Loading Time Is Super-Important! Here's How To Track It On REAL Users (25)
Loading Time Is Super-Important! Here's How To Track It On REAL Users
I've mentioned how important loading time is for affiliate marketers, right?
Just to recap: it's really important.
It's really really REALLY important.
Most other variables on your landing page are chancy - if you change them, they might improve your CVR, they might lower it. But if you improve your loading time, it will almost always improve your ROI, anywhere from a little bit to an ENORMOUS bit.
The Problem With Loading Speed Testers
But the problem with loading times is that testing them - which you should be doing a lot - can be really tricky.
You can use a speed tester site, but they aren't 100% representative. They aren't running the same connection or system as your users. They aren't on 3G or WAP. And they probably don't even have the option to test from the country you're targeting!
Shockingly, the loading time results you get from a server in a datacenter in New York probably won't match the actual loading time that someone in Manilla viewing your lander on their smartphone sees.
If only there was a way to just track what loading speed they - and all the rest of your visitors - saw!
Spoilers: there is.
And it's free, and surprisingly easy to set up.
What You'll Need For This Tutorial
You don't need to be techie at all for this tutorial.
You'll need to be running a campaign with a landing page in it - doesn't matter if it's the first part of the funnel, like a pop campaign, or the second, like a display campaign.
You'll need to be able to edit the HTML code of that lander and insert a bit of Javascript. You don't need to know Javascript or anything more than the very basics of HTML.
That's it.
How To Enable Real User Monitoring For Your Current Landing Page
To start with, head to http://www.pingdom.com .
You're probably used to these guys from their website speed tester, but that's just the tip of the iceberg of what they do.
Sign up for an account, and log in, then go to Real User Monitoring:
From there, click "Add New" at the top right:
The "Add Real User Monitoring Site" window will pop up.
- For Website Domain, enter the domain you're currently using for the lander you want to test.
- Website Name can be anything you like.
- Their load time thresholds are somewhat optimistic. Drag the slider all the way to the left.
- Set the timezone - either your timezone or the timezone of the geo you're targeting, whichever you'll find easier to understand.
- Hit Create Site.
You'll see a window like this pop up:
Copy that code, and paste it into the lander you want to track just below the <head> tag in the HTML. Upload that to your server.
And you're done!
Wait for some traffic to run to the lander, come back, and you'll see you have a page like this:
And yes, that big number in the middle is the average load time that real users are experiencing! That's live data from a real test I've been running for another tutorial.
From there, using the tabs at the top right and scrolling through the various pages, you can drill down into all sorts of data:
- Using the "States" tab, you can figure out whether any slowness is from DNS, redirect times, download times, or even from render times (something that a loading time website will never tell you).
- You can break down the devices which are visiting by load time, to tell if specific platforms have a problem with the landing page.
- You can check your stats by country if you're running a lander in multiple countries.
And, of course, you can tell how fast, on average, your lander arrives - not to a datacenter that's pretending to be a visitor, but to the same people who are (or aren't) converting on your offers!
I hope that was helpful! If you've got any questions, comments, or suggestions, let me know below!
Great stuff as usual Caurmen! Webpagetest.org is another good choice for this type testing 
Originally Posted by northernlights
The other services I would recommend checking out are:
- NewRelic for those of you with a server and slightly tech inclined, but be aware on the free plan Browser RUM has a 24 h window of rentention
- Site 24x7 (a bit of a crappy interface but data is decent)
- Cedexis (if you want to contribute to global data on ISP > Server / CDN performance and have access to these stats for research)
Wow thanks Caurmen! Definitely will test this
Originally Posted by zeno
Do they offer something more than Pingdom or just a second reference test?
exactly what I needed. Thank you Caurmen. So far I was trying to make different copies work on my lps, but most likely not so many ppl waited them to appear... Speed is very important! Ty
Cool beans! Now time to figure out if my direct isp vps strategy genuinely works.
great info. thanks
Thanks, good post!!
Coool! But isn't real user monitoring actually slowing your page a bit?
@cz_voyager - good question!
I looked into it, and I'm pretty sure that the answer is "not significantly". Pingdom's RUM code loads asynchronously, meaning that it shouldn't block loading on the rest of the page. And when I ran a conventional non-RUM load time test on the page I was testing, I didn't see a difference in terms of load time with the RUM code enabled.
I'm reasonably confident that it has minimal impact, but I'd be interested to see the results of a more rigorous test, perhaps using Gatling.
