Home > Paid Traffic Sources > Native

---

Thoughts and suggestions on big scale Taboola's campaigns? (10)

---

I'm running ecom ads on Taboola for the past few months now. Spent well over $1M, had many ups and downs. I'm running on SmartBid - Optimized. However I am still struggling to create successful campaigns. There are still lots of questions I don't have the answer for, although I'm testing new ideas every day. Here are a few of them:

• Blocking bad performing banners or adding new ones to an established campaign (70+ conversion pixel on the campaign, running for more than 3-4 days) seems to do more bad than good to the campaign. It usually changes performance quite a bit and almost always kills the campaign. Why could this happen and is there any way to prevent it from happening?
• Changing Site bid %s on an established campaign doesn't do any good either when using Smartbid. If I have a placement that has 20 sales and I increase CPC % by 10-20%, very often it goes into negative on the next day. Why is this happening?
• New campaign performance is usually very random. Sometimes it takes off sometimes it doesn't. Do I have to start multiple clones of the same campaign and select the best peroformer?
• How do I know if the banner has a good CTR and will it keep that number? If I get a certain CTR on a banner on an established campaign (xx,xxx spend on banner + running for multiple days) and I either clone the campaign or upload the exact same banner to the same campaign, the CTR and performance of the copied banner will be totally different (even though nothing else changed, it's just how the algo optimized it). Is it worth using Even vs Optimized to get a rough estimate of the banner CTR?
• Do I have to start with higher CPCs and lower them down by roughly 10% daily to get the best out of the campaign? Is this working?

I know a few people who are doing crazy amount of sales (x,xxx/day/offer/geo) on Taboola/OB right now, but I just can't seem to get that kind of CR/profitability on the campaigns. I suppose my CPC is too high and they're getting cheaper clicks, even though we run very similar (if not the same) creatives and LPs. I usually target 1 GEOs but have had similar results when running globally. Starting with higher CPC and decreasing it every 1-2 days. I cut the sites that are not performing after 4-5x payout and cut the obivous shit ones. I cut the shit banners, but that usually has a bad effect on the campaign. My campaigns are usually spending quite a bit (over $xx,xxx/day) but they are not profitable/close to break even. I feel like I'm still missing something on the campaign setup/management side of things. My ads seem to be pretty good and my LP is usually better than what others have.

What am I missing here? Is there anything you would recommend to run a successful big scale campaign on Taboola?

Any help would be massively appreciated. Anyone else who's spending mid $x,xxx to $xx,xxx/day on TB/OB and interested in sharing ideas/tricks/thoughts, just send me a PM.

---

Do you completely delete the ads or pause it? I have a feeling that if you delete the ads it might effect the performance of the campaign.

I've read that for your sites that are performing well, it's good to actually put a fixed bid above your actual CPC of that site.

Yea that bothers me too running Smart Bid - Optimized. I hate it when all the traffic goes to one banner without giving other banners the chance to perform. The Taboola rep that I spoke to told me that when you're running smart bid - optimized never have more than 10 ads in the campaign. I didn't ask him at the time whether he meant 10 ads actively running or more than 10 ads(paused or unpaused in the campaign).

I start off with a pretty high bid too at 0.75 for Desktop traffic. I believe the reasoning behind this is so that at the start of your campaign you can pull in the most premium placements and over time you can lower the bid by 10%. @jack_l is there more to this?

Perhaps maybe changing the bid too often can cause the campaign to perform worse? Changing the bid probably every 3-4 days or once a week would be more ideal I would assume. In terms of cutting sites, I think the new approach to take is significantly reducing the bid on those campaigns instead of cutting them - like if a site bids 0.75, I'll reduce it by 70% instead of cutting it. By significantly reducing the bid, at least you won't be spending too much on that placement because the bid is way too low. Cutting/Blocking the sites might mess with the algorithm.

All these methodologies I've mentioned are still in the testing phase. I'm not an expert by any means or launched 100s of campaigns, but I've read some stuff here and there about certain strategies to combat what you're going through and I really hope this helps you out or gives you a new perspective.

---

Hey man -

We haven't done quite as much volume as it sounds like you're doing, but we have been consistently profitable this year (peace be upon the internet marketing gods) and have run a very diverse set of verticals. We do higher payout lead-gen stuff but we also do search arb, which is nice because it provides insanely large amounts of data since the cpa on those camps is in the $0.05 - $0.50 range typically. It also allows us run in lots of smaller, less competitive geo's, which is nice.

