Optimizer Strategies

Might be a simple thing to ask but not always so simple to answer as you may be trying to achieve more than one thing when introducing Bidstream Optimizer.

Define a Strategy - What does success look like?

While there may be more than one ambition it's good to have an overarching objective. That way you can be clear on what success looks like and set targets to achieving this.

To help we have described here some popular strategies worth considering. More than one of these may be important but aim for one to be primary.

• Demand Strategies
• Supply Strategies

Demand-side Strategies

These strategies are most suited to a demand platform (DSP) or a supply platform that is ingesting inventory from a third party via API.

Reduce OPEX without affecting spend

Objective

You have steady/fixed demand but the volume of bid requests you have to manage is rising. You want to remove bid requests that aren't generating any value from your bidstream to reduce infrastructure costs and platform load.

Background

There are underlying causes for the rise in bid requests: header bidding, bid duplication by exchanges/SSPs and arbitrage. There are also many bid requests that are passed in streams that have zero value to the demand partner. For example, if you only buy video inventory then you only want to receive video inventory in your stream. But often this is not the case.

Strategy

With Bidstream Optimizer you can implement a commercial strategy to remove supply that is generating no bid activity. This will slowly remove from the bidstream those requests that are not relevant ensuring that spend levels are not affected in any way. There is always a percentage of inventory that is allowed through in the bidstream as a control group to ensure that inventory is not filtered out that may meet future demand.

KPIS

Setting the Strategy

Different levels of filtering can be applied to Optimizer to reduce the bidstream. The more aggressive the filtering the greater the risk to Gross Spend, but also the greater the reduction in OPEX. Deciding what the "threshold" should be is individual choice.

This is best illustrated in this chart, which shows for a single campaign the impact different filter rates have on spend.

[insert the elbow chart]

In this chart the filter rate has an effect on spend. However, the line between the dots is not linear and so it's best to set your initial filter rate somewhere near where the line creates an elbow. This is where maximum value is being generated from optimizer. If you want to be more conservative initially then bring the filter rate lower than the elbow. If you want to be more aggressive then keep it on the elbow or higher.

It's important to note that this is a model. In reality what will happen is when the filter is applied, more inventory that is a better match for the campaign will be introduced to the stream, and maintaining spend levels.

Monitoring

There are two ways to monitor your strategy to make adjustments accordingly.

Runtime Monitoring

Monitor in real-time your two KPIs compared to the control group to ensure Spend / QPS is inline with expectations:

Analysis Reporting

Monitor the elbow compared to your filter rate and ensure that you are setting your filter rate in line with business goals. Adjust accordingly by contacting your account manager or managing directly using the Optimizer API.

Remove low quality inventory

Maximize top line revenue but not increase infrastructure costs

Enrich bid requests with new data to maximize revenue (as a supplier)

Turn on new supply

Profile demand partner bidstream to inventory that generates value to a fixed output QPS

Enrich bid requests with new data to maximize revenue

Turn on new demand

Supply-side Strategies

These strategies are for a supply platform that is managing the outbound bidstream it is sending to it's demand partners.