Home > Paid Traffic Sources >

---

TikTok - Should I Continue With This Offer? (51)

---

This pertains to TikTok. I'm running a personal loan lead gen offer on Revshare basis. The offer has different payout stages dependent on the credit score the user has when they sign up for the offer.

Bad Credit - ~$1.xx per lead
Medium Credit - ~$7.xx per lead
Good Credit - ~$12.xx per lead

The ad groups has exited the learning phase(50+ Conversions) and I started to scale, but increasing the budget by 30%. But I feel clueless as to whether or not I should continue with this offer. The CTR is good above hovering around 2-3%. CPA is around the $3 range. I'm thinking that I don't understand what metrics to really look at when determining to continue to pursue a offer. This offer has varying payouts so it's much harder to decide whether to continue or not.

Most people that sign up for this offer is within the bad credit range so most of the time I'm getting paid out around $1.xx per lead. For the first 1-3 days I was profitable, but yesterday didn't pan out as well(not like terribly bad). I guess what I'm trying to get at here is understanding whether or not to keep pursuing this offer.

By looking into the stats, I found one interesting thing that showed Apple phones took most of the spend while generating a lower amount of conversions, but Samsung generated the most amount of conversions while the lower amount of spend. Although this data seems to be conflicting when I look at the third party tracking software(Binom) I use. It shows iPhone with the most amount of conversions. So I don't know which source to trust here. I'm leaning more towards TikTok data but don't want to make the wrong decision here.

Binom



TikTok Stats





​

---

Your issue is the bad credit leads. Can you set Binom not to report those back to TikTok? I suspect it really isn't considering conversion value and is simply optimizing for number of conversions.

You may not have a $3 CPA after that, but it still may be low enough to get profitable. Plus you'll still get paid on the bad credit leads, you simply won't be reporting them as conversions to TikTok.

---

I think it is sending conversion values back to TikTok with the "value per submit form shown". From what you're saying, is TikTok possibly delivering impressions at a cost of around my average conversion value? Although, you suspect it isn't the case. If TikTok is delivering impressions based on conversion value, then ideally the bad leads would be good, if I were sending them the high value leads then my CPAs would increase. This is quite interesting you brought this up as I never thought about it that way. I would love CPAs to be low, but then again that means I'm getting a majority of the bad credit leads which isn't attractive.

I think I can not report back the conversion value back to TikTok from Binom by adjusting the postback url link. https://functions.funnelflux.pro/tiktok-s2s-api?ttclid={externalid}&access_token={t7}&event_na me={t8}&currency=USD&revenue={payout}&pixel_id={t6 }

Getting rid of the &revenue={payout} token.

---

No, what I'm speculating is that TikTok doesn't care about the conversion value when optimizing. It only cares if it converted or not. If so, removing revenue won't change things. You need to simply not report conversions on bad credit leads, don't tell TikTok they converted. Then TikTok will only see medium and good credit leads and optimize towards them.

So see if there is a way you can set Binom not to report any conversion below a certain value, and then test that and see what happens.

---

But, TikTok can't differentiate which leads are good and which ones are bad. The only way I know they're good or bad leads is based off the payout value.

And just to be clear here.

The payout varies by whoever applies for the loan.

Bad Credit Score - 669 or below - $1.xx per lead
Medium Credit Score - above 669 to 739 - $7.xx per lead
Good Credit Score - 739+ - $12.xx per lead

---

I feel like you're reading the answer you want to hear, not what I'm actually saying.

Can you tell Binom to not report any conversion with a value below $3?

The question I'm asking and the solution I'm suggesting all take place inside Binom.

---

@iwanttofly That's an interesting thought! So basically what you're saying is that when bad credit leads convert, just DON'T report that to Tiktok, so the algo would stop targeting people with bad credit (or at least similar audience pools).

That's an intriguing possibility - I can see that happening for FB. Don't know how sophisticated Tiktok's algo is, but worth a test!

@stungads I'm still a Tiktok newbie myself - but several things you can try:

1)Test another offer. There are so many personal loans offers - wouldn't hurt to test more to potentially find one that converts even better! Now would be the best time since you have creatives that are still converting. There's a revshare offer on Madrivo - loans for people with bad credit - worth a test!

2)Test more creatives. As you must know, creatives go stale pretty quickly on Tiktok.

3)Target iOS and Android separately. That way the data would be more clear.

