When Should I Stop Running A Campaign Or An Offer? (11)
Knowing exactly when to stop running a campaign or an offer can be hard: it's something that you will get a feel for over your career as an affiliate.
Having said that, here are some guidelines for various situations you'll encounter, whilst you're developing your own instincts.
The Four Key Rules Of Cutting Anything
Rule 1: Achieve Statistical Significance. If you cut things before you have enough data to be sure they'll continue to perform as they have in the past, you're just throwing away money. Don't do that. Follow our Ads and Landers guides for statistical significance. Also, be careful how long you've run them over - if you've run a campaign for 15 minutes the data isn't significant even if you got 10,000 clicks in that time. 2 days is minimum.
Rule 2: Don't Run Losers Beyond Statistical Significance. If a campaign, an offer, a lander or an ad isn't performing and you have statistically significant data to that effect, any more money you spend is wasted. Be sure to figure out when statistical significance will occur, if you can, and check regularly enough that you're not wasting money.
Rule 3: Don't Fall In Love... With Your Campaigns. When you're running campaigns you want to be Spock, not Romeo. It's very easy to just keep trying to improve a single campaign that you're convinced is super-close to profit. Remember, angle and offer are far more important than ad and lander. If you've tried a bunch of ads and landers and the campaign still ain't cooking, kill it and move on.
Rule 4: Don't Cut Then Paste. When you cut anything, make sure that when you create its replacement, you're trying something different. If a lander didn't perform, changing its CTA to blue probably won't make the difference. If your angle was "Win Vouchers For Beans" and it achieved -90% ROI, "Win Vouchers For Heinz Beans" is unlikely to be a killer. Make BIG changes.
How To Know If An Offer's A Loser
First of all: always, always test an offer on every network you've got access to. Don't just run the offer with one network and assume it doesn't work: split-test across all of them and see which one wins.
On Web, if an offer+angle combination has received traffic equal to 4x its payout without a conversion, after you weed out any terrible placements (see How To Run A Campaign On Any Traffic Source), then cut it. On mobile, let it get to 8x payout to account for the greater range of factors in play. Note that you should treat each offer/network combination as a separate offer here - see the examples below.
If an offer didn't work with one angle, then test it again with another, very different angle. If that also fails to get a single conversion after 4x or 8x payout after placement weed-out, then you can assume the offer's a loser - at least for you and your available traffic sources, approaches and networks - and move on to a new offer.
You can also try the same offer in a different country - but generally moving to a new offer will be a better approach. Come back to the other offer in a different country later if you really want to.
Do not assume that an offer "must" work for you just because it's getting high volume on your network. That offer may be running on very traffic sources, it may be being run by an affiliate with a particular expertise in that area, or that affiliate may have a far, far higher payout that means they can afford to spend a lot more testing. When you're a very experienced affiliate, you can sometimes stick with an offer until you make it work - as a newer affiliate, you're far better moving on.
Finally, if the offer consistently gets some conversions, but nowhere near enough to get profitable, and you've tried six or more different campaigns with that offer, move on to a new offer.
Example 1: Adi has run "Platemail For Android", an offer he saw in his network's newsletter, on a single mobile campaign with the angle "Don't Let Your Whatsapp Contacts Get Stolen By A Virus". After he weeds out a couple of crappy placements that were eating his spend, he runs a further $6 worth of traffic to the offer with no conversions. The offer pays out $1.20 on a conversion. If this was a Web campaign, he would cut that campaign and angle, because it has spent more than 4x the payout ($4.80), and try again with a new angle but the same offer. However, because this is a mobile campaign, he lets it get up to 8x the payout ($9.60) before he makes any decisions on the campaign and angle.
Example 2: Laura has run the same offer, "Platemail For Android", on two mobile campaigns, and has split-tested it between two networks. On the first campaign, neither network's offer managed a single conversion. On the second campaign, she has spent $20 after removing crappy placements. Divided between the two networks' offers, she has spent $10 on each offer. That's more than $9.60, so she stops running the offer on the first network (the same network that Adi's running on) altogether. However, the second network is offering a higher payout - $1.50. She keeps the second network's version of the offer running, until it hits $12 in spend - 8x $1.50.
How To Know When To Kill A Campaign
Most of your campaigns will be losing campaigns, particularly in the beginning. If a single campaign with a single angle isn't working after a couple of rounds of optimisation, kill it and move on.
You may be competing with someone who has a 5-year direct relationship with their offer and can outspend you. You may have an angle that will just never appeal.
Keep launching new campaigns and move on.
First of all, follow the rules above about killing offers. Generally, if you reach a point where you should kill an angle + offer combination, you'll also kill the campaign, unless you're split-testing multiple very similar offers.
After that, if a campaign is at -30% ROI or more after you've completed a couple of rounds of split-testing, and you've genuinely tested a wide variety of ads and landing pages, kill it unless you're very sure that you can make it work.
Finally, if a campaign is eating your time and you can't scale it to the point where it gives you the income you're looking for, you should kill it. If you've got a campaign that's in profit $2 a day, and you've tried every way you can think of to scale it, and it's eating 30 minutes a day of your time, kill it. Likewise, if you're running another campaign that's making $2,000 a day, and you have a $20 a day campaign that's eating half your available AM time, kill the $20 a day campaign and focus on the thing that's working!
How To Know Which Offer To Choose If You're Split-Testing Several Offers
If you have a campaign that's running a few different offers or the same offer from a few networks, and all of them have passed the tests above, load up this split-test calculator. As with landing pages, tick the "include" boxes equal to the number of offers you're running.
Now enter the number of clicks (from a landing page to the offer, across all landing pages in the campaign) each offer has recieved as "trials", and the number of conversions each offer has generated as "successes". Hit "Calculate".
