In this article series, we’re sharing our thoughts about rethinking and increasing yield in the current complex and fragmented digital advertising ecosystem. In the last two weeks, we wrote about:

So far we’ve mainly discussed yield tools. When it comes to yield management, tools are important but they’re not as impactful without the right organization. So that’s what we will address today: the right decision-making process to increase revenue.

The right KPI: the opportunity cost
We hear a lot about choosing KPIs, so here’s our two cents on the topic: to increase yield, the right KPI is the opportunity cost.
Quick definition before we go further: the opportunity cost is the loss in goods that you incur by affecting available resources to another use. This criteria is critical when deciding to invest. The quick rule of thumb is that the opportunity cost should be lower than the expected return of one investment.
The common mistakes publishers make are actually often misevaluation of opportunity costs:

  • If you don’t assess the opportunity cost of a PMP, you might end up underpricing it (and seeing a graph like the following, if you’re an Adomik user). In this case the opportunity is the revenue that Publisher would make by leaving the inventory in the Open Auction.

Screen Shot 2018-05-02 at 17.28.51
Source: Real data from a Dutch publisher – February/March 2018

  • Similarly, when opportunity costs are not evaluated we see poor pricing strategy between SSPs and Backfill – for instance, SSPs default floor rules can end up being lower than the Backfill RPM.

That said, it’s not that easy. In  a complicated stack, you have two kinds of opportunity costs:

  • “Horizontal ones” (within one level/dimension) – how much net incremental revenue am I making by adding one SSP to the stack? If I add a new proprietary tag will it cannibalize my advertising revenue?
  • Vertical ones (between layers such as Open Auction, Direct, PMP) – what’s the holistic impact of adding a new floor in open auction, creating a new PMP, keeping an annual guaranteed deal…?

And to make it even more complex, you need to take into account monetary costs ($$), but also operational costs, such as setup, management, monitoring.
To make this lat recommendation more actionable, let’s look at one practical case.

One example of reasoning based on opportunity costs
We’ll ask ourselves a question that most Publishers must have been faced with: in May 2018, should I renew a priority campaign ensuring me a monthly revenue of 10k$ in revenue and 5M of impressions during 12 months.
From the number above, the campaign expected revenue is 10k$*12=120k$.
What matters in order to make the decision is the campaign opportunity, i.e. what is your expected revenue if you choose not to renew it ( is it higher that 120k$?).
Answering that question is a 4-step process:
Step 1. Analyze the campaign distribution environment
First, to nail down the opportunity calculation you need to understand how the campaign is distributed (geo, format, device, audience segment…) as each environment has its own opportunity cost.
In our example, the monthly 5M impressions are distributed as follows (per month, format and geo):
Screen Shot 2018-05-28 at 14.57.49
Step 2. Calculate the opportunity of each environment
For each environment you have identified, you now need to evaluate how much revenue you would earn, without this campaign running. The metric you should look at is your RPM, excluding guaranteed campaign.
The formula is:
RPM without guaranteed campaign = Revenue without annual campaign / (total ad request – campaign sold volume).
If you’re using multiple SSPs, this calculation might be difficult – you’ll need to merge your Adserver and DFP data – but actually, you can use the clean database and management system you’ve built, if you’re following our #1 and #2 advice to rethink your yield 😉
Taking our example above, these are the RPMs we see:
Screen Shot 2018-05-28 at 15.07.49
Step 3. Calculate the total opportunity
You can now easily calculate the campaign total opportunity by multiplying the diffusion volume (step 1 table) by the RPM for each environment (Step 2. Table)
For each environment, the formula is: (Guaranteed Campaign sold volume * RPM excluding guaranteed campaign)/1000.
Screen Shot 2018-05-28 at 14.58.16
According to the above table, the opportunity is $90k. Even if the overall results is lower than the planned 120k$, the figures show that we would have achieved higher advertising revenue without the guaranteed campaign in May 2017 – this campaign might not be such a great deal after all.
Step 4. Consider non financial aspects!
We have

  • on one side, the guaranteed campaign with $120k in revenue
  • on the other, $90k in expected revenue based on your observed RPM on the campaign’s environments

—> So far, the expected net gain is $30k. However, we’ve only taken into account financial aspects. Before making the final decision, it’s critical to weigh other aspects as well (especially as using non-programmatic channels can entail additional work for teams). Among these:

  • Operations cost: will this campaign generate long and tedious setup?
  • Does this campaign put you at risk somehow (data leakage…)?
  • Will it entail heavy monitoring and reporting?

This is where we are: guaranteed revenue $120k – financial Opportunity costs ($90k) – non financial costs = your net profit. That’s it, you have everything in hands to make an educated decision!
A few limitations
We believe this approach it quite simple to implement, but it still shows a few drawbacks:

  • Analyzing a campaign distribution environment requires to be very granular – our example above is too simplified.
  • Calculating the RPM can be a quite the headache if you have to combine waterfall and header bidding data, especially as different platforms don’t necessarily have the RPM calculation method.
  • Even if you manage to be super granular, you’ll still be looking at averages RPMs (per month, per inventory type…). To be even more precise, you’d need to look at the RPM distribution in each environment… but let’s keep that fun part for next time 🙂

To conclude, when trying to increase yield management in a multi-product /multi-channel advertising ecosystem, the opportunity is the best metric – even if it’s has its drawbacks.
But beyond the metrics, database etc., what really matters is implementing the right organization.

  • If all your teams (Programmatic vs. Direct) are competing against one another, by definition it may not be the best setup to make sure you adopt a holistic approach to yield management
  • However, with an integrated autonomous team with decision making power on the ratecard, you may have a better chance!

Since a diagram is better than a 100 words, here is an example of target organization:
Screen Shot 2018-05-28 at 14.58.28

Tags :

holistic yield management programmatic yield management publishers yield management Yield Teams