Bidding Strategies Part 1: Mapping The Landscape (34)
How much should you bid on your traffic source?
That's a question that rarely has a simple answer. And yet, bidding strategy can be absolutely key for your campaigns: the difference between positive and negative ROI, or between a $100/day campaign and a $500/day one.
Over the next couple of weeks I'm going to be covering several bidding strategies for different platforms. But we'll start with probably the most important strategy: how to scout out the landscape of your traffic source and figure out where the best ground – the best bid – is.
Traffic Sources This Guide Covers
This guide is only useful for traffic sources where there's no kind of "quality score" metric. On any platform where the price of your impressions is adjusted by the response of viewers to your ad, this guide won't work. Watch for Part 2 of this guide if you want bidding suggestions for traffic scored guides.
(Adwords and Facebook are the most obvious examples.)
It also won't work on any platform which has a winner-takes-all bidding system. If you only get traffic if you're in the top spot, it's far easier to monitor potential traffic volume, so just bid up to that top spot and then take a note of how much traffic you see! However, very few systems use a winner-takes-all system these days.
So what does that leave? Most traffic sources. Most adult sources; all DSPs (as far as I know), including platforms like Go2Mobi and Decisive; POF; and 90% of other traffic sources can be mapped out with this system.
I'm assuming a CPM traffic source here, but this works equally well for a CPC traffic source provided the CPC bids are flat (not influenced by quality score). It won't work with a 'Smart CPM' or other algorithmic bidding method.
Finally, the traffic source needs to have realtime reporting, and you need to be able to adjust bids and expect those changes to take effect more or less in real time (plus or minus a minute).
What's The Problem We're Trying To Solve?
Put simply: we want to know where the sweet spot between ad price and ad performance is.
Almost all traffic sources work approximately like this:
At very low bids, you'll get no traffic, or almost no traffic. We want to avoid that, because no traffic means no conversions.
At very high bids, you'll be put in 'first place' in the bidding, and get the maximum amount of traffic possible. Once you get above that maximum amount, you'll get no more traffic no matter how much you bid. If the maximum is $2 and you bid $10, you'll get exactly what you'd have gotten for $2. We want to avoid that too because we don't want to waste money, and that's a total waste.
In the middle, there are sensitive points where a small increase in bid results in a large increase in traffic. We want to find those points, and position ourselves just on the right side of them.
Unfortunately, most traffic sources don't publicise the curve of their graph - and it'll be different for each set of targeting options.
So we need to test.
When Should You Do This Test?
Personally, I usually use this test immediately after a campaign shows any promise, so that I can optimise my bid along with other variables.
However, you can test at any point after that in a campaign's development. If you have a mature, profit-making campaign and you haven't done a bid-mapping exercise like this, it's valuable to do one, because you may find that you can squeeze out considerable additional ROI.
Mapping The Landscape: Step By Step
You'll need a couple of hours available to scope out your bid curve for each campaign with different targeting. However, you can do other things at the same time. This time should be during the waking hours of your target geo – so not between about midnight and 9 a.m. for your target geo.
It should also be as close to the middle of the day as far as your traffic source is concerned as possible - in other words, not right after or right before daily budgets reset. Ask on STM or ask your traffic source's support team if you don't know when their budgets reset.
You'll need a timer – the simplest way is just to Google 'Timer 5min' or whatever you need. I've suggested 5 minute intervals here, but that's dependent on the size of the traffic source. For a smaller traffic source, you can go for 10 or 15 minute intervals.
You'll also need some budget available. You won't need much; you can go through this entire process with the cost of about 5,000 impressions, sometimes less.
Make sure you have a daily limit set on your campaign, just in case one of these steps goes wrong. It should be reasonably high but not so high that if the campaign suddenly spends up to that limit you're in trouble.
Part 1 - Finding The Top Of The Curve
1) Start with a very high bid – $5 cpm is my usual opener, as very few traffic sources will go higher than that. On pop sources, you can start around $0.15 per pop. Watch your campaign like a hawk for 5 minutes. With broad targeting, traffic can come in very suddenly indeed, so you need to be prepared to act fast if you suddenly get a thousand impressions or more in the first minute. As soon as you hit 500 impressions, temporarily pause the campaign. If you don't hit 500 impressions in 5 minutes, note down how many impressions you got in those 5 minutes and pause the campaign. (For pop sources, reduce these numbers considerably - aim for about 40 pops at this level.)
2) Change your bid to half your previous bid and repeat. Again, pause when you hit 500 impressions, or note down how many impressions you had at the end of 5 minutes and pause then. Also note down the bid that caused that many impressions.
3) If you had approximately the same number of impressions per minute for 1) and 2), halve your bid again and repeat the test. Keep doing this until you get noticeably less impressions for one of your tests. Note down bid amounts against each impression amount each time.