Hope that helps!



Amy

---

**Prior to typing out my thoughts, this is what I keep on thinking**

After re-reading again and again, I kinda get what @iwanttofly is saying. What I'm trying to get at here though is how does TikTok know if it's a bad credit lead. Does the offer send back certain data(audience) to TikTok? Scratch this line, I just had a epiphany and hope it's right.

**This is me typing out what's going on in my brain now**

TikTok shows conversions coming in. The majority of those conversions are from an audience that has bad credit history. Since TikTok is seeing a lot of conversions coming in from the ***bad credit*** group it'll keep serving my ads to this group.

So your suggestion is to get Binom to not report back any conversions to TikTok that are $3 below because that's bad credit audience. Rather, it should feed back conversion data to TikTok with conversions within the medium and good credit audience. So that TikTok will continue to serve my ads more tailored to that audience.

As I'm typing this out right now, I got a holy shit moment in my head. Please tell me I'm understanding this correctly now lol.

I will have to chat with Binom to see if that's possible to not report under $3 conversions to TikTok. And because I'm using FunnelFlux's API for postback I'm not sure if it's possible.

---

Hey @vortex, and here I thought FB would take conversion value and ROAS into account. Interesting...

So @stungads, if we assume that TikTok users have a similar credit score distribution as the regular population, or at least similar to Facebook's users, and we assume that nothing untoward is going on with the offer, then it leads me to a few thoughts.

1. TikTok's AI doesn't care about conversion value or ROAS. It may show up in the dashboard, but the AI doesn't care. That's your problem, it just finds people that you say are valuable because you marked them as a conversion.

2. While the AI does consider conversion value and ROAS, it is so heavily weighted towards CVR that it will keep sending you people with bad credit because they convert.

3. Your creatives are highly geared towards bad credit individuals and so the AI just keep sending them because otherwise you wouldn't get any conversions.

I have a strong suspicion that a lot of it is the 3rd, but it is definitely worth testing to find out if 1 or 2 are also having an impact.

So again, check inside Binom and see if there is a way you can set a floor for reporting conversions. See if there is a way you can tell Binom that if Conversion Value < $3, don't report to traffic source. Please ask if that is unclear.

---

You got it. TikTok doesn't "know" that someone has bad credit. Well, they may buy that data or infer it, but it really doesn't matter. What TikTok knows is that these group of people, bad credit, have a lot of commonality. And the more members in the group, the better the AI is able to define their commonality. So if TikTok seems that group as valuable to you, it is going to focus more and more on that group.

So I'm saying, stop telling TikTok about people who are in that group. Only tell it about people in the groups you care about. Now of course, that assumes that the AI doesn't care about Conversion Value or ROAS. But it is definitely worth some testing to find out. Because knowing that has tons of value when it comes to eCommerce or variable payout offers.

---

@stungads Another thing to experiment with is the angles you're using in your ads.

For example if you're using the "Need a loan? Bad credit OK!" angle, then you better make sure the offer you're running knows how to make the most profits from people with bad credit, in order to pay affiliates well.

Oh I wouldn't know for sure! I don't think I've ever promoted a variable payout offer on FB. I only made that comment based on speculation and my knowledge on how FB works. The FB algo does take conversion value and RoAS into consideration for sure, but no algo is perfect, and manual optimization can still potentially get better results. I'd be interested in seeing some actual test results - too bad this type of loan offers can't be run on FB.



Amy

---

And I bet this is a lot of it, even if not so blatant. The creative is just attracting people with bad credit. Which honestly are the people who are most interested in short term loans.

---

then you better make sure the offer you're running knows how to make the most profits from people with bad credit

So, understanding the flow of the offer? Once a user signs up, look at how many upsells there are and therefore understand that the offer owner should be able to pay me more than the standard payout rate for a lead?

I have now taken a more streamline approach to creating videos. I'm not creating 8 different videos at a time. Instead I'm creating 4 videos and all of them are totally unique in some way. And whichever one crushes, I have several opening hooks to switch out. So the idea of bad credit angle, I have thought about filming specific opening hooks with that kinda verbiage i.e. How with my 599/650/750/etc.... credit score was able to be approved for a $x,xxx.xx loan.

Does any tech expert know how to filter out leads like iwanttofly suggested? I spoke with Binom chat about trying to get conversions under $3 not postback to TikTok and they suggested that I will need to write my own script and upload it to my server in order to do that.