We don't really 'rev the engine' on Taboola with a high bid to start. A little high, yes, just to make sure we're getting good clicks, but it's not like Outbrain where it's hard to get traction at all if you don't start super high and slowly reduce.

We do 100% SmartBid, auto-optimized ads, etc. We never even look at what sites the conversions are coming from. I used to try to optimize the sites individually when we started, but eventually took @thedudeabides advice and stopped and it was the best thing we ever did. My understanding now is that the smartbid is literally bidding on the individual person (ip address/email address) on the other side of the screen, and taking into account like 30 different variables (age, gender, location, time of day, income, past conversion history, etc, etc, etc), so the actual site they are clicking from isn't the hugely important thing that it is on something like Revcontent where that's the only variable we can really optimize on.

I agree 100% adding ads/deleting ads/etc can tank a campaign's performance really quick... it's a Catch-22 though because ads burn out after awhile too so if you're not constantly trying new ads you lose the longevity factor...

If we're running a smaller volume campaign (like low to mid xxx/day) we just continually upload new ads in it.

Once we get something up to xxxx/day though that's when we'll usually have multiple versions of the campaign going at once, that way we can continually test new ads/optimizations/etc without it massively affecting the overall performance, as it would if there was just one single camp instead of a bunch of them. And I don't believe we've ever made it to xxxxx/day without doing so via an assortment of camps rather than one big one.

But yeah, if you're doing millions of dollars of spend running ecom and haven't lost your shirt I'd say that's quite an accomplishment in and of itself! I have trouble getting most ecom offers profitable even at xxx/day in tier 2 and 3 geo's... So yeah take my advice above with a big grain of salt! lol

But yeah, feel free to shoot me a dm would be happy to compare notes as things continue into Q4

---

So when you run campaigns Jack, do you normally blacklist/reduce bids on sites that don't yield any results? So let's say you run a new campaign with 6 ads and after running 1-2 days, only one of your ads is getting all the impressions do you assume that's the best creative and not test the others thoroughly?

When you're cloning the successful campaigns, I'm assuming the other campaigns you're either splitting them between mobile/tablet - the optimization part. And another campaign would be used to test creatives, once you find 2-3 ad creatives, you'll dump them back into the old campaign to see how well those ad creatives stack up against the winning ad from your initial campaign. Do you also use the same testing variables - smart bid, optimized, and same bid?

---

We've spent a lot on Taboola as well and see the same thing.

Basically if you touch anything, it will fuck up the campaign.
Even if you don't touch anything, one day you can have 300% ROI, the next day 10% ROI.
It's random as fuck, there's no pattern.

The only way to stay sane on Taboola is to look at things at a weekly timeframe vs daily. If we're profitable in a week time frame, good if not we need new creatives.

It's sad how they hate if you optimize things, but that's how it is. They just want to take all your money by making sure you buy crap too not just the best performers for you.

---

So we don't typically do any optimizations at all to individual sites. The SmartBid itself will block sites now. If you get a campaign up to say 10k spend and go look at the sites you'll see some have been blocked by the SmartBid itself, and usually by that point none of the sites in the top 30 will be 'horrible' ones because the smartbid would have quit sending clicks to those.

Again though, my understanding is that the smartbid is bidding on the person themself, over like 20 variables, not just the site. So if I'm running a Donald Trump commemorative coin ecom product, the smartbid is going to eventually focus on Conservative people with medium to high incomes with a history of converting on ecom products (to use a made up example). In that case it doesn't matter that much whether they are on site a or site b, what matters is the above-mentioned factors. I'm sure the site still matters to some extent - since some sites do have more 'fat finger clicks' and the like, but that's just an example of how the site is only one of a ton of variables, unlike on Revcontent for instance where it's the only variable or main variable.

If it was profitable I'd keep it in the rotation and then try to find new ads that could beat it... if it's not profitable I'd take it out If there was another ad that had a good conversion rate but much lower ctr, then maybe I'd switch to that one but raise the overall bid I don't think there's any hard and fast rules, just trying to find creatives that work basically. Ideally you'll end up with 4-6 good creatives that are all profitable (although that happens only rarely of course lol).