4) Once you have found a bid at which the traffic reduces, enter a bid half-way between that bid and the previous bid. So, for example, if you find that the traffic reduces at $1.25, bid $1.88 (half-way between $1.25 and $2.50). Repeat the test.
5) If the result of 4) was approximately the same volume of traffic as in 1), then drop your bid down to half-way between your bid in 4) and your final bid in 3). If it was noticeably lower, then increase your bid to half-way between your bid in 4) and the next highest amount you've bid already.
6) If the result of 5) is lower than the result in 1), then the next highest bid that you've tried is the approximate bid 'ceiling' for this targeting. If the result of 5) is the same as or very similar to the result of 1), then the bid you used for 5) is your approximate bid 'ceiling'.
Example
1) I start with a bid of $5, and get 250 impressions in 5 minutes. I record that.
2) I repeat with a bid of $2.5. I get 256 impressions in 5 minutes.
3) That's approximately the same as 1), so I record that and the bid and halve again to $1.25. I get 124 impressions this time. I record that, alongside the bid.
4) Since the number of impressions was lower, I now choose a bid half-way between $1.25 and $2.50 - $1.88 - and test again. This time I get 176 impressions.
5) That's still lower than 1), where I got 250 impressions, so I increase to half-way between $1.88 and the next highest bid I've used, $2.50. That gives me a bid of $2.19. I test at that bid, and get 239 impressions.
6) 239 impressions is pretty close to 250 impressions, so I note that $2.19 is my approximate bid ceiling. I know that if I bid any higher than that, I'll not get any more impressions for my money.
Part 2 – Mapping The Curve
This part is much more straightforward.
1) Adjust your bid to 80% of your ceiling bid. Restart the campaign and run it for 5 minutes, noting how many impressions you get against this bid level. Pause the campaign.
2) Adjust your bid to 60% of your ceiling bid. Test again as above.
3) Repeat for 40% and 20% of your ceiling bid.
4) Find the two points between which the number of impressions changes the most. Adjust your bid to exactly half-way between those, and test again. Note the result.
5) Plot all of those points on a graph in Excel. You now have a good map of the available impressions vs. cost per impression for this campaign.
You can now use this knowledge to choose your best bidding point, whether that's the cheapest bid at which you'll get actionable amounts of data or the best balance between price and volume.
Notes On This Approach
Bidding systems are super-complicated, because they're a live competition between advertisers. So you should be aware of these things when you use this method:
- This testing system does not address quality. On a lot of traffic sources, quality of impressions goes up as bids increase. If you're not sure if your traffic source is one of those, ask on STM. That doesn't make this test useless – particularly because the ceiling bid is likely to also be the quality ceiling – but you should bear it in mind when picking your bid, and try testing higher bids to see if CVR increases. We'll go into more detail on testing for quality in Part 2.
- The bid landscape will change over time. The bidding landscape here is entirely determined by what other people are willing to pay. As such, it changes over time. On most traffic sources it won't change fast, because hundreds of people are contributing to the landscape and most of them are likely to change their bids rarely if at all, but you should periodically re-test.
- Bid volume varies over the day as advertisers run out of budget. That'll be MUCH more pronounced on some platforms than others, but it's a thing all over. We'll go into more detail on this in Part 2 of this guide as it's a big topic - for now, just be aware of that, and as mentioned above, don't do your test either at the very start or the very end of the bidding day.
And that's it! In the next part of this series, I'll look at competitive bids and strategies for different types of traffic. In the meantime, if you're confused, you have a question, or you think that this approach could be improved, let me know in the comments! I know that the approach outlined here can be a bit confusing at first, so don't be shy about asking if you're stuck!
Waiting for the next part. Thank you.
Really appreciate this guide Caurmen! Not too much out there on bidding strategies/optimization.
Any thoughts on whether it's constructive to dupe campaigns at the 80%, 60%, etc levels and run for a day, instead of changing bids inside the 1 campaign?
I usually dupe at different bids, but not in such a systemized fashion (until now!). The goal is to get more reliable data (including traffic quality), but there's always the thought in the back of my mind that I'm invalidating my tests because I'm cannibalizing or competing with myself...
Just what I was looking for.
Thanks for the strategy, Caurmen!
Caurmen, thank you for this. Great guide and I can't wait to see part two.
Great, glad it was helpful, everyone!
@barton - look for Part 2 to expand on that! There are definitely reasons to do that test, specifically traffic quality, as you mention.
Hi Caurmen,
Thanks for the great write up to this guide! So I was trying to test this method during peak times on ZeroPark and noticed that this method may not be viable for ZeroPark.
I did the $5 CPM test in the initial test and received 81 redirects and then the 2nd test I did with $2.5 CPM I received 27 redirects. With this method, I blew away $33 dollars even when setting a campaign budget to $5 for each test. The surge of traffic I received being the highest bidder(IMO). Before zeropark even gets a chance to pause my campaign when I already went past budget there are already several placements sneaked in and spent $5-10 dollars.
