Outbrain Tutorial for Native Noobs - Part 4 (5)
***************
How to Pause Ads Manually
The way Outbrain's algo works when it comes to ads, is that it will automatically assign more traffic to ads that have higher CTRs. (Remember how the winning of auctions depend on CTR and CPC? Ads with higher CTR will win more auctions, CPC being equal.) Although the algo is effective in identifying high-CTR ads, we can help to improve efficiency and save on budget by cutting ads manually to optimize for revenue instead of CTR.
This is what this part of the tutorial is about!
To see ad stats in Outbrain, click on "Campaigns" in the left menu > "By Campaign" tab > "By Content".
First thing I want to point out:
The columns in lighter blue are up-to-the-minute, whereas the ones in darker blue are delayed by up to 2-4 hours. Just something to keep in mind when reading campaign stats.
Here's an ad-cutting approach you can start with, and tweak to your liking:
1)If you're testing lots of ads in the beginning (like I did) and need to cut them down to the recommended 10-15, wait until some of the high-CTR ads have clearly emerged (hard to define which point this would be - say 20-30k+ impressions or so for the ad that has received the most impressions?) Then, gradually cut ads that have gotten at least 5-6 clicks (so the CTR wouldn't be too inaccurate) and have the worst CTR of all the ads.
Cutting based on CTR instead of CR or ROI is obviously not ideal. But if you're wanting to test a lot of ads without forking up the cash, then the inaccuracy may be a worthy sacrifice.
2)Once you're down to the recommended 10-15 ads, check to see whether at least one of the ads is profitable for at least some of the sections. To check this, either a)drill down two levels into your tracker stats (break down by ads and then further by sections), or b)in the Outbrain interface just change the date filter to show stats from AFTER you've cut down to your best ad(s).
If none of the ads is profitable for at least a few sections, it would be a waste of time to continue. As I've explained before (while talking about one of the rules in theOptimizer), it's probably because your ads and/or landing page and/or offers are not good enough.
Doing this check at this point in time is important, because if your ads+lander+offers are not good enough, you wouldn't want to continue running the campaign and spending thousands of dollars on cutting placements just to end up with a handful of placements making small profits that will never make that money back.
You may ask at this point: How much traffic do we need to run, and how many profitable sections do we need to see, for us to decide to keep the campaign running as-is and focus on cutting placements and tweaking bids? Unfortunately, I don't have a definite answer. I haven't run enough campaigns on Outbrain to provide any rules of thumb.
But the more data you have, the more you can be sure whether your best ad is "cutting it" or not (no pun intended!) And if you don't think you have enough data to make this decision, run traffic for longer - meanwhile, you can still continue to cut ads as described below.
3)Next, you may want to cut more ads using this statistically-sound method.
I would recommend using the bayesian method to cut ads that have at least a 90% chance of being inferior to the "current-best" ad. I've described the method in step-by-step detail in this post.
When using the bayesian calculator, use the number of impressions (number of times an ad is shown) as "Trials", and the number of conversions as "Successes".
Please note that this method will only cut ads that are INFERIOR to the top-performing ad. The evaluation does not take into account of revenue or ROI at all.
Optional: 4)Once you've cut down to only the best-performing ads, if you want to test that rule in theOptimizer that sets the content bid to 70% (or whatever percentage you like) of the EPC, now may be a good time. (Again, only if you're not using Outbrain's conversion optimization algo, i.e. only if you're not using "Semi Automatic" or "Fully Automatic" optimization. Otherwise, just let Outbrain's algo do its job without interference from theOptimizer.)
IMPORTANT: Sometimes, it may be good to keep ads that have higher CTR but lower ROI than your best-ROI ad, because the higher-CTR ads can win more auctions and get traffic from the more competitive placements - which the lower-CTR best-ROI ad may not be able to get. So, if the higher-CTR lower-ROI ad isn't "wasting" too much money compared to the best-ROI ad, it may be good to keep it running, just to continue getting a sample of the best traffic.
Once you've cut down to the best one or more ads, focus on evaluating and cutting placements towards a positive campaign ROI - as described below.
