Thread: Day 36-37: Testing Bids

**************************************************

ACTION & EXPLANATION



Why Test Bids?

First of all, I want to explain the effect that the campaign bid has on a campaign.

Every pop traffic source buys traffic from many websites, which are referred to as "publishers" or "ad placement sites", or simply as "placements". ProperllerAds calls them "zones" and Zeropark calls them "targets". They're all basically the same thing: Websites that will be showing your ad to their visitors.

Now, every traffic source has a different algorithm that determines how much they will charge us for traffic, and how much traffic to send to our campaign. But in a majority of cases, the higher you bid, the more traffic you will receive. This is because traffic sources will try to maximize their revenue, and so will sell their traffic to the highest/higher bidders.

Like many newbies would in the beginning, you may be thinking that if a campaign is in loss, then you should automatically lower your bid to make the ROI higher. Would make sense, right?

It sure would! Except there's another very important trend to consider: Bid prices are driven by the market, i.e. they are dependent on how much your competitors are bidding.

This is the gist of what typically happens: Your competitors, like yourself, would target all placements in the beginning. But as they know which placements convert well and which ones don't, they would blacklist the bad placements, and then bid higher to get more traffic from the remaining good placements. The result? Average bid prices for good placements are higher, and average bid prices for bad placements are lower.

In turn, when you bid lower, you'll get traffic from worse-converting placements; and when you bid higher, you'll get traffic from better-converting placements.

So then your next thought may be: Great - then I'll just bid high to get traffic from the best placements! Although this may sound like a good plan, it isn't always.

Remember that in order to get this traffic from the best placements, you'd also be paying high prices. Similarly, when you bid lower, sure enough, you won't be getting much traffic from the very best placements, but then you won't be paying much for that traffic either.

Therefore, generally speaking, you'd be getting what you pay for: Bid higher = pay more for better traffic, or bid lower = pay less for worse traffic. And either end of the spectrum can make you profits.

Now, let's look at this from different perspective: Zoom in on a specific placement, and look at how changing the bid will affect how much we're paying for this placement and the traffic volume we'd get from this placement. Let's call this Placement X.

Let's say Placement X converts really well. So those of our competitors that have been running campaigns for a while, would already know this. They would either start a campaign to target JUST Placement X (possibly along with other good placements) and bid high to get more traffic, or they would blacklist the bad placements and then increase the campaign bid to achieve the same (i.e. get more traffic from the remaining, good placements).

Let's say we've set a low campaign budget. At this budget, the traffic source may not even allocate Placement X traffic to us, because they would want to sell this at higher prices to competitors that are bidding higher.

As we increase our campaign budget, we will gradually see traffic from Placement X. The more we increase our budget, the more Placement X traffic we should see. But in order to get more traffic volume, we'd be paying higher costs.

So when should we stop increasing this bid? Obviously, we want to maximize our profits. The lower we bid, the less money we pay, but the less traffic we'll get from Placement X. The higher we bid, the more money we pay, but the more traffic we'll get. The trick is finding that sweet spot - the exact bid that will maximize our profits for that placement.

The scenario above, where I was changing the bid to observe the change in prices and traffic volume from Placement X, was designed just for illustration purposes. Each campaign we set up can get traffic from thousands of placements. We can't possibly set up so many campaigns to test so many placements, just to find the "sweet spot" bid for every placement.

Fortunately, there are more efficient ways to test bids. Below I'll present 2 that I regularly use - they should serve you well.



Bid-Testing Approach A

This approach will save you money, but will require more time to execute, compared to approach B below.

Here are the basic steps:

1)Run your original campaign (assuming at around average bid or slightly higher) until at least most of the larger placements have received enough traffic for you to decide whether to cut or not.

2)Once most of the bigger unprofitable placements have been cut, clone the campaign (so that the new campaign has the same placements blacklisted from the start). For the new campaign, set a higher bid, $0.50-$1 higher than the bid in the original campaign.

3)Run this new campaign until, again, most of the bigger unprofitable placements have been cut.

4)Repeat: Clone the 2nd campaign, set a higher bid ($0.50-$1 higher than the 2nd campaign), run and cut placements.

5)Can repeat until you have 3, 5, or even 10 campaigns, all running at different bids.

Central idea of this method: In the end, you may have multiple campaigns running, each containing placements that are profitable at that particular bid.