Ah, interesting - thanks for bringing that to my attention.
The key issue here is it looks like you've bid $5 per pop rather than bidding CPM, which would be $5 per thousand pops (or $0.005 per pop)!
The recommended numbers I've given here don't quite work for pop traffic, as $5 CPM is too low. I'm going to consult with the STM brain-trust to get a solid number for max bids on pops - watch this space!
Quick update - we'd recommend starting with a $0.15 per pop bid as a maximum, and rapidly reducing that. Probably even more rapidly than recommended in the guide, given the cost is higher.
Thanks Caurmen, I'll give that a test tonight during peak time in Australia and will post back the results later tonight. 
@caurmen - This is quality material. Really appreciate your guides, I've gone through most of them and many of the other contributors' guides over the last 3 weeks. It's brought me up to speed rapidly.
@mrpayne - cool, glad they're helpful!
So, even while up at 2am last night I decided to go ahead and test my campaigns bid ceiling since it was peak times in my geo.
Here is my screenshot and I've also included a Dropbox link for those who want to download the excel file I made and just input their bids/impressions to see their own graph.
Oddly, I seem to have had a spike in traffic at bid points .007 and .019 which makes the graph look a little odd but everything was time to be exactly 5 minutes even though it was a pop traffic source. It definitely helps me visualize the bid range I want to test to balance traffic/costs.
Dropbox link for bid/impressions excel file -- > Download Here
I was also up late last night as well trying to test the bids at peak times. I chose the peak times at around 6-8PM in Australia.
Things I've Noticed:
- Between the highest bid of .15 and second lowest bid .02 there isn't much difference in terms of the amount of redirects I receive. But, also the amount of spend for 0.02 is significantly lower. Also, with the bid at .15 & .0264 the amount of redirects received are not that far apart from each other.
- In Part 2, I noticed that bidding lower may actually yield more impressions than bidding higher.
- Picking a sweet spot for bids in my opinion is around the range of .013 - 0.02 and mostly likely I'll choose a value in between those two. So on going forward my bids at this point would be .17.
Hi guys. I added a variable {bid} to my tracking link (not the one that considers the costs, other var for itself).
Now I am very easy to play with prices, draw graphics and seek the best price for each site from source.
@simon_89
Originally Posted by simon_89
Originally Posted by il3n1n
Now I am very easy to play with prices, draw graphics and seek the best price for each site from source.
Originally Posted by mrpayne
I do not understand why
Do you need to follow the peak times of your GEO or the country where the ad network is in? Peak time of the Geo right?
And
@spartanen - See Part 2 for more discussion of timezones. In short, no - just make sure you're testing during the time that people are awake 
Hi @Caurmen... I saw part 2 but i think i did to much reading so i overlooked it... just saw it...
offtopic... love that i am almost ready to start "cooking" after al this absorbing of information for about 2 months! Only thing that i read all kinds of new interesting stuff all the time so i wait a bit more... almost time to not making excuses and gooooo!!! 
@spartanen - Always, ALWAYS start on a big project before you feel ready (unless you only get one shot, which isn't the case in AM). You'll learn so much faster by doing than by reading alone!
Is popads a ad network where i can use this strategy?
@spartanen - yes, but see my comments about pop pricing and appropriate bids above.
Is there just 1 curve or are there more curves.... for example from 0.001 till 0.007 and from 0.007 till 0.012 etc...?
Hi @Caurmen are you sure Popads works like this?
Because when you go into your campaign and go to summary you can see "We estimate you'll receive XXXXXX impression(s) per day" and if you change you bid that will change also and then it looks like its just a straight line from low to high when you bid from low to high.
I dont want to doubt you offcourse so probably you can explain.
I know spartanen has pivoted to other things, but for those who were wondering the same, I wouldn't 100% trust popads' numbers from the campaign summary tab.
At best, they have been a broad ballpark number.
Newb checking in. You do this high/low strategy with EACH URL/KW you are bidding on? For example, if you were bidding on 5 URLs you would change the bid for each URL after each received 500 impressions?
Hello caurmen, thanks for awesome post. But let's say that at $1 CPM and $3 CPM bid I get same amount of traffic, but are they of equal quality? Maybe traffic I get for $3 CPM will convert better?
@bnvltd - absolutely, that's something you should test. Sometimes the traffic will be better quality, sometimes it'll be exactly the same - depends on the precise bid, geo, targeting and source.
Originally Posted by bnvltd
Originally Posted by caurmen
Do you know if this would work for Native sources like Taboola?
Assuming you can adjust your bids as rapidly as suggested here, there's no reason it won't work for native.
Sir, this knowledge is awsome. Thank you for helping our newbie
Home > Affiliate Marketing Forum >