***************
How to Pause/Reactivate Publishers and Sections Manually
To see placement stats in Outbrain, click on "Campaigns" in the left menu > "By Publisher" tab > and click on either "By Publisher" or "By Section", whichever you want to see.
To pause/exclude a publisher or section, simply toggle the "Exclude" to the right:
In the screenshot, CNN has been excluded, but Gray TV is still active.
In this part of the tutorial, I want to suggest some criteria you can use to decide which publishers and/or sections to exclude.
1)From the beginning of the campaign, you can use the four theOptimizer rules for cutting publishers and sections (Rules 1-4). These are like safety guards that can prevent any "runaway" publishers or sections from sucking up your budget. Some of the publishers and sections can be reactivated later on, but by cutting aggressively in the beginning, we can potentially get better ROI faster.
Of course, the other option would be to just allow Outbrain's auto-optimization algo (semi/fully automatic) to take care of things, without using theOptimizer rules.
OR, you can use a mix of both. It all depends on how aggressively you want to do the cutting.
2)I also want to offer another way of cutting placements, that is statistically-sound. Please see below.
How to Cut Placements Using Statistics
It's the same method I've described in this post. If you have further questions after reading the instructions below, that post can provide further insight.
Firstly, please download the "Kill/Whitelist Calculator" excel spreadsheet from here.
Open the calculator spreadsheet and go to the tab "Kill-Whitelist CPC".
Next, we need to download stats from Outbrain for pasting into this calculator.
In Outbrain bring up either Publisher stats or Section stats (whichever you're wanting to analyze and potentially exclude). For best results, set the date filter to only show stats from AFTER the best ads have been cut. For example, if I started running on April 13th, and cut ads down to the last 1-2 best performers by the start-of-day on April 14th, then I would set the date filter to only show traffic from April 14th onwards.
Doing so will ensure that you're treating the publishers/sections fairly. A placement can only perform as well as the ads allow them to. If you include stats from subpar ads, you'd be cutting placements that may become profitable by running only your best remaining ads on them.
Go to "Reports" > "Export Reports":
Open the Outbrain stats spreadsheet you've just exported. These are the headings you'll see.
You'll need to copy and paste the following columns of data into the calculator spreadsheet, as follows.
-Outbrain spreadsheet's "Section Name" or "Publisher Name" column (depending on which stats you chose to download) into Calculator spreadsheet's "Ad/Placement Name (Optional)" column.
-Outbrain spreadsheet's "Amount Spent" column into Calculator spreadsheet's "Cost ($)" column.
-Outbrain spreadsheet's "Impressions" column into Calculator spreadsheet's "Impressions" column.
-Outbrain spreadsheet's "Total Conversions (view & click)" column into Calculator spreadsheet's "Conversions" column.
Then, copy the excel formula from any of the cells in the "VERDICT" column and copy all the way down to the last row of data.
As for the values at the top of the Calculator spreadsheet:
-Confidence: Can leave that at 80%. If you want to cut more aggressively but with less certainly/accuracy, you can decrease that to 70%, but I wouldn't recommend going any lower. Setting this higher than 80% is also possible, but will require collecting more data to arrive at a verdict.
-Payout: This would be the revenue per conversion. If you're promoting several offers - in a listicle for example - you can calculate the average conversion value by dividing your total revenue by the total number of conversions, and putting that value in this field.
-Min. ROI: This would be the minimum ROI you want to achieve for your campaign. Typically this would be at least 10-30%.
At this point you should look through the values in the "VERDICT" column. Entries that say "KILL!" means that the placement would be highly unlikely to achieve the minimum ROI you've specified (with the exception I'm about to point out). If you're lucky you'll see the "WHITELIST!" verdict, but a placement would need to have outstanding performance in order to receive that verdict. "KEEP RUNNING" means there's insufficient data for either verdict to be reached.
Needless to say, the idea is to pause the placements that are marked "KILL!" But before you do so, consider the following:
1)The calculator makes predictions based on the assumption that the same conditions would persist. So if you were to change the section bid for example, there is a slight chance you could still make that section/publisher profitable.