So usually with cloning camps or running multiple iterations of something, it's just kind of random. Like if the first camp does well and is scaling upward then maybe I'll create a clone with slightly different hour-parting... then another clone without a different mix of devices... then another clone where I focus on really high-ctr, clickbaity ads... etc... and then yes if I find an awesome creative in one I'll usually try to add it to the others then yes

But yeah - take all of the above with a grain of salt - it's entirely possible any success we've had this year has just been due to stumbling upon a few really good offers or something like that - I know of other people doing way more volume/profits on Taboola than us doing things 100% differently... I think there's probably lots of ways to make it work...

---

When you take the ad out, do you completely delete it or pause it? I'm doing a test now on smart bid - even allocation and I'll be done testing the creatives tomorrow. Once I see 3-4 creatives that either have a conversion or the highest CTR - I'll switch the campaign over to Optimized to let it compete with those 3-4 creatives to see which one comes out on top. Basically right now, I'm just trying to debug what makes or breaks a campaign.

Yea, there's definitely various ways to skin a cat here, just trying to develop the best approach to find campaigns that work. I'm launching ecommerce campaigns but they all seem to be a miss. All the offers I'm running, they're either yielding 1-3 conversions a day and I honestly don't know if that's normal. But, I also think that makes sense since the budget I have on it daily is $100. I definitely appreciate the insight you've provided here and gives me something to refer back to when running campaigns.

---

Actually Taboola and many other native ad networks have changed a lot recently. All of them have introduced some sort of algorithm and that for sure will definitely influence the performance of a campaign at some extent, so better not stick with old fashioned optimization strategies.

When it comes to Taboola, I'm under the impression that they are somehow trying to get to something similar to FB when it comes to their SmartBid or optimization algorithm. However, that should take some time and considering the ad format differences is not going to happen any time soon.

Ads burnout is quite slow on native, so you won't necessarily need to refresh ads every few days. Plus Taboola uses the ads as the main drive on their learning phase and blocking an ad that has enough conversions to influence their algo, will certainly reset the learning algorithm - especially with 70+ conversions. Better focus on sites.

Whenever you're changing site bids with SmartBid on, you're practically telling Taboola that you are willing to pay more. So when you have a converting site, it will increase the upper limit of the bid trying to get the conversion on potential sites too.

Here's an example: Let's say site msn-uk-politics running on SmartBid with a bid level at around $0.35. Having SmartBid on, it will allow the algorithm to play with the bid (if I'm not wrong about the percentiles) from -50% to +100%. So practically your bid will vary from $0.175 to $0.7. Now let's say this site is profitable and you're increasing the bid by 30%. In this case the new bid range will be from ~$0.228 to ~$0.91.

Generally speaking on Taboola new camps should start on a high bid, then slowly drop it. If the targeting/settings of the campaign are the same on multiple clones, then odds are that these campaigns may compete with each other in terms of bids. If you have more than one account, I would try running clones from different accounts or at least sub-accounts.

When you have creative traffic allocation set to optimized, Taboola will focus major part of the traffic on a single creative and keep it as the champion for as long as the campaign runs. It may send a few clicks to the other ads, but nothing significant to turn change the champion. Then when adding a new ad, the winner will be the same, it will just send some traffic to the new ad to see if it can win over the existing champion.

Setting the creative traffic allocation to even, will give you more control over the traffic the ads receive. But, even in this case, it's worth noting that it would be good to avoid blocking an ad that has a good portion of the conversions. If done so, the learning phase of a campaign may be reset.

Ideally yes, this way your ads will get tested on premium sites rather than on 2nd or 3rd tier sites of Taboola's inventory.

Outbrain too has changed a lot recently, so I'd say the best approach would be to go for semi-automatic optimization keeping device targeting separate while getting rid of low quality section as soon as possible. Then after a certain amount of spend, it would be a good idea to ask their support to increase the Publisher and Section limit to the max. It would be pretty hard to get a campaign profitable with the default 30 pubs and 100 section blocking limit, so better ask for a rise of the limits as soon as you reach the blocking point.


I'm not 100% sure, but these fluctuations could happen even due to the new campaigns being launched with high bids. It's just a thought tho, but wouldn't be surprised if for some reason the newly launched high bid campaigns will be competitive enough to cause such fluctuations.