Unfortunatly its not free anymore... Only 1 free month and then: https://www.pingdom.com/pricing
Awesome share. Was wondering how much connecting to a SQL database and running a couple of checks on a small database would impact page speed, I'm guessing since its s2s it should have minimal impact?
Originally Posted by caurmen
My page is averaging 2.7sec load time.. see the screenshot below.
What I'm running:
LiquidWeb VPS
Rackspace CDN
Lander:
83kb html file
Minified css/js/html
Only 3 external requests (18kb audio file, 2kb lander javascript and the js for pingdom)
Any suggestions on what I can do to get it below 1 second load time? I do see the redirect is very slow, is that the network?
Hmm, that is very slow indeed.
The redirect's unlikely to be
What geo is this?
First thing to try - pull the CDN out of the equation and run a bit of traffic again, then see what the results are.
Also, if you're running on a geo near to one of the geos they offer, try the Full Page Test on Pingdom as well - http://tools.pingdom.com/fpt/ - and see what results that gives, and if there's a specific asset or stage that's blocking.
Let us know and we can diagnose further!
....and that's why I love STM, great post Caurmen...catching up here.
@caurmen, I'm curious if you've actually tested this and see if xx ms improvement is statistically significant as far as CR goes?
Originally Posted by caurmen
Ok, so I did a test as you suggested without using the CDN and it didn't improve the redirect time.
I'm currently running in Indonesia and as far as the full page test goes, there isn't a location very close but I chose the closest one. The lander I'm running with is small and is only referencing a small audio file, everything else is in the index.html file, including all css and javascript.
Besides the CDN, are there other ideas I can explore testing or diagnose? Does it make sense to test a VPS based in Indonesia?
Great stuff! We'll get this fixed.
Can you show us the waterfall graph from the Pingdom test? That'll help track down what's going on.
Originally Posted by caurmen
The waterfall test doesn't reveal much as there is many files to analyze. Two things I will test today and report back are using a VPS in Indonesia or close to that geo, and changing the 1 image that is being used on the lander from a data url and just reference it in a directory as a src img. I've seen a mix of information regarding some browsers having a difficult time decoding data URI and that it can take longer to load than just linking to the image. This particular image is a small icon of about 3kb, not very big.
Also, one observance is that Android browsers seem to be having a much longer load time which would confirm why that browser type has a CR around 48% less than Chrome Mobile.
Ok so here's a positive update.
I found a VPS in Indonesia, currently waiting for it to be fully setup by them.. should be within 24 hours. Does it make sense to use a VPS along with a CDN or no? All my files are static html/css/js/images. I decided to specifically look for a CDN service with a presence directly in Indonesia to test the effect it has by running it straight from them. The results were great and reveal what I think is the issue behind all of this.
The new CDN definitely has had major improvements across the board as you will see below in the updated Pingdom screenshots. The ONLY area lagging is the same Android Browsers but it did improve. I feel that browser is what is skewing the overall lengthy network redirect and I'm not sure if that is common for this browser or what the issue may be.
Perhaps its my landing page code? Perhaps its just that browser and/or service usage on that browser in Indonesia?
Here is a screenshot of my best lander on the new CDN, big improvement... compare it to the last post I made:
Now, in an effort to try and improve the Android Browser load time.. I wanted to try and make my lander as W3C compliant as possible, just to test. In the header I used:
<?xml version="1.0" encoding="UTF-8"?><!DOCTYPE html PUBLIC "-//W3C//DTD XHTML Basic 1.1//EN" "http://www.w3.org/TR/xhtml-basic/xhtml-basic11.dtd">
Still have a few errors but down to 5 from 18 errors. It does seem to have had an improvement on the Android browser load times while everything else remained mostly the same. The same data is small, I sent about 200 visitors to each test so a large sample may change the results but I don't think it would do so significantly. Here are the results:
What are your thoughts?
Looks good!
It may also just be that the area you're targeting has pretty slow Internet connections on phones. Are you targeting wifi or just carriers?
With the local VPS - no, don't bother with a CDN on that. A server local to your target geo will (usually, assuming it's well-peered at the datacenter) be slightly faster than a CDN.
I'm only targeting carrier traffic so that certainly has an influence. I guess I just find it odd that android browsers is the only one so sluggish.
The CDN has been a good improvement though.
@mrpayne - that IS odd, you're right. Can you dig into your stats and find out what model of Android phone they're using?
Interested to see what results you get from the local VPS!
Originally Posted by caurmen
Home > Tracking Campaigns > Tutorials, Tips and Guides