---

Or - just test more offers and let the data talk.

I don't have a solid understanding of the loans industry just yet. But based on what I know about leads, the selling price is dependent on what buyers are willing to pay.

I believe how these loans lead gen offers work - and I may be wrong - is that each website works with a slew of lenders that can provide these loans. And each lead is matched up with lenders that can provide loans for that particular lead based on the information on their form (i.e. amount borrowed, credit score, location, amount of risk based on whether they have a job/assets, etc.)

So - I'm speculating that certain loan offers may be connected with more lenders that would be willing to pay good money to acquire bad credit leads. And chances are those offers would pay us more.

Testing more offers in general is something we should be doing anyway. No matter what happens in the back end, all that really matters to us is the ROI.


Sounds interesting! I'd be curious to see results!


@jeremie?




Amy

---

Not going lie. I probably broke the #1/2 rule of affiliate marketing. Test multiple offers. 2 reasons I haven't been doing that, 1) Creative Approval - Sometimes I have to constantly follow up with the affiliate network to see where we're at with creative approvals. 2) Testing with different networks means payments are spread through different networks and in order to get paid out sometimes you have to reach a certain threshold.

But, what I didn't test was multiple offers within the same network. Yikes.....will get cracking on that now.

Yea, I always think the leads I generate worth more than just the $1-12 range. I always think about the LTV of the customer and how it should be more. But, tbh I'm not that skilled at lead gen yet to know the ins and outs of what to expect. Once I get to that level, I do plan to re-read kjrocker's thread about going direct.

---

This tip from iwanttofly is solid, let's see if the TT algo can actually be smart enough to select higher level clients... I doubt it, but we never know until the theory is put to a test.

This is something you need to check with the affiliate network, since they are the ones who are reporting the conversion. Even the advertiser has to be able to report the various levels of leads separately for this to work. It would be the best if they could just turn the postback for the low level leads off

Or you could create something like a "dummy" postback script, which would be a script that would grab the value parameter and only launch your trackers postback link when the rate is higher than $1.

This should be very simple for any coder.

This is your current postback : https://functions.funnelflux.pro/tiktok-s2s-api?ttclid={externalid}&access_token={t7}&event_na me={t8}&currency=USD&revenue={payout}&pixel_id={t6 }

The dummy would be https://domain.com/dummy.php?ttclid={externalid}&access_token={t7}&ev ent_na me={t8}&currency=USD&revenue={payout}&pixel_id={t6 } and you would use this dummy postabck at the affiliate network.

The dummy.php would parse all the parameters from the url and as long as the payout value is higher than $1, it would then launch your real postback and fill in all the parameters that it parsed from the url.

Know what I mean?

---

I dont think TT understand conversion value because for me to rapidly get the algo to work I tend to throw false conversions at it after 5 conversions... I tend to match the 5 conversions device, location at times, browser etc to reach the recommended 50 conversions TT need. Also I started to rotate in CPC offers to get the algo to learn and those values are obivously all at random. from .05 to 9$ cpc... but it does help the algo.. the longer it reaches TT to get 50 conversions, creative burn and CTR start to drops... so not sure if is understood by TT - best way is to word the angle on the creative a bit more to target users that have good credit...

---

Yeah that's one of the major differences I've noticed between Tiktok and FB - getting out of the learning phase is so much more important for the former.

And this is why budgeting is so crucial. Basically if you don't exit the learning phase within the first week, the chances of success decrease considerably.



Amy

---

The first part looks easier I'll reach out to my affiliate manager to see if that's possible. However, if I ask them to do this wouldn't they scrape any bad credit score leads I get?

For the second part, I'd have to hire a coder who understands affiliate marketing that would create the script right? Plus the postback I'm using is currently utilizing FunnelFlux's Events API which I don't have within my own tracking domain so that's an issue there as well.

---

I get that... yea creatives burn out very quickly from what I keep hearing and being able to generate 50 conversions quickly to get past that learning phase is crucial. It took me four days to reach 50 conversions so the sooner you get to 50 the better.

I messed up when I launched my campaigns by 1) not testing a couple offers at once, 2) not adding disclaimers to the videos. I had to put a disclaimer on one of my videos and upload it to the ad group that passed it's learning phase. Probably not ideal, but I'm going to see how it pans out. The conversions are still the same range per day around 10 - 15. I have just upped my budget too around 20-30% for scaling. We'll see how this pans out in a few days.

