Home > Paid Traffic Sources > Native

---

Optimizing Taboola with SmartBid On (13)

---

Hi fellas,

Recently, I started running on Taboola and one thought that keeps bugging me is the approach to take for optimizing pubs with Smart Bid being on.

I noticed that Smart Bid algo tends to be sensitive fragile to even minor changes. Although, I see that some pubs did not bring me any conversions even after 50 USD spend on their placements, I am reluctant to interfer with Smart Bid algo.

Am I just being paranoid about it? Do you just apply the same optimization methods regardless if Smart Bid is on or not.

Would love to hear your thoughts.

- J

---

At some point as you scale from 3 to 4 then 5 figures a day on a campaign/offer on a big source with lots of placements, you're going to have to give up trying to micro-manage everything, namely small site placements as there are hundreds if not thousands of them spending a fraction of your total/spend per day. I know some have a different way of approaching things, running many small to mid sized campaigns, but I prefer running a few big campaigns.

Standard 'best practices' of things like "Cut after 2-3X CPA spend" start to fall apart in that case. You also have to be conscious of the monte carlo effect in regards to a large pool of outcomes/placements and not read too much into a few outliers that seem really bad or really good; that may have just been the way things landed for that day or period. But $50 is really nothing to be too concerned about for a placement unless your CPA is like $5, and even then a few of those shouldn't be making or breaking your campaign.

Better to instead focus on your overall campaign CPC and numbers and trying to dial in your CPA that way. It's completely fine to make big adjustments to it, say dropping bids drastically at night then raising in the morning, but really only after the campaign has spend a few thousand minimum and collected enough conversion points. Mainly just want to make sure you've given time for your ads to establish themselves in the algorithm.

For the big volume sites eg MSN I still make manual adjustments to the smart bid %, but I'm always looking to reduce my reliance on that and focus more on campaign CPC and let it do it's thing. If a placement has spent a lot, say $3-500 and had no conversions, my first instinct is still to reduce the smartbid % a lot before cutting it entirely, unless I know from experience running it before that it doesn't convert.

It can be a tough mindset switch thing to turn over so much extra control to smartbid but it really is the way to go. If you're pushing big ROI it's much easier to give up control to it, less so with small margins. I can't see myself ever not using it, and all the top native guys I know make heavy use of it.

I know that's kind of a really broad answer here without giving too much specifics, but those are my thoughts without hard data to comment on.

---

@thedudeabides, thanks for a thorough answer to this question. I have also heard extremely positive feedback on smart bit from big spenders, so it is definitely here to stay. If we were to go a step further, Savanah's presentation at the most recent AWA has shown that machine-learning (ML) based advertising is a big time save and after a few days overperforms even a very seasoned media buyer (note, her presentation was on FB ML algo).

I'm just interested in how advanced is Taboola's Smart Bid. Is it just a matter of spending a few k for the algo to tune itself and get going while trimming off obvious non-performers or there is still a way to tell early on if an offer is just crap and test the next one.

@jack_l, @jaybot Do you guys have any opinion on this?

J

---

Assuming you're selecting the right type of offers and have your campaign structure and bids in order (namely just bidding high with low cap to push ads to top at launch), you're going to see right away whether there's potential for the campaign to crush it often just based on ad CPMs and a few conversions.

I don't really think of smartbid as something that helps you find winners, moreso just a tool to help you scale and extract more profit with the least amount of effort, and it seems VERY good at that in my experience thus far.

---

Another member @nativeq shared the same sentiment and suggested that smartbid is more so for scaling. "My feedback is this: smartbid is better for scaling(if you are tracking correctly), fixed bid is best to start with, but only if you have an automation tool behind(and setup automated rules there). It's very easy on native to lose money in a blink of an eye. Another thing is, taboola's reporting is not always very accurate with no of conversions so optimization can be compromised also sometimes."

With that said, is smartbid not beneficial to run when running initial testing campaigns? I would think as long as you set the ad creative targeting to even distribution instead of optimized you can figure out which ad images/headlines perform the best. But then again by the time you think about changing from even to optimized I don't there's a need to do that as you would've found the image/headline combo.

---

@stungads - just split-test both and see which works better. Do one campaign on fixed bid and an identical one on SmartBid.

For what it's worth, I lost WELL into five figures on Taboola when I started, trying to make Fixed Bid work, trying to make Whitelists work (when they still let you send your rep a list of pubs to whitelist), and trying to do SmartBid but 'Override it' and do my own optimizations on top of it.

Once I quit doing that and just 'embraced the smartbid', it instantly became our number one traffic source, and I've remade the money I lost countless times over in profit.

So yeah, I vote SmartBid 100%. Pass your conversions to them and let their algo do the work. Set the ads to 'Optimized'. Their algo is analyzing tons of variables relating to the individual person clicking your ad, which you would never even have access to. It's really more like Facebook AI than it is anything else.

Selfishly I'd prefer everyone else NOT use the SmartBid, since that would be better for me, but at the same time I don't want you to have to lose all the money I did.

