OK, so you’ve learned more about header bidding and how it works. You’ve also discovered why it makes practical sense for you as a digital publisher to set up header bidding, since it helps you get better returns for your digital inventory. But the question remains: what are the steps you need to take to set up the right header bidding strategy?
Honestly, it depends on where you’re starting from and what you’re hoping to achieve as an online publisher. To help figure out what makes sense for your organization, AppNexus has compiled some useful pointers on implementing a successful header bidding strategy.
1. Choose a Good, Open-Source “Wrapper”
2. Optimize Your Wrapper So It Works Even Better
Once you’ve found the wrapper that works best for you, you’ll want to configure the code in accordance with the partners that you’ve chosen to work with. To do that, the first question you need to ask is how many header-bidding partners should you be working with ideally?
The honest answer is that it really depends. Popular belief places the ideal number of bidding partners at around 10, but that figure’s a little misleading. It really depends on what your business goals are – and how minutely you want to use header bidding to make sure those goals happen.
Here are two major factors to consider before creating your final line-item list of header-bidding partners:
Factor #1: Timeout rate
One of the great advantages of header bidding for publishers is that it can eliminate the problem of latency. For more about how it helps, we recommend you read Ben Kneen’s article, “Flattening Your Waterfall with Prebid.”
But publisher beware: header bidding can still create latency if you don’t select the proper bidding partners. Different exchanges will bid at different speeds for your inventory. As a publisher, it’s up to you to optimize the timeout rate of your instant auction. As a rule of thumb, we recommend publishers in the US, Canada, and Europe opt for timeout rates of 400 – 800 milliseconds for display and 800 – 1200 milliseconds for mobile. That gives most exchanges enough time to place bids on your impression, but still keeps things fast enough that users aren’t left waiting too long for your ad to load. For a more in-depth look at optimal timeout rates, you can check out prebid.org and leverage AppNexus’ Headerbid Expert, a chrome extension that allows you to analyze each header bidding partners’ performance in order to earn more and negate latency.
Factor #2: Pricing Strategy
This is a big one. In the same way different bidding partners bid for impressions at different speeds, they also bid at different prices. As a publisher looking to optimize your inventory yield, you want to develop a line-item strategy that captures the bid prices with the high degree of accuracy, while limiting the required setup work.
Most bids tend to be relatively low ones (around $1 – 2) and cluster together. In other words, for lower value impressions there’s a high chance your bidders will value it within pennies of each other. After these come mid-range bids (anywhere from $2 – 5) followed by high-level bids (between $5 – 10), where you are more likely to have an outlier from a specific buyer. That is, bidders may all see that impression as valuable, but are more spread out. Meaning, your highest bidder is much more likely to value the impression at least $0.10 or even $0.50 higher than another. Because of this, you can save time when setting up your header-bidding strategy by using “price buckets” with different ranges that take these behaviors into account.
For lower bids, we absolutely recommend you configure your price buckets at the granular “penny level” so that you’re able to capture the exact yield of your lower bids. If winning bids of $1.22 and $1.25 come into play, you want to deliver the impression to the $1.25 bidder and get that full value. If you were to use dime buckets instead, both bids would just be rounded to $1.20 in the ad server, and would randomly deliver, potentially costing you $0.03 in revenue. It doesn’t sound like a lot but over millions of impressions it adds up!
For mid-range bids however, while penny buckets would be most accurate, using dime buckets will let you use 1/10 the number of line items while having a minimal impact on revenue. That way, if winning bids come in at $5.30 and $5.39 (which would be rare in itself), the most you can possibly lose is $0.09.
Using the same logic, at the high end, you can move to a $0.50 price bucket. If you end up receiving a rare auction with bidders at $10.00 and $10.49 the most you can lose due to ad server selection logic would be $0.49 cents. Far more likely is you’d see one bidder submit an auction at $10.00 and others bidding at much lower levels, meaning an approximate value will more than suffice in selecting the optimal line item.
3. Set up Your Code and Start Monitoring its Performance
After you set up your line-item code on a particular webpage, you want to target each line item so that it produces the greatest revenue gains possible. That means monitoring your bidding partners’ performance over the long run. To accomplish that, here are three KPIs you need to look out for:
- Response Rate: If one of your bidding partners doesn’t even respond to your auction alerts (be it a real bid or a zero bid response), they can’t possibly win an auction. This may indicate your timeout rate is too restrictive, or your partner’s tech isn’t up to snuff.
- Bid Rate: Start tracking your auctions in general to see which bidders can submit a real bid most of the time. Frequent bids can indicate a high level of interest on the part of bidder for your inventory. But even if a bidder isn’t frequently, if they aren’t bidding high enough, it might be time to bid them farewell.
- Win Rate: Of all the bidders actually bidding for your inventory, how many of them are actually – and frequently – winning their bids? These are the keepers – the ones who’ll help you stay in business over the long term.