved. It was also pleasing to know that most of this research was done with free tools that anyone can get online. They are listed below.

  • Similar Web – Great free tool which allows anybody to look at key facts about a website including visitors, growth etc… They also have a pro service which can allow you to get even more information.
  • HTTP Watch – Great free tool which allows you to see what’s happening when a page loads. They also have a pro version which allows you to get more info.
  • Google PageSpeed Test – This is a totally free tool which allows you to get a Performance Score based on how anyweb site is coded. This works just as well for Desktop and Mobile.
  • W3C Validator – This is the definitive way to check if aweb site has any coding errors or not. It’s totally free as it’s open source technology.
  • Evidon – This is a great tool to see what Ad-Tags are on any page. The technology was previously called Ghostry. They also have a pro-tool which allows for even more information.
  • SpeedTest – This is also a free tool which allows you to measure the speed to connect between yourself and a server. If you create a login then you can use the data to see global benchmarks and global trends.

The Conclusion

One second of faster page load time improves viewability by about 1%. Therefore:

  • Suggestion For Publishers – Optimize pages to load as fast as possible.
  • Suggestion For Advertisers – Be mindful of to work with ad-tech companies which are designed to work in your local market.
  • Suggestion For Ad-Tech Companies – Colocate technology in your local market, or use local cloud providers where possible.

Further Reading

Previous post

INTERVIEW: Ad Tech’s ‘Godfather’ Dr. Boris Speaks

Next post

Amazon Australia – All Huff, no Puff?

Join the digital transformation discussion and sign up for the Which-50 Irregular Insights newsletter.