For example, if you've spent $50 on a section and made $25 back, in theory by decreasing your bid by more than half, you may be able to make the section profitable.
The reminder here is that by decreasing your bid, the traffic level from that section will drop and possibly cease. But I'll leave that to your experimentation.
Note that if you had chosen the automatic conversion optimization option (semi/fully automatic), then Outbrain would already be tweaking your section CPCs. So, if you wish to change the bid manually, you'd need to override Outbrain's algo.
Here's how to change a section's bid: In Outbrain, display section stats, find the section you want to change the bid for, click on the sandwich icon in the "CPC Adjustment" column, click the plus/minus sign to increase/decrease the bid until you reach the desired target CPC, checkmark the "Override automatic system adjustment applied by Conversion Bid Strategy" statement, and click "Save".
2)If you're analyzing publisher stats using this calculator, be aware that it would still possible for one or more of the sections under that publisher to become profitable if you'd run more traffic to them. If you're not sure, it may be good to keep that placement running to collect more data before taking another look.
In addition to checking the "KILL!" entries, you may also want to look for placements (publishers/sections) that have been paused out by theOptimizer, to see what the "VERDICT" is. If the verdict says "KEEP RUNNING", you may decide to run them further to collect more data. At this point, you may want to turn off one or more rules in theOptimizer (for example the more aggressive Rules 1 & 2) and go to Outbrain to toggle some of those publishers/sections back to active to run more traffic. Alternatively, you can leave them paused until later on when you have a profitable campaign and ready to scale, at which time you may decide to re-test all the publishers and sections that have gotten paused prematurely.
***************
Other Optimization Opportunities
"Pseudo-Whitelisting"
Outbrain actually does not offer the option to whitelist placements. However, Bryan taught me what to do in order to achieve the same thing (at least similar) - so I'm imparting this wisdom to you.
To get traffic only from certain placements: Change the campaign bid to a very low amount (I think the lowest allowed by Outbrain is $0.10 CPC), then set custom bids for placements you're wanting to whitelist to reasonable amounts.
Drilling into Other Stats
If you go to Outbrain's stats and click on the "By Platform" dropdown:
You can see which platforms, OSs and browsers are converting the best/worst, and include/exclude from campaign targeting, by going to the "By Campaign" tab in Outbrain, selecting your campaign, and clicking on "Edit":
You may also want to go into tracker stats, which will allow you to drill down to up to 3 levels deep. There is the standard data that
(V6 is "Referrer section ID" and V7 is "Referrer section name" - these got cut off. For the complete list of tokens, you can always go into Voluum's traffic source details for Outbrain.)
You can play around to see what additional insights you can gain by drilling up to 3 levels into various stats. For example, you can drill down first by "V4: Ad ID", then by "V7: Referrer section name", then expand the stats for your top-performing ad, to see whether the section has promise of becoming profitable. (ie. If not even your best ad can make a section profitable, then you should either pause the section, or test more ads.)
One important thing to point out is that Voluum can't track by ad image - it only tracker by Ad title and Ad ID. If all your ads have different ad titles, you can just drill into "Ad title" in Voluum. But what if you have multiple ads with the same title but different images?
I've confirmed with Bryan that there's no way to tell from Outbrain's interface, the Ad ID of each ad. So what we'd have to do to make the correlation, is display stats by "Ad ID" in Voluum, then look at the ad state in Outbrain by going to the "By Content" tab, and comparing these values to match up ads in Outbrain with Ad IDs in Voluum:
-Outbrain ad's "Clicks" and Voluum's "Visits".
-Outbrain ad's "Conversions" (Click) and Voluum's "Conversions".
Excluding Non-Traditional Publishers
Bryan told me that Outbrain has different types of inventory. There's "traditional" inventory like CNN, NY Post and Time. However, they now have additional, non-traditional inventory such as in-app and push placements. The quality and user-behavior from these placements are different. So, make sure to optimize your ads accordingly. This could mean bidding down on them because the traffic is typically cheaper, or blocking them out altogether if the conversion rates aren't up to par with the rest of the publishers within a campaign.