---

Depends on the setup they have. If the advertiser is reporting the leads to the network just by sending a postback that they let's say just forward to you, then yes, disabling that wouldn't be the best idea. But maybe they have a different setup which makes it possible to still credit the lead to you, but not posting it back to your tracker.

I dont think neither of this is an issue actually. The coder just needs to understand how to parse/grab data from a url and then generate and redirect to a new url with the right parameters in it.

Same with the FF events API. This dummy postback would basically just serve as one extra redirect and it would launch the FF url exactly the same way as when used as postback in the network, as long as the payout value is high enough. The FF url "doesn't care" how and where it's launched/opened either, it just parses/grabs the tokens from the url, that's all.

---

are you just cloning the ad group with 1 creative that is doing well? to scale I tend to just clone the ad gorup with the same creative X times my budget.

---

Going back to your OG post, here's my current tracking postback url I have at my affiliate network: https://trackingdomain.com/click.php?cnv_id={subid2}&payout={payout}&cnv_stat us={status}

What I'm understanding from this is that we don't have to adjust anything in Binom, the only thing I'd need to adjust is the postback url in the affiliate network.

So the coder would write a script and put it with the php file in which I'll upload to the root folder of my domain. Once the user goes through the offer flow and actually converts it will go through my tracking link first and immediately redirect to the funnel flux domain: https://functions.funnelflux.pro/tiktok-s2s-api?ttclid={externalid}&access_token={t7}&event_na me={t8}&currency=USD&revenue={payout}&pixel_id={t6 }

Nope so once an ad group completes the learning phase I increase the budget by 20-30%. Your method looks to be the preferred method. I'm just testing this out. So far, the CTR and conversions seems to hold the same even though I threw in a new creative and paused the other one. The only difference about this new creative is that I added a disclaimer at the bottom.

---

Oh wait, let's set one thing clear so I dont create any confusion here.

So as long as I understand the setup correctly. You are using this Binom postback at the affiliate network now: https://trackingdomain.com/click.php?cnv_id={subid2}&payout={payout}&cnv_stat us={status} to pass the conversion to Binom and from there it's passed to TT via the funnelflux link, correct?

Assuming I got it right then yes, you would only change the postback link at the affiliate network. You would use the dummy "postback" there. So something like https://trackingdomain.com/dummy.php?cnv_id={subid2}&payout={payout}&cnv_stat us={status}

The dummy.php would do just one thing, parse the token values from the url and then generate the Binom postback url as long as the "payout" value is higher than 1, if not, it would do nothing. As long as the value is 2 or 3, it would just redirect to the regular Binom postback with the tokens it parsed from the url filled in.

It should be really simple coding... grab token values from url, check it payout is higher than 2, if so: use the tokens to generate the real binom postback with the values filled in and redirect to it.

Are we on the same page now?

---

I wouldn't do it that way.

I would strongly, in the strongest language I can find, urge you to leave the affiliate network postback alone. You want to know about all revenue from the network, and not just by looking at their dashboard. You also want to be able to see it in Binom. Also, that makes it easier to spot creatives that generate a lot of low payout conversions.

I would have the script sit between Binom and FunnelFlux. Call the script as the traffic source postback in Binom and let it call the FunnelFlux postback as appropriate.

---

Thats a valid point, and yes, he can use the same trick and place the dummy php between binom and the FF tiktok postback.

---

I think it's better to post all the screenshots of the links to see what exactly needs happen here. So if I'm putting a dummy php between Binom & FF. How does that work? I don't have access to FunnelFlux's servers. Now I'm just talking out of my ass

Campaign URL Link:

Offer

Full links within offer

Offer URL: Postback URL:

Affiliate Network

Full links within affiliate network box

Postback URL:


Traffic Source

Postback URL:

---

So dummy.php would be placed with the adv/network.

In there, you would need some logic to grab that payout amount, & if >x amount , exit else fire money pixel?

am I understanding this correctly?

If so, that makes sense. I am going through something similar in another vertical right now. TikTok def seems to start optimizing for those lower dollar amounts once a few come in. to the point it tanks my campaigns. the adv explained there was nothing they could do except move me to a flat-rate cpa offer which just doesn't work as well for w/e reason.