You won't just be getting traffic from each placement at the "sweet spot" bid, but you'll get the next best thing: You'll likely see traffic from the same placement at multiple bids, at different traffic volumes.

Example - Let's use Placement X again. We may see this trend from the 5 (let's say) staggered-bid campaigns we set up:

Bid = $1 (original camp); no traffic from Placement X because bid is too low

Bid = $2; 500 impressions/day from Placement X, ROI = 150%; placement profitable so keep running

Bid = $3; 1000 impressions/day from Placement X, ROI = 100%; placement profitable so keep running

Bid = $4; 1500 impressions/day from Placement X, ROI = 30%; placement profitable so keep running

Bid = $5; 2000 impressions/day from Placement X, ROI = -30%; placement unprofitable so cut it from this campaign

This example serves to illustrate that you may see traffic from the same placements in different bid-camps. A placement can be profitable in some camps and not others - and you would of course keep the placements that are in profit and cut the ones that aren't.

It would be up to you how many camps you want to set up to test different bids. The more bids you test, the more profits you can stand to make. The flipside is that the most bids you test, the more money you'll need to spend.

Worth mentioning is the reason why I suggested for you to test bid points that are $0.50-$1 apart: Usually if you test bids closer than that, you'd get traffic from pretty much the same set of placements, which isn't what we're wanting to achieve. By testing different bids, we're wanting to trigger traffic from different placements.

Important: Not every bid-camp would be worth continue running! In the previous lesson, we covered how to estimate daily profits, and listed some positive and negative signs to look for in placement stats. If stats don't look promising, pause the campaign without spending further money. Don't chase after pennies and dimes. Our time is valuable too.

Note: This testing approach assumes that if a placement isn't profitable at a certain bid, that it probably will not be profitable at higher bids either. This is why when we clone a camp and set it to a higher bid, we continue to blacklist the placements that were blacklisted in the lower-bid campaigns.

Also note: In addition to testing higher bids, you can set up campaigns to test lower bids as well. Say you were bidding $1 in the original camp, you could test $0.50 and/or even $0.30. When testing lower bids, you may or may not want to inherit blacklisted placements from the original average-bid campaign, because there's a chance that the worse-converting placements may end up profitable at the lower bids. However, if you do keep those placements blacklisted from the start, you could save some money if those placements still turn out to be unprofitable even at the lower bids. So - pros and cons, and I leave this decision for you to make.



Bid-Testing Approach B

This approach will cost you more money, but will save you time, compared to approach A above.

Here are the basic steps:

1)Run your original campaign.

2)Once most of the bigger unprofitable placements have been cut, clone it several times (3...5...10...), and set them to different bids. (Again, the higher-bid camps can inherit the blacklisted placements from the original camp.)

3)Run and cut placements for each campaign.

Everything I said in the Approach A section above also applies to Approach B. The only difference with Approach B is, instead of cutting placements and then increasing the bid and then cutting more placements before increasing the bid further, we are setting up all bid-camps at the same time. This will optimize your campaign more quickly because we're testing multiple bids at the same time, but this approach will also be more expensive, as we would be testing and cutting some of the same placements in multiple bid-camps. If you're wanting speed and wouldn't mind spending more money, try Approach B.



Bid-Testing Approach C

There's one issue you may run into with Bid-Testing Approaches A & B above:

Some traffic network algos will aim to maximize their profits by sending a BIG portion of traffic to your highest-bid camp, when you have multiple camps running with different bids (for the same targeting).

This would result in your not being able to test the lower bids effectively - because they won't get enough traffic - because it's all being "hogged" by the highest-bid camp.

If you observe this, try Bid-Testing Approach C:

1)Run your original campaign.

2)Once most of the bigger unprofitable placements have been cut, increase the bid. Note in your campaign journal a)what the current estimated daily profit is, and b)the exact date and hour you increase the bid.

3)Run for a while to again cut the bigger unprofitable placements. Then estimate the daily profit at the new bid, i.e. since the bid increase.

4)If the estimated daily profit at the new bid is HIGHER, repeat steps 2) and 3). If it is LOWER, revert to the previous bid.

To summarize: Use the same campaign to test higher and higher bids until your profits start to decrease, then revert to the last bid to maximize profit levels.

**************************************************

And that's a wrap for the optimization module!

The tutorial is coming to an end - just a few more lessons on troubleshooting and scaling. We're almost there!





Amy