I would recommend to just block these out at the start to conserve budget, and retest them for scaling after your campaign becomes profitable.
However, there's no known way to block these out ourselves. But what you can do is click on each publisher's hyperlink to check whether it's an in-app or push placement:
And if it is, then toggle to pause it.
***************
Additional Suggestions
Here are some additional ideas I have on how to improve on this campaign (that haven't already been covered).
-It's always a good idea to split-test landing pages from the start. I didn't do so due to my campaign being a very small test, but you are strongly encouraged to do so!
-If possible and relevant, try to incorporate some loan offers in your listicle - lots of people can use some temporary financial help to get them through the covid situation. If I had decided to run this campaign further, I would definitely test this.
-Instead of starting by finding out which offers were hot from affiliate networks and friends, I could also have done research on Adplexity to see what other people were promoting, made a list of offers and then tested those.
-On hindsight: Since most of the insurance campaigns on Adplexity Native were pushing single offers, perhaps I shouldn't have insisted on testing listicles. But this is something you could test!
-Once you've blacklisted some of the biggest publisher and sections in the initial few days, it may be good to test more ads/landers/offers or even completely different verticals on that whitelist, to see which ad(s)+lander+offer(s) perform the best, before spending further money on cutting placements.
***************
Great Native Threads on STM
A lot of wisdom can be found in some of the threads in the Native subforum. The following in particular are must-reads:
https://stmforum.com/forum/showthrea...26%23129297%3B
https://stmforum.com/forum/showthrea...26%23127940%3B
https://stmforum.com/forum/showthrea...ith-Voluum-DSP
https://stmforum.com/forum/showthrea...rt-with-native
https://stmforum.com/forum/showthrea...e-optimization
***************
Outro
And that's all I have to say about that!
As usual, this tutorial has turned out to be a lot longer than I had originally planned - thanks so much for bearing with my long-windedness as always!
I hope that this small experiment has sparked your interest in giving Outbrain's traffic a try. And if any of you veterans in native traffic are reading this - please don't be shy - any tips you can share on how to make "my" process better would be very much welcomed!
Thanks for reading!
Amy
Huge thanks for taking the time to create this great tutorial Amy! Really appreciate how thorough you were explaining everything.
Hey @vortex -
Gotta love the senior listicles amirite?!?! They are just so fun to do!
Anyway.... fantastic follow-along!
What were the results though?
Did you get a decent number of conversions?
I would add that we've been using the 'Conversion-Optimization' auto-bidding feature of Outbrain lately and it is working very well. Seems to take several hundred conversions to work decent, but once you get to that point it's pretty darn good.
Originally Posted by jack_l
As for the results: I decided not to publish them because I made so many mistakes that readers of this tutorial can now avoid, I don't want people to look at the negative ROI and conclude that native doesn't work or Outbrain's traffic is bad (when Outbrain has quality traffic, being one of the two biggest native networks out there). I WILL say though that results were promising - with the $1000 I cut down to a good ad and enough placements to start a whitelist campaign. And there are so many placements that got paused out early that I can reactivate to run some more.
(Actually Outbrain doesn't allow whitelists - but there's a way to do it as suggested by Bryan - which reminds me - I'm going to add this to the tutorial now. It's simple: Just set a low campaign bid in the campaign settings, then once those placements appear in the dashboard, set a reasonable bid for those specific placements.)
Plus there are so many other optimization and scaling opportunities I haven't taken advantage of yet, including testing more ads, more landers, more offers, scaling to more placements (retesting prematurely-paused placements plus Outbrain's Extended Network), adjusting placement bids, increasing the bid after most of the unprofitable placements have been paused to trigger higher-demand (potentially better-quality) placements, etc. etc. etc.
I'm seriously considering expanding into native, but we'll have to see, because I've started more projects than I could finish. But with a promising campaign on my hands, it would be hard not to continue optimizing. We shall see though.
Amy
Original post above edited to add "Pseudo-Whitelisting" section.
Amy
Home > >