You may still lose lots of money up front. Its not like doing this is easy - indeed it's probably the most difficult thing I've ever done in my life (not most stressful though, just most difficult the way a tough game of chess or writing a book is difficult), but yeah, if you're worried about what method is best, just split-test them against each other

---

Resident expert jack_l chimes in, always love reading through your posts. That's my plan to split test. With Smart Bid - Optimized, how many ad creatives and LPs are you uploading to that campaign in your initial test? When Taboola finds the best ad creative, do you throw in new ads or do you optimize it at that point(increase budget)?

With the SmartBid approach, it's ideal not to mess around with the campaign as much so my plan is the following:

1) Test Offer with x # of ad creatives and x # of LP
2) Once offer is proven with a winning ad creative and LP, throw in more ads to see if I can beat my current winning ad.
3) Will keep on repeating the process of finding the winning ad until I can't beat it and then move on to the next step of testing a winning LP.
4) Once winning ad creative and winning LP is discovered I will then start to scale.

I guess my understanding of how scaling works is to take a winning ad and LP *singular* and then scale by increasing budget and bids(as long as it's below EPC).

---

Just to clarify something - are you running all your Taboola Smartbid campaigns on full RON with absolutely no whitelisting at all? That's 60K+ Taboola placements and I imagine that the vast majority of those are going to be low quality and low volume, with only a tiny proportion (less than a hundred?) being profitable

My reps talk about Premium whitelists and Performance Whitelists which can cut the 60K sites down to a few hundred or a few thousand sites...are you using these?

Smartbid sounds good in theory but spending at lower amounts at the start of a campaign (say $50-$100 per day) it seems to spend only a minority of the budget on the the sites that converted and most of the budget goes on other sites of variable quality, which is very frustrating. That's where the temptation to use black & whitelists comes in.

What is the approximate amount of overall spend or number of conversions that needs to be reached to get Smartbid to work?

---

Yeah I believe we do full RON... I don't ask the rep to upload any whitelists or anything... and we get all sorts of crazy random pubs showing up...

You are absolutely right about the 60k+ sites and many of them not good, but I think that's why SmartBid works better... with fixed bid you're getting all of those thrown at you... whereas the SmartBid is purposefully bidding higher on the good placements... it also now blocks any widgets that do badly.. in addition to just bidding less for them or not bidding on them at all...

So yeah... at the start sometimes it will seem like there are 'bad' sites, but if you look at the site data once the campaign has spent 5 figures or so, all the top 50 sites will all be pretty good, and there will be plenty of sites that the SmartBid itself blocked..

Plus, again, the SmartBid isn't just bidding based on site, it's literally taking 20+ variables like city, state, time of day, ip address, history of that user, all the demographic groups and interests, performance of different ads, etc and combining all of that to maximize the chance of conversions. It's like Facebook's pixel basically.

Lol I probably sound like I'm shilling for it for some reason but seriously, I HATED the idea of SmartBid when I started... it just seemed totally antithetical to everything that was good about natives vs facebook or google... but yeah in our experience it just works way better... whether it's a 20$ a day campaign or a 2000$ a day campaign...

I think it's very possible that a network could have a SmartBid that is NOT good though... it all depends on how good the algorithm and AI is... for instance we also use Outbrain's SmartBid equivalent... but I don't think it's as good as Taboola's... still pretty darn good, but definitely not on the same level...

---

That's the thing...the name of the game seems to be to find those top 50-100 profitable sites as quickly and as cheaply as possible.

That's why I'm keen on putting on the performance whitelist to let Smartbid do it's magic on just the top 1000-2000 sites, rather than asking Smartbid to to sift through 60K sites most of which are junk?

---

Here's some info from someone who spent $100,000 and lost around $30,000.

1) If you kill bad ads, it will fuck the campaign.
2) If you block bad placements, it will fuck the campaign.
3) If you add new ads, it will fuck the campaign.

The best method that worked for me so far is to lower campaign bid and upbid only on the top using the taboola interface. +10% etc

Sadly Taboola seems to me like they want you to buy EVERYTHING, all the CRAP. They are after all, experts at making money and making money is all they want to do. So yes, that means you must buy all the shit too else they will punish you and take your money anyway.

My winning camps I don't touch, and I don't do anything. Instead I create new camps with the winning combo and multiple campaigns. Then stop the shitty ones that arent doing well out of the gate, just like how you stop bad adsets on FB.

I am no expert on Taboola yet, I know guys personally that made million plus in profit from just 1 offer on taboola. I know 2 for sure, maybe 3. I stil did not figure out where I fuck up. I am not confident in my game on Taboola like on Facebook. On FB, i'm so sure of the way to go that i dont even second guess myself.

---

Kinda sounds like Taboola is becoming more like Facebook? A lot of what you described is exactly how FB behaves when it comes to making changes to campaigns. Weird.

I know in terms of account bans, Taboola has fully embraced the Facebook approach - banning for no reason and not accepting appeals. (And being impossible to contact and talk to a real person.)


Sent from my iPhone using Tapatalk

---

I second this. Exact same expereince about blocking stuff, adding new ads, etc. Not quite sure if others have experienced the same.

---

Home > Paid Traffic Sources > Native