Home > General > Affiliate Marketing Forum

---

Optimization process & when and how to optimize? (6)

---

Hey guys & gals


I was wondering if anyone can give some case study, article links where I can learn more about optimization.

when it comes to optimization process itself:

• How many visits should each element (such as carriers, LPs, Browsers, Browser Version etc) before starting increasing bid on each or excluding?
• I've came a cross this situation many times. Day 2 I could see (for instance Carrier Vodafone) has performed best and O2 performed worst. So I excluded O2 and increased Vodafone BUT day 3, Vodafone performed worst of anyone else. So my question is when start/what needs/after how many visits or clicks I should exclude.
• Or another situation was after day 4 I could see Vodafone & O2 had conversions and good amount of clicks so I did a "bid-bucket" and tok them to its own campaign (with same blacklist & whitelist) BUT then they stopped converting. So my question is when/after how many visits I should start doing "bid-bucket"?

Appreciate any help

---

Hey Navidaffiliate,

first, have you read this tutorial/explanation by Vortex?: Optimizing a Direct Linked Pop Campaign EXPLANATION READING at the end of that post, there are many other links with posts regarding optimization.

1: This depends on many factors, in my opinion it depends a lot on your budget, and how much the payout is. Usually a offer with a low payout converts much quicker than a offer with a high payout, such as a mobile content or sweepstake offer compared to a nutra trial offer. This makes the payout important, because the lower the payout, the more data you will gather for optimizing.
Let me give you another example: (given that these offers are working) if you are running an offer with a payout of 36$ and you spend 30$ on one placement without any conversion, then it would still not be enough data to cut out that placement. However, if you are running a offer with a payout of $0.50 and a placement has spend 30$ without any conversion, then you must cut out that placement.
2: This happens a lot when you don't have enough statistical significant data: for example if Vodafone has only made a couple conversions, there are many other factor besides the mobile carrier.
3: I usually don't do a whitelist campaign, i feel a lot of times the traffic dies out when you don't have a big enough whitelist. I like to do blacklists, and slowly keep adding non-profitable placements to the blacklist.

I would suggest also to focus on one variable at a time, instead of trying to cut out landers, carriers, domains, etc.. at the same time.
Also, try to make 2 campaigns, 1 targeting only android and another only IOS.

Hope this helps!

---

Thanks a lot man. Appreciate all your help

I have one another question regarding the whole process. Lets say I have a sweepstake SOI offer paying out $3. How much should the testing budget be and How much each placement should use before I start excuding them (I like the theory of blacklisting instead of whitelisting)?


Btw I looked at your platform Optifyme.comit look good. Anything it can help a newbie with or you think I'd need it?

---

Hello

Here is another simple strategy to follow in terms of optimization : https://stmforum.com/forum/showthrea...-of-my-methods

As for the budget, depends on what you want to find out. In order to find out which offer from a batch is the best or shows most promise, you need to spend at least 2x offer payout for each of them, to see if any of them starts to fall behind. After 3x payout spent on each of them, you should be able to exclude the worst ones and pick those with promise.

When cutting placements, you should give each of them at least 2x payout to be sure they do not perform well with the current funnel. The problem here is that many small placements are not able to send such an amount of traffic in a reasonable time frame, so you will have to focus on the bigger ones first and keep the small ones running and decide on them later on.

I'm also a fan of the blacklisting approach, it usually works better for me.

---

You should be able to figure it out by reading the following posts:

https://stmforum.com/forum/showthrea...l=1#post336229

https://stmforum.com/forum/showthrea...844#post334844

It would be hard to determine a test budget, because you'll need to first find an offer that converts decently, then use it to test landers to find a winner, and then use that to test more offers. There are WAY too many factors involved to even suggest an approximate budget.


Optifyme has provided some good answers to your questions. Below are mine:

I'm afraid you can't judge based on visits - it would mainly depend on the cost you're paying for traffic vs. the revenue from the conversions (i.e. the offer payout and conversion rate).