---

Let's break this down some more so everyone understands what I'm suggesting.

The problem we have is that EVERY conversion is posted back to TikTok and that's what we want to prevent from happening.

In this particular case, we want to ONLY postback the conversion to TT in case the lead is either medium or good score which means a payout of $7 or more. This value is stored in the "revenue={payout}" token.

Right now the setup is like this :

I will use @stungads screen to copy the links from.

Traffic from TT is going to the campaign tracking link: https://trackingdomain.com/click.php...&3=___EVENT___ and the clicks land on the offer.

Once a conversion happens, this postback url fires and tells Binom that a conversion has happened: https://trackingdomain.com/click.php?cnv_id={subid2}&payout={payout}&cnv_stat us={status}
At this time, the postback already contains the real values and not just the placeholders, so the payout part could look like "payout=7"

Binom parses this url and takes all the real values from there and stores it in their DB so you can see them in stats later on.

The next thing that happens is that Binom uses the traffic postback url, fills the tokens with the data it has collected from the affiliate network and launches the URL to tell TikTok a conversion has happened.
This is the url that Binom works with: https://functions.funnelflux.pro/tiktok-s2s-api?ttclid={externalid}&access_token={t7}&event_na me={t8}&currency=USD&revenue={payout}&pixel_id={t6 }
And again, Binom will replace the placeholders such as {payout} with a real value that it has now stored in the DB.

Now TT knows a conversion has happened from a particular click.

What we need to do is prevent the FF traffic postback to fire, in case the payout is lower than $7. But since Binom doesnt offer conditional traffic postback setup, we need to use the "dummy postback" trick.

This will sit between the Binom traffic postback and the FF api link.

Let's say the dummy php will sit in the root of the tracking domain so it's url will be domain.com/dummy.php

We need it to match the format of the FF link, so the full url would be: https://domain.com/dummy.php?ttclid={externalid}&access_token={t7}&ev ent_name={t8}&currency=USD&revenue={payout}&pixel_ id={t6}
And you would have to set this url as the traffic postback in Binom.

So once a conversion happens, Binom would launch the dummy postback url now, with all the tokens filled with real data again.

Now, all that the dummy script needs to do is parse the url and grab all the tokens from it. It would check if the payout token has a value of 7 or more and in case it does, it would simply generate the FF url, add all the token values and launch it. If the value is below 7, it would do nothing.

So basically, it would do the same as Binom does with the FF url template only with the condition that the payout value has to be higher than 7.

---

Here's a serverless JS option you could use with cloudflare workers.

Go to your Cloudflare account > Workers > Create a service.

Give it a name, 'postback' anything it doesn't matter, click 'Create Service'.

Click 'Quick Edit' on the service page and paste the JS below.

Now use the domain for your worker, you can see is on the service page under 'Routes', and use the URL parameters from the original postback

So your postback will now look like:

https://workername.youraccount.workers.dev/?ttclid={externalid}&access_token={t7}&event_na me={t8}&currency=USD&revenue={payout}&pixel_id={t6 }

NOTE: I haven't tested this at all but it should work.

Also, change MINIMUM_REVENUE to what the minimum you want the postback to fire on.

---

Yes I am doing @iwanttofly method, still waiting on DNS NS to fully propagate.

---

Because the AI is garbage and not nearly as smart as anyone gives it credit for.

Most of Tiktok's interests and categories are made up guesses and have little bearing on what people are actually interested in.

Which is why if you do fresh camp with interest targeting, even after narrowing it down based on a broad target camp and checking TT stats, it will usually perform like shit compared to a broad camp.

Sending clickIDs for similar users based on region and device (iOS) is still a solid strategy as it is giving the AI hints as to where it should dig deeper. When you have a broad camp targeting all 25+ M/F in USA its still has to narrow stuff down. If there are 1-2 conversions in a San Diego, CA on iOS that have good credit, there is a higher chance that there will be more, and searching there rather than continuing to test Boise, ID would be a better use of the AIs traffic and your money.

I could be wrong, of course.

---

Yup, that's what we started with. Then @ianternet mentioned he is sometimes firing fake conversions to get the algo out of the testing phase quickly, so we kinda started to discuss the option too.

But yes, the original idea was to make sure the postback only fires for the better quality leads to test whether the TT Algo can make any use of that.

