(The First) Nine Easy Things To Forget That Can Boost Your Campaign To Profit (5)
(The First) Nine Easy Things To Forget That Can Boost Your Campaign To Profit
AM has a lot of small details. Lots of small, easy-to-forget things that can add up to some major lost profits, or even an unsuccessful campaign.
As you may know, I'm a big believer in checklists. They're proven to improve performance in everything from surgery to the practise of law.
So here's a quick checklist of the top nine or most easy-to-forget items that can cost your campaign real time or money. I've added a short explanation after each.
And stay tuned! In discussion with the rest of the STM team, we came up with even more things that it's easy to forget - so look out for Part 2 next week!
Check Your Funnel Works
If I asked from the main stage of an Affiliate World Conference how many people in the audience had never launched a campaign with a broken link, I don't think I'd see many hands go up. No matter how expert you are, it's super-easy to screw up one of the complex steps that goes into launching a campaign. (I last did that... oh, on Thursday.)
So grab your tracking link and GO THROUGH each step. I usually do that immediately after submitting the campaign for approval at the traffic source. If you find a bug, don't fix it and move on, fix it and start again - half the time you'll find you didn't actually fix it as you thought you did.
Check Your Loading Speeds
And by that I mean the loading speeds at each stage: tracker, lander (each lander you're testing), and offer. If the offer's super-slow, either consider using a prefetch/prerender link or ask your AM if there's anything they can do. If the lander's slow, debug it yourself - start by checking if any external links you've used have mysteriously gotten slow for some reason. And if your tracker's slow, either talk to your hosts or, again, debug it yourself.
You can often catch a loading time so slow it was going to cause you real problems this way.
Oh, and don't be lazy - test load times from as close to the geo you'll be running as you can. Dotcom Tools is good for that.
Check Your Load Speeds On Real Users
Loading time tools often don't tell the whole story. There's virtually no reason to not run Real User Monitoring software on your landers at least whilst you're optimising. It doesn't (or shouldn't, anyway) affect your load times.
Here's a great tutorial for RUM testing.
Analyse Javascript Clicks
If you're running a Javascript element on your lander - a quiz, a carousel, anything your users can interact with - you can dig tons of additional optimisation data out of tracking your users' interactions with that.
Here's a tutorial on tracking JS events on your lander. Particularly for quizzes, I highly recommend this - it lets you sharpshoot problems that previously you'd just have been attacking with a shotgun.
Bot-Testing
I've talked a lot about bot testing recently, so I won't go on about it more here. Suffice it to say that in 2016, if you're not bot-testing your traffic you're probably getting fleeced. So many placements are boosting their traffic with clickbots that it's getting ridiculous, and eliminating the 90% bot placements as early as possible is a great way to save money.
Here's a step-by-step guide to doing a basic bot test.
Split-Testing The Same Offer On Multiple Networks
You absolutely, absolutely need to be doing this. A lot of newer affiliates will just trust a single network and not split-test: with all due respect to my colleagues in the affiliate network business, that's a massive mistake.
Almost every experienced affiliate has seen huge - up to 100% - swings in CVR between different networks running exactly the same offer. Never just run one offer from one network unless you've scoured the globe to find other networks with the same offer and come up short. Otherwise, always split-test them against each other.
Rigorously Stick To Statistical Significance
There's two sides to statistical significance: spending too much, and cutting too early.
Cutting too early is easily fixed. Never cut without testing against a split-test calculator, unless you're an expert and have the experience to know you're seeing a pattern you've seen before.
Spending too much money is actually more common. I see quite a lot of people running split-tests - notably lander split-tests - well beyond the point where they have the information to make a decision. Using a Bayesian split-test calculator you can often make decisions startlingly early: don't do it after 1 or 2 conversions, but if you have 4 or 5, you're good to at least plug the numbers in, so do so early and often. Remember, using Bayesian testing there's no problem with repeatedly testing for significance - and running a test past the point of significance is literally spending money for no reason.
Cut No-Hoper Placements Early
Remember that the one exception to the "don't check stats too often" rule is if you've just started a high-volume campaign. You want to catch no-hoper placements as early as possible: if they've spent more than your hurdle amount (usually around 3x payout) with no conversions, kill 'em.
Don't just leave a RON or multi-placement campaign to run for a day - it's all to easy to come in, learn that most of the cash was eaten by 3 terrible, high-volume placements, and have essentially wasted most of your first day's spend.
Check Your Lander Against ALL Devices You See
There are basically two stages of laziness with landers. The most egregious one is to design a mobile lander on desktop, and then not check it on a phone at all. That's really really bad, and really easy to do if you're used to responsive design and just assume your lander will work.
Never assume a lander will just work on a phone. Test it.
The second stage is even easier to forget. Check your lander on all the common handsets you expect to see. You can do this using Google Chrome's developer mode as a quick stopgap, but you should really check using something like BrowserStack which lets you see how the lander actually renders on real devices. It's incredibly common to find out that your CTA is below the fold on half the devices you're targeting, or some Retina weirdness makes it look ugly as hell on an iPhone, or something else along those lines.
Once you have your stats in, check your devices list. If there's anything on there getting significant traffic you haven't checked already, check it now.
And that's it for Part 1! Next week in Part 2 I'll reveal another nine things that if you check as you launch, will make a measurable impact on your profits. See you then!
Another epic post! And the tips about implementing RUM and js events tracking - I'll definitely test those.
Caurmen could you please clarify what you mean by this:
Amy
Nice post Caurmen. I think all these can be automated can be saas. good idea!
Originally Posted by vortex
@vortex - yep, as francug says. Once you've "fixed" the issue, go right back to the start of your funnel testing, reset all cookies from tracking, and test it again as rigorously as the first time.
Half the time you'll discover the fix you added either didn't work, or caused another problem which you need to fix - at which point you should fix that, then go back, reset everything, test again, and so on...
Home > Affiliate Marketing Forum >