This could be a good approach too. Practically you can use the avg cpc of a site over the last few days and set it to fixed bid for specific sites.

When only one of the ads keeps most of the traffic, it means that Taboola has decided which ad to use as a champion. In general a high ctr ad should yield a lower CPC while a low ctr one, higher CPC, but yet again, here we don't have any insights of the criteria used by TBL internally to determine the winner. However, I'm under the impression that besides the ctr it takes in consideration even min publisher RPM, margin, competition, promoted product category, etc.

Last but not least, I wouldn't ignore the importance of the funnel. Quite often, some small changes on the landing page and offer page can turn the results upside down.

---

I’m still relatively new to Taboola (only been running a month, just rounded $1k/day spend), but I’ve had some experience and feedback so I’ll try and weigh in.

What I’ve been told, and have used as a principle in my campaigns, is to treat vCPM as the main metric. Low vCPM ads seem to lose impressions to others in their campaign anyways, and once they fall down to under 5% or so of daily impressions it seems safe to pause them without issue. Adding a new widget does seem to throw things off a bit, as Taboola immediately sets the campaign back to Learning mode and allocates a very large percentage of the impressions to the new ad(s) for a few days. If the performance on these isn’t as good or better to the previous ones, it seems to kill the campaign


[QUOTE] I’ve noticed the exact same thing, and it makes no sense. Raising a site bid either causes my campaigns to spend more for the same volume of traffic from the site, or oddly, in some cases causes it to do worse inexplicably. I guess in situations where you’re seeing the avg CPC for a site way above your campaign avg it might work, as an indicator your bid on that particular site is too low, but in general I agree with you and I’ve avoided raising bids for specific sites even when profitable and only doing so by the campaign level bid.


In terms of bad sites, what my account manager advised me to ad, and I’ve done thus far without killing campaigns, is to decided how to handle the site based on the site vCPM vs campaign vCPM. If the vCPM is much lower, and especially if the site isn’t at the top in terms of volume, it seems safe to block entirely. If the site has a high vCPM, I was warned that blocking it could kill the campaign, and its best to reduce the smartbid for it by 10-70%, which will still give some traffic from the bad site, but less and cheaper. In my limited practice so far this method has worked.

[QUOTE] Hard to tell. Banners seem to have their performance vary wildly in the first 3-7 days before settling in on the winner(s). Several people have told me to avoid Even banner distribution at all costs, and that Taboola were planning to remove that option entirely in the near future


[QUOTE] From what I have seen so far Taboola Smartbid seems to always spend 90%+ of your bid (may not be the case if you’re bidding extremely high), so bidding high to ensure conversions then dropping slowly looks to be the best approach. While I don’t have enough evidence to be certain how best to approach bidding, I can share what happens when your bid/vCPM is too low. Both for new and existing campaigns, a low bid will work… initially. However, after a day or two, it will do a few things. First, it will send the actual CPC price way above your smartbid and seems to trigger the campaign going into Expedited Learning. A day after the overbidding, the campaign spend drops and the quality of sites sending traffic reduces too, effectively killing the campaign.

---

Thank you everyone for sharing your thoughts on this. It's very interesting to see that others are struggling with similar problems and it's not as straightforward as it should be (if they want us to spend, they should make it easy for us to spend). Some people have their unique approaches and I guess everyone's doing it differently.

As far as I see here you have to come up with your own strategy that works for your special case. What I gathered from the comments for example is that it seems like blocking the top 3-4 banners (traffic wise) is not a good idea. Also not a good idea to increase the CPC of a certain site. Optimized allocation should be used more often than Even testing. I've also learned that not changing the campaign after launching it is often much better than trying to optimize it manually. Like others said Taboola's trying to become FB in terms of auto-optimizations. It must take the algo much more time to learn about our campaigns though, so we should probably leave it keep learning for a bit more than we would think.

I'll keep sharing my observations in this thread and please do add your own ideas, so this thread can become a great place to learn for everyone who's running on this ever changing native network.

I appreciate everyone's thoughts, thanks for spending the time to share!

EDIT:

I have created a Native mastermind group for people who are actively running Native traffic and would like to share thoughts. Add me on skype and I'll add you to the group: live:.cid.c805ad289d0a0c4a

---

Home > Paid Traffic Sources > Native