---

In this case, the speed doesn't matter because we're just posting the conversion back to the source, as long as the script get's executed properly, all is fine. No real users are involved in this process, it's basically one script sending information to another one. And that goes for any postback, speed is irrelevant, it just MUST execute properly. So it's still not the best idea to use some very shitty hosting as the script might timeout and not execute. I'm not sure if bluehost is that bad, probably not.

There is some problem yes, likely at the DNS level because I can't even ping the domain or run a trace route.

---

I chatted with Bluehost/NameCheap support today and they said everything should be good to go. I've asked their support team to see if there was anything off and they said no so I proceeded with the campaign creation today.

Ok all, nonetheless I created 2 separate campaigns to test the 2 theories(Postback only 50% of the conversions & Filter out Postbacks under $3).

It will start running tomorrow.

I'll run this campaign for as long as possible to see when TikTok Algo kicks in.

From what I've observed from 2 different ad groups, one that has passed the learning phase and one fresh new ad group. There was no difference in performance and they seemed to perform identical.

---

Ok now it seems to resolve and it pings to this address: 162.240.77.136

Looking forward to the results of your tests

---

Same! I'll probably run this test for 3-4 days to get enough data. Since I'm using the best creative/offer and that I'm kinda filtering out the low payout conversions, it'll take longer to complete the learning phase for the campaign. So, I'll most likely stop the campaign before the learning phase is complete, but I should have enough data to see if any of the theories hold.

---

You can still aim to exit the learning phase in 7 days (in case you get lucky and things pan out!)

Simply increase your daily budget accordingly - you already know how many of each type of conversions you were getting on average, and the conversion rate.

Curious to see results!



Amy

---

Ok everybody! The results are in.......
Well the first thing I realized when I ran my campaign after 2 days was creating a separate pixel for each of the test(because of the below - see item 1). Or at least do a separate pixel for the postback > $3. Yay mistakes!

Anyways, my plan for next week is to create 2 separate campaigns.

1) Run the same campaign again postback > $3, but with a new pixel. After running it for a week, I would use that data to build a custom audience which will then turn into a LAL audience. And then test that. Hopefully, this will yield an audience that has a higher tier credit score.
2) Run a new LAL's campaign with people that completed the SubmitForm conversion event. Since my pixel has went through a lot of data, hopefully this will garner a audience that's most likely to convert.

I've ran this campaign for 4 days. Monday to Thursday. Started at 8AM to ~9PM. Here are the results. I think the most important metrics to look at here are CPM, CTR, and conversions.

1st Campaign: Payout > $3 will post back to TikTok

Total Conversions/Revenue

05/23/22: 22 $68.68
05/24/22: 12 $41.40
05/25/22: 19 $41.20
05/26/22: 13 $38.88

Total Revenue: $190.16
Total Cost: $310.88

I was close to breaking even on the first day. And the days after that is when performance started to plummet. What's very strange to me though is even though my CTR is going down day by day which is understandable, my CPM is decreasing as well. Usually I would assume CTR goes down and the CPM would increase. On 5/25 & 5/26 TikTok is delivering more impressions, but I'm not sure their intention behind it. Are they trying to get to find people to convert? Maybe if I ran a few more days and the pace they're delivering the impressions might imply that "okay, this isn't a good campaign since I'm delivering so much impressions and the CTR is decreasing and not much conversions, therefore we'll start to increase your CPM."

So like....maybe TikTok is seeing that I have a budget of $100 a day and that I'm not really getting much conversions so they're trying to deliver more and more impressions to see if they can get people to convert. By them delivering more impressions, it might be a undesirable audience that they're delivering impressions to.

Nonetheless, I'm going to move forward with my item 1 that I listed above. It'll be interesting to see if that has any impact.



2nd Campaign: Postback only 50% back to TikTok

05/23/22: 29 $75.64
05/24/22: 12 $47.60
05/25/22: 11 $41.60
05/26/22: 13 $37.72

Total Revenue: $202.56
Total Cost: $315.03

Kinda the same performance as campaign 1. This campaign has an inverse performance in terms of metrics. The CTR kinda fluctuated all over. Same with the CPM as well. It would increase and then slightly decrease. My thoughts on this are that even though in the beginning TikTok is delivering a considerable amount of impressions to me on the first day, they see a trend where people are scrolling up immediately after they reach my video. So the more people that skip my video the more likelihood TikTok will want to deliver fewer impressions to my campaign. The average watch time on this video is 4-5 seconds which is my opening hook.



