Every advertiser wants to increase conversions without violating its ROI goals. While this case study examines one campaign in particular, the story is universal across Right Media clients: moving from a static CPM to a dynamic pricing model removes typical inventory constraints and allows campaigns to deliver more efficiently.

Dynamic Pricing allows each individual impression to be priced flexibly, on every ad call, based on its true value to the advertiser. Instead of being optimized out of lower quality inventory, the advertiser can simply bid less for it. Instead of being outbid on higher quality inventory, the advertiser can offer to pay more for higher value impressions. Whatever the case, the system monitors ROI targets every time it sets a bid for an advertiser.

Goals and Method

• Goals of the two-phase test were to acquire new customers, and
• control CPA while maximizing conversions and scale.

The test looked at the performance of two ad units (300×250, pops) across two time periods of equal length (18-19 days) and scale (10-11mm impressions). The campaign was priced on a flat CPM in phase I, and a dynamic CPM in phase II.

Phase I

Not yet optimized, the campaign started at a high CPA and delivered broadly. CPA dropped progressively as the system learned and became more selective about when to serve. The campaign averaged around 9 conversions a day, and it hovered around this level as long as its static price restricted it from higher and lower priced inventory.

Phase II

Upon the switch to dynamic pricing, conversions skyrocketed. Instead of ruling out impressions of different value, the advertiser could now bid on them relative to that value, as long as its campaign targets were not violated.

Delivery increased, and the campaign garnered nearly 5x more daily signups, on average, than it did when it was priced statically. At the same time, CPA continued to trend downward as the system predicted response more accurately and got smarter about how to spend the campaign’s budget.

Summary

Once the client could pay according to value, instead of paying the same, flat price for every impression, the advertiser opened itself up to much more inventory and won considerably more new signups. Cost per user was controlled, and ROI was maximized on the campaign. While nearly every network started on the buy, only Right Media remained at the end.

On its surface, it seems counter intuitive to adjust the CPM bid price, often adjusting down for an advertiser, but in fact, there is a compelling reason for it. How many times have you heard a media buyer say after their campaign has been running, “publisher X, you’re not hitting my goal, you need to ‘optimize or we’ll have to cancel the buy.” If an advertiser is not hitting reasonable ROI metrics established on the campaign, there are 3 options: 1. keep price the same, miss the target goals and eventually lose the advertiser, 2. optimize potential inventory out of the campaign through frequency capping or manual removal of sites, sections or placements. The third option is a new Right Media option - 3. adjust price to match the value of the inventory (based on ad call variables), do not remove any inventory and hit the ROI goals.

Before Right Media’s Yield Manager, the choice with all ad servers was optimize by negation or lose the campaign. Adjusting price based on the value of the inventory to the advertiser is the only way to keep the advertiser relationship AND maintain enough access to inventory for a long term scalable campaign. It works in theory and in practice. Several of the same advertisers that started working with us 2 years ago are still with us, and we’ve added many more.

Bringing it back to the publisher, by allowing price to vary on each ad call based on the advertiser’s ROI goals, we create competition for every single ad call from a publisher. Competition among hundreds of advertisers is better than selling your inventory to one buyer at a large discount. 

Publishers get competition for their inventory. Advertisers get ROI driven Dynamic Pricing. Long term relationships are built based on the fairness to each and their goals.

There is no hard and fast answer to when it is appropriate to kill a creative.  The answer can be different for every advertiser and possibly even different for each advertiser line item.  Usually, though, an AM can figure out a general pattern over time for a particular LI or group of LIs.  I’ve learned that figuring out this pattern is more of an art than a science. 

Yield Manager’s most powerful “quick” optimization tools are the target CPA setting and the learning controls that set how much learning is allowed for each creative (aka, ROI slider). Other more detailed controls such as geo, channel and publisher targeting, and frequency caps, to name a few, are all part of the overall larger concept of optimization, but continually moving an account towards or keeping an account at the desired CPA target is a balance of adjusting both the CPA target and the ROI slider.

As a rule, at a given point in time, I try to adjust only one or the other so that I can see which change has the most impact, but when I have an account that doesn’t react in an obvious way to a change in only one of these variables, I may do both. I’m selective about employing this tactic because it is possible to cause a drastic or undesired movement in activity such as spending too much at a very high CPA or not spending enough at a high CPA. Who knows which outcome is worse!  

If I’m happy with impression volume but want a lower eCPA, I’ll adjust the CPA target setting down 1 to 5 cents, depending on how much it is currently above the target. If I’m happy with the eCPA but want more volume, I may simply adjust the learning slider to open it up for more learning impressions. If I’m not happy with either the volume or the eCPA, I might adjust the target down slightly while adjusting the slider to allow more learning, hopefully finding more sites/sections where the creatives will get positive results.   

Once an account reaches the point where it has a collection of creatives that have tested successfully and are performing consistently, I’m more likely to adjust for fewer learning impressions while keeping the CPA target constant or adjusting up very slightly so it can continue to win the impressions that learning has already proven work well towards the target. Regardless, I’m always mindful about adjusting the CPA target so low as to potentially make the creatives less competitive for the valuable impressions, or allowing so much learning that each creative risks too much budget at an eCPA outside of the advertiser’s goals.

I recently started a Right Media managed advertiser campaign. The account was booked as a dynamic CPM campaign with a CPA target. We employed daily spend caps of approximately $50 per line item to control the advertiser’s spending for the first several days.

The campaign had an incredibly successful launch; the first 24 hours provided the client with well over 100 leads, at an effective CPA 1/5 of their target.  This is an immediate signal to an account manager to begin to scale the campaign, but first one important issue must be addressed. Rather than make changes, I first verified with the advertiser that their targeting pixel was on the proper page. By checking this from the beginning, we were able to avoid unnecessary spending with poor results. Had the pixel been on the wrong page, Yield Manager would have optimized indefinitely on improper data, and the dramatic success would be negated within hours.

Once the client and I were sure that the pixel was properly located, I was free to increase the daily spend while maintaining the current level of ROI success.

The advertiser’s spend has since increased twofold and still regularly comes in well below their CPA target, thus demonstrating the scalability and power of the Yield Manager platform.