<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=822494444590823&amp;ev=PageView&amp;noscript=1">

Apr 10, 2018 8:53:02 AM

3rd Golden Rule to Header Bidding Success: The proper use of header bidding analytics

In this article, we'll discuss the proper use of header bidding analytics in order to optimize the output of your header bidding setup. 

Publishers implement header bidding to increase the level of competition in their programmatic auctions and as a result, increase their revenue.  

That said, in practice you'll only maximize the revenue uplift if your header bidding stack works well. And since this can be complex to manage, relying on the right analytics tools will make a big difference.

Not everyone has the time to spend ages on analytics, running reports per country, partner, format... testing new partners and identifying issues. We see publishers adopting two different strategies:

  1. "Basic header bidding monitoring" to make sure everything runs well.
  2. "Advanced header bidding stack management" - to adjust your stack and continuously test different setups

 

Basic header bidding monitoring

Scenario 1_Stable publishers

Getting this header bidding “basic monitoring” right is a two step process (and it’s not so basic by the way).

First step - Get visibility on true HB metrics

Before even thinking of what metric(s) to look at, you need to make sure you’re looking at your true data.

Most of you see what we’re coming at with this, but yes, monitoring your discrepancies is critical when you’re in a header bidding setup.  You need to make sure the revenue you see in your analytics tools is in line with the numbers you see in your SSPs and ad servers - otherwise, any decision you make will be biased and could potentially harm your revenue.

There are several types of discrepancies:

  • Volume discrepancies, i.e. discrepancies between the impressions reports from the programmatic partner and the publisher ad server impression reports - often linked to integration/implementation issues.
  • CPM and revenue discrepancies, meaning that you don’t see the same revenue in your header partners and in your adserver. These can have several cause, for instance:
    • Your different reporting tools do not show data on the same time zone (ie if DFP is in PST and your SSPs in ET, you might have an issue).
    • Some platforms show you “gross” metrics, and other “net”. Make also sure that all your partners bid in net - or at least that you manage well gross bids.
    • Your bid buckets are not granular enough.

In this quest to solve header bidding discrepancies, the right analytics tools (and the right account management team), can be a game changer to help Ops teams get visibility on their true header bidding metrics.

Second step - Discrepancies Monitoring

When you finally get your hands on your real data, to make sure everything runs well, start by monitoring the evolution over time of very simple metrics such as:

  • CPM
  • Revenue/Volume repartition between your partners and your Ad Server / Primary SSP (DFP/AdX usually).

As an example, if one partner displays a CPM way higher than others or too close to your primary SSP (AdX), or if the volume of one partner drops without being compensated by an increase in the same proportion on another partner, this is probably that there is something wrong in the setup.

Being able to monitor page latency is also a plus due to the negative impact that can be generated by Header Bidding.

That said, to be able to make educated comparisons between partners, what you critically need is an analytics tools that connects to all your partners (including ad server) and joins all data sources to give you a holistic view on your stack.

Without this 360° vision, when faced with an issue, you most probably won’t be able to properly identify issues, dig deeper and figure out the root cause/workaround. To go one step further, the right analytics package should let you breakdown your KPIs by partner, but also by buyer/brand to make sure you're investigating and troubleshooting issues at the most granular level.  

Another piece of advice from our side: whatever tool your programmatic team ends up choosing, to keep the maintenance burden manageable, it might be worth keeping your number of demand partners stable, and limited. You can check our article on the topic here - hint: there's no magic number.

 

Advanced header bidding setup management

Publishers_changes

To go beyond simply making sure the setup works well and make continuous optimizations to your header bidding stack, you might need a more advanced analytics package. We've been rambling about this in this blog already, but if this is your case, we would even recommend building a testing framework (typically with an A/B test) to assess the impacts on your revenue.

Let’s explore the (frequent) case of a publisher willing to add a new partner in the header. To make sure the new partner is ramping-up well, this publisher will need to look at “basic” metrics (revenue/CPM - as said above) but also:

  • Participation rate, ie is the new partner able to participate in Header Bidding auctions.
  • Win rate - participation is not enough, to translate into revenue, the new partner also needs to win some auctions.
  • Response time - is the partner’s one in line with your timeout? If not, would a small increase of your timeout generate enough additional revenue?
  • Incremental revenue - if the partner is able to participate and win, the question is whether it’s bringing any additional revenue (vs. just cannibalizing existing demand sources). Here, it can get tricky. What we recommend is:
    • Plugging the partner on a fraction of your inventory only. This fraction should be randomly determined via random A/B test - let’s call it, the “A group”.
    • Comparing this inventory with exactly similar inventory (the “B” group - determined by the same random A/B test). This will allow you to see the true revenue and CPM uplift from the new partner - if any.

Another method that we recommend when setting up an A/B test is to work on Header Bidding bid level data to estimate the additional revenue through simulations.

With such a framework, it gets quite simple to decide whether to keep the new partner or not. You simply need the right tool to support this advanced analysis.

This is just a typical example where using the right product as well as implementing a solid testing framework can make a big difference. Other cases we hear about a lot are for instance “how to troubleshoot the case of a partner with a very low participation rate?” or “is it worth moving a header bidding partner from client wrapper to server-side wrapper?”.

 

If as a publisher, this is something you're also seeing/struggling with and would like to give us your thoughts, please reach out to the Adomik team or find us in the AdOps channel in Slack by tagging Adomik. We look forward to hearing from you!

 

 < Previous Article - The Number of Header Partners           Next Article - Managing Prices in a HB setup 

Topics: programmatic publisher, header bidding, header bidding optimization, 5 header bidding golden rules, header bidding analytics, Yield Teams

Get the latest in programmatic yield management:

Subscribe Via RSS
Recent
Topics

see all

If you are a programmatic seller looking to:

  • Have a unified understanding of your stack
  • Increase your revenue by up to 30%+
  • Better leverage private marketplaces
  • Maximize your header bidding channel

Reach out to us, we can help. Click below to get in touch. We look forward to hearing from you.

Schedule Your Demo