Overall Thoughts

So I'm not getting the feeling that any of this testing made any difference to the algorithm. However, there's one thing that kinda resonated with me when I saw Ian, Scotty, Cassssshy posts. They all kinda mention or at least I think they said that, don't quote me on it lol. Your campaign has to work wonders out the gate or else it's not going to perform as well. Which brings me back to the earlier discussion on this thread. Maybe if I try Ian's method of faking the conversions on the first day of running the campaign can boost the performance of the campaign overall. @matuloo you posted about how to fake the conversions. Maybe you understand fully about what Ian is saying...

When you run your campaign, the first 5 conversions you get. You'll take those values from Binom/whatever tracking source and plug it into the FF postback link. Do it repeatedly until you reach 50 conversions. @ianternet, what do you mean by match 5 conversion devices, locations at times, browser. If you match those conversions by like the same region let's say California, wouldn't the TikTok algorithm start to deliver the majority of impressions to like California then? I like this idea, but if I have to consistently do this when launching my campaigns then this sounds like an indicator that I need to work on my creatives.

---

Thanks for the update.

So it looks like posting back just the "good" conversions didn't really help the overall performance much, if at all
Well, at least you have tried and we have one less variable to worry about.

It would be the best to hear from Ian, but let me still say what I believe is the point here.

So whatever source is doing some initial testing period, it behaves differently during this period than it does with normal campaigns.

In some cases, the source is actually sending better traffic, in some cases it's the other way around. The traffic might be limited, the Ads might show to somehow limited audience etc... so by faking the conversions, you try to escape from this testing phase and get the "real" traffic and ad impressions as soon as possible.

Yes, I believe this is what Ian is after. He posts back conversions for clicks that match the actual converters as closely as possible. I'm not sure how deep you can analyze the clicks in Binom, but the closer the selected click for fake conversion is to an actual converter, the higher chance that it will be from the same audience you want to target.

This is basically the same as any smartlink is doing. They check each click, and analyze it down to device, carrier, geo, location etc... then send it to an offer that they know (from past data) converts the best with a similar user data profile.

Hope it makes sense and hopefully it's actually in line with what Ian is doing

---

That's something I'm willing to test and see if it makes a difference. The first 5 conversions I get I'll quickly take the clickids and generate 50 conversions out of those and see if that changes anything.

From running campaigns, it **seems** like the first day your campaign is running if you're getting a bunch of conversions in the 1st half of the day your campaign will most likely do well. If not, it'll just tank in the following days. It's like TT is quickly optimizing and finding you the best audience to convert the first day but if that audience doesn't help, then TT is just going to throw you into a bad pool the following days without really giving your campaigns a chance. This is all speculation, but this is something I'd have to test as well.

I got my hopes up by trying to run this campaign as you need 1,000 SubmitForm events triggered to build a custom audience. I'm far from that at the moment. I think this is geared more towards ecommerce. Either way, this week's focus is going to be geared on testing the best variation of my creatives in terms of opening hooks and then I'll be taking the best creatives and split test between IOS/Android and then dayparting and then scaling!!! *fingers crossed*

---

Great thread -- answered some concerns I had about 0$ conversions.

Ty for doing the heavy lifting @stungads

---

Does this really work? For me currently, I am using manual bids and my accounts are not spending so for now it's difficult to get out of the learning phase. Also I tried to use the higher funnel concept and that didnt worked for me, when I got 50 clicks and then I moved to conversions, there the same learning phase started. So now I wanna move to a different strategy that actually works

- - - Updated - - -


- - - Updated - - -

---

We're waiting for the legend @ianternet to chime in

In your case, if your accounts aren't spending instead of using manual bids try lowest cost and that should probably work.

So with the idea of the higher funnel concept, I think you actually want to build a custom audience with that so it should require 1,000 VC instead and then you can create another ad set to retarget this group for the next event, maybe initiate checkout and then once you hit 1,000 initiate checkout go for Complete Purchase.

The learning phase is meant to stabilize your ad costs. Or at least that's what I learned.

---

Isha what's your audience targeting like? Broad/narrow? What's the audience size?