If your offers all have the same or very similar payouts, then you can choose which one to run the same way as you chose landing pages, above.
Otherwise, it's spreadsheet time! Don't worry, this calculation's pretty easy. You're just going to calculate the earning per click from the lander (EPCL) of each offer to figure out what the best one to run is.
For each offer, take the minimum and maximum possible conversion rates (as listed on the calculator) and multiply them by the payout, then multiply by 0.01. That gives you the offer's minimum and maximum EPCL.
So, if an offer has a payout of $5, minimum possible conversion rate of 5.2% and maximum possible conversion rate of 8.6%, its minimum EPCL is 5.2 x 5 x 0.01 = $0.26, and its maximum EPCL is 8.6 x 5 x 0.01 = $0.43.
Now, compare all the offers. If any of them have a maximum EPCL below another offer's minimum EPCL, stop testing that offer. If they've had 10 or so conversions or more each and the maximum EPCL of an offer is very close to another offer's minimum EPCL, stop running that offer.
If they have each run for at least 40 conversions and their maximum and minimum EPCLs are very similar, ask your AMs when you can get a quality review on each offer. Pick the offer you're closest to getting a quality review (and hence payout bump) on, and run that offer exclusively until you get to enough conversions for a quality review - then ask for a review and a payout bump.
If the above is true AND your AMs tell you there's no room for a bump on any of them, run the offer with the network that you're running the most other offers with, to build your relationship with them.
Question about the split test calculator.
Suppose i have 2 LPs and 2 separate offers.
Would I be entering 4 (i.e. 2x2 permuations) trials along with their respective conversions? Or do we "group' and enter two trials w/aggregate conversion total?
@taewoo - Good question!
Usually I treat the LP and the offer as seperate stages, and so group them - both landers for each offer, both offers for each lander.
If you wanted to be more rigorous, however, there's nothing wrong with taking a multivariate approach and doing the 2x2 strategy. It'll highlight any interactions between individual landers and offers, but at the cost of requiring more data.
Experience suggests that 90% of the time, given similar offers, there are no strong interaction effects, which is why I recommend the "treat landers and offers separately and cut both at the same time" approach, but either works.
Originally Posted by caurmen
Originally Posted by caurmen
@caurmen
If for example, I am split testing 10 offers from different networks.
Same vertical. Pin submits. Payouts are $2.40 each. Pop traffic.
Each offer has received 6k impressions
Overall 60k impressions over 1 day.
Offer A - 0 lead
Offer B - 1 lead
Offer C - 1 lead
Offer D - 0 lead
Offer E - 0 lead
Offer F - 2 leads
Offer G - 1 lead
Offer H - 0 lead
Offer I - 1 lead
Offer J - 0 lead
Should I:
a) Cut all offers with 0 leads and split test the offers that converted.
b) Run all offers again for another day to get more significant data?
@whtang: add up the clicks OUT from each lander to the offer. So if LP 1 had 3 clicks to the offer, LP2 had 14, and LP3 had 12, the total would be 29 clicks. Does that make sense?
As for best timing: there's no really solid answer on this, but make sure you have a decent amount of data. I'd go by payout and test after about 4x payout to each offer.
@ibong92 - Impossible to say without more data. How much did the impressions cost you? If 6k impressions cost you $3, run for longer. If those 6k impressions cost you $300, kill the entire campaign 
'On Web, if an offer+angle combination has received traffic equal to 4x its payout without a conversion, after you weed out any terrible placements (see How To Run A Campaign On Any Traffic Source), then cut it. On mobile, let it get to 8x payout to account for the greater range of factors in play. Note that you should treat each offer/network combination as a separate offer here - see the examples below.'
Is this referring to the amount of cost based on visitors seeing the offer or the cost I'm spending overall? Well, I guess it must tbe first, otherwise how else can I calculate the cost.
Originally Posted by jaydenuk
In case of multiple offers, inconsistently performing LP can initially screw up the data, that's why its important to let each funnel combination run long enough for the numbers to balance out.
To explain better what I had in mind: In many cases, the first 100 hits can result in 5 LP clicks, then next 100 clicks can produce 3 and the next 100 can do just 1 click to the actual offer.
Under certain circumstances, these 9 clicks from those 300 hits can be distributed unevenly between 2 offers : 7 vs 2 for example. But 1000 hits later, the numbers will be more or less even - that's what statistical significance is about from a large part and that's why "you need to give it some time"
I have not understand this:
If they have each run for at least 40 conversions and their maximum and minimum EPCLs are very similar, ask your AMs when you can get a quality review on each offer. Pick the offer you're closest to getting a quality review (and hence payout bump) on, and run that offer exclusively until you get to enough conversions for a quality review - then ask for a review and a payout bump.
Originally Posted by Phoneix
https://stmforum.com/forum/showthrea...211#post289211
Just use the calculator to compare offers, cut losers until you have a winner.
Amy
I think there are general Thumb rules:
1) For high CPA offers, Like Nutra of CC submit - You have to give it time, But don't go overboard. The deposits are quite difficult to achieve and it might take longer. These are the perfect example of offers that you have to throw some money at them just to see what sticks, but $500 should pretty much give you the general direction of the offer.
2) Fester flows - SOI, Lead Gen and so on - One day and a budget of even a 100$ gives you sense if this offer works on the traffic source your testing. Low payout SOI offers are usually quite easy, If you got complete Zero - Then there's either a tech issue along the road or it's just not converting.
But the main suggestion I can give is, Be Patient and make sure you look into the traffic source prior to committing a large sum. Talk to an account manager and inquire about what works.
Home > >