It's important to focus on the big traffic segments only - for example for OS that would be android and ios and maybe windows phones - and ignore the smaller segments. With pop traffic it's mostly about being efficient - i.e. getting quick results without spending too much - because pop camps don't tend to stay profitable for long. Mind the 80/20 rule.

And only break out a segment in a separate campaign (i.e. a profitable OS/browser) if it's responsible for a lot of traffic. Otherwise, traffic may decrease drastically. Also, it will take time for you to manage an extra campaign - so make sure the profits are worth that kind of trouble.

Regarding carriers, I would suggest to target one carrier per campaign. Makes it easier to optimize bids.

I see that you've listed LP with targeting options. It's important to distinguish between targeting options and elements you need to split-test. Targeting options are traffic segments like OSs, browsers, carriers, devices, etc. - for those, you would want to target as many as are profitable for you. Elements you need to split-test are mainly offers and landers - stuff that each visitor will only see ONE of usually (in other words, you can only send each visitor to one lander and one offer), so you would need to split-test to find out which offer and lander are the best, and send visitors to those.


Again - you would want to target as many segments as are profitable or at least break-even for you, and NOT split-test them by cutting the worst and keeping the best.

Also - there will always be volatility in performance. Campaigns will perform differently from day to day, hour to hour etc. This is exactly why we need to look at stats from the same time-frame when split-testing anything. For example we can't run only offer 1 yesterday, then run offer 2 today, and compare their stats. We'd need to run both offer at the same time, and compare them using stats calculators (please see links in my signature on how to cut various things).

I would also recommend to cut the segments with the most-negative ROI FIRST, then wait to see how the rest of the segments will react before cutting further. For example, if you cut out IOS and it's ROI is in the toilet compared to Android, many/most placements' ROI will likely increase. And when you cut some big placements with very negative ROI, the ROI of other segments like OSs and browsers should go up.

It's hard to provide rules of thumb on when to cut a segment such as OS/browsers. Start by cutting the worst placements first as mentioned above. If you have specific stats to show please do - it's always easier to comment on specific stats than formulate general rules.


I definitely agree with optifyme - blacklisting usually works better than whitelisting when it comes to pop. I talked about it here (although I was talking about placements specifically, the same can sometimes happen to other traffic segments as well):

https://stmforum.com/forum/showthrea...l=1#post335598

You can set up your own rules-of-thumb for blacklisting/whitelisting. The larger portion of the total traffic the particular traffic segment is responsible for, the longer you should wait until you decide whether to whitelist/blacklist it. Having said that, do keep in mind that pop camps don't usually last very long - so don't wait too long before making a decision.

Again, if you have stats, it would help to show them so we can dissect them.




Amy

---

You are welcome!

If you are running a $3 SOI Sweep, I would suggest splitting between a couple other similar sweeps, then you can spend on each offer about 3x the payout, as mentioned by Matuloo, then you see which of these offers converted best.
By any chance, are you using landing pages? Then you would also have to split test between the landers..

Then you can start optimizing based on placements..
For each offer I usually like to spend about 10-15X total payout before deciding if an offer is worth pursuing. After spending 10-15x payout I look on getting at least -50% ROI, so for a 3$ sweepstake after spending $30-45 I would like to have at least 5 conversions, or $15 revenue. Of course when I do this, I make sure that the type of landers are working for other people, by doing some manual spying or even paid spying (which I don't usually use much anymore).

Aside from this, I believe that the method shared by Matuloo's to be very complete!

Thank you! Well it is a tool to help affiliates automate their optimization process. by connecting to your tracker and traffic source Optifyme automatically blacklists placements based on your rules. For example you can create a rule that blocks a placement every time it spends 3x payout without any conversion, like this: If Spend > $9 & Revenue <0 then block the placement.
You would usually use a tool like this when you already have a working offer, and don't want to continuously checking your data to find placements that are losing you money. We can also notify you through email or Telegram messenger in real time of your tracker statistics.

---

Home > General > Affiliate Marketing Forum