Amy

---

It's completely broad targeting, Amy. The audience size is 74,164,000-90,647,000. I did split testing but broad worked well for me so I left narrowing down it

---

I tried doing this. I have one account suspended with TikTok and as mentioned in my earlier post my new account isn't spending these days. I used this account prior and spent around 150 euros but didn't get any conversions in past. So I tried to copy the HIT ID of the conversions which I had in my last account and pasted it to update the conversion in TikTok but it didn't really work. I am not sure why it isn't working, is this because I haven't got any initial conversions (but it should not matter if I am forcing it from the tracker itself) or because it is the hit ids of the old conversions so indirectly duplicates?

Also, I have a question if we force this way then TikTok algo will not be having any info like who to target even if it is out of the learning phase it will not give the proper results maybe. All I wanna know is how exactly will this work?

---

I tried to explain what I believe is the reason behind this here: https://stmforum.com/forum/showthrea...l=1#post433527

To put it simply... fake conversions will end the learning/testing phase asap in order to lift any limitations that might be tied to that particular campaign phase.

But yeah, let's see what @ianternet has to say on this, once he finds these replies from us

---

OK that rules that out!

What about your creatives? Were you using 100% original creatives?

I know you were using ripped creatives before - I'm guessing that that may have had a negative impact on your account history as well.

It may be worth it to do some testing on a fresh account just to rule that out.

Also: What was the offer payout, and what was your daily budget? Did you have any idea (e.g. from your AM) what the cost per conversion is approximately? I'm trying to get an idea on whether your budget was set high enough. The algo can be pretty fickle about the learning phase - if it's obvious that your daily budget isn't high enough for the campaign to exit the learning phase in the first week, the algo can just "give up" on the campaign and stop sending traffic.



Amy

---

catching up here... sorry for the late notice... one of my FB ads is pissing me off...

Yes, @matuloo is correct, I try to find the most accurate clicks to feed it fake conversions. so if it's an android tablet with firefox then I'll try to find that exact click... I just use my gSheets template to help filter the clicks and tell me which one is a good base on the previous set of conversions. If you have 1 conversion and a low budget, I wouldn't suggest, you want a template of several conversions. there is no point to turn 50 clicks into 50 conversions. the idea is to get enough conversions within 5 days, TT says 1 week, but if you can do it within the first 3 days it is important because I found from my learning after 3 days impressions start to drop and CTR starts to deteriorate.

also, you do not want to use like yesterday click and convert it the next day... TT understands that conversion might not be quality. so check every couple of hours and do not bulk send fake conversions. At the end of the day you shouldn't probably use this strategy too much. Slowly feed it and see how the campaign reacts (*example if you are getting 1 conversion every hour, then you should do 1-2 'fake' conversion every hour), at the end of the day if your creative engagement is shit, TT will not feed it... it essentially isnt all about the conversion it is about engagement.

you can probably see that theory in my initial case study when I started on TT... but in the first 3 days, I found it pretty important, probably critical.

the idea is to get a really good creative first... ha thanks @stungads

---

Oh yes, I used my own creatives. Also, I want to tell that my account started spending today. I initially increased my budget from 50 eur to 100 eur per day and my bid also up to 12% so yeah it started spending and also I tested the lowest cost and it also spent well with a budget of 20 euros. So yeah talking about the offer it is auto insurance (which is not being displayed on spy tool) so I doubt whether it is working or not. On the spy tool, it shows that the last it was running was on the 21st of may but my AMs are telling me that it's working well with 44% CR (mixed from all sources, as they don't have particularly of TikTok).

But the offer has spent almost 7 times the payout and I've got 1 conversions so far, I am doubting whether it's dead or should I still give it a try. Also I fear because auto insurance is over-saturated these days. So I definitely can't get rid of the learning phase with this slow speed and also with Ian's method as I've not got an initial conversion too. So now I have this big question "Should I continue with this offer or not?"

Also, I am up to a new offer of personal loan and have the same issue with it, AM's are telling me that it's performing the best on TT and CR is 28-40% and it's revshare. But I am seeing its last run on the 24th of May. It's the same as @stungads said the payout varies from $3 to $25 depending on credit. So this thread is related to me too. So I am doubting whether to believe the AMs or not.

Once I had one AM who said that one of their loan offer is running so well with 20% CR and I tested it and I was in a total loss, when I asked him about this situation he said that "TBH no one has tested it and I wanted you to test" so I am fearing out. I know we have test everything but currently I am more focused to test the offers which are already converting.

---

Home > Paid Traffic Sources >