In order to avoid that your budget is spent on the campaigns with less performance, you should control manually the budget allocation. Shift the budget to the best performing campaigns and consider pausing the under performing campaigns if you have a limited budget.
You should try to spend the budgets throughout the day when you have big audiences, like in prospecting campaign or retargeting with more than 500 thousand unique cookies. Most of the DSP´s have the option to spend the budget evenly throughout the day. Use this option when you have a big audience pool and/or a large inventory. Consider only spending ASAP when you have a small inventory or a small audience, and you are not spending all the budget for the specific strategy.
After 24 hours you should check if your creatives are approved in all ad exchanges. At the moment, the main ones are Doubleclick Ad Exchange and Appnexus. Both have specific rules that you should follow. If the creatives are disapproved on these two ad exchanges, your campaign reach will be too small, specially on desktop.
Pixels have loads
Check weekly if all pixels are having loads. Normally all the DSP´s have reports for the pixel load. This is very useful for tracking and to create audiences that you will use on retargeting, especially if you use an audience strategy with a conversion path.
Consider excluding the worst performing sites, instead of exchanges. If you see group of sites that perform much better than others, consider have them in a single strategy with higher bids. Break out the lower performing sites and reduce the bids on them.
Bidding plays a big roll in achieving the right users. You need to win the most valuable impressions, however keeping your performance. Find the balance can be difficult and most of the times a good creative will allow you to bid higher. Raise the bids in the best performing strategies (normally retargeting) and try to use the bid optimization from the DSP’s. Each DSP built different bidding algorithms and you will be able to see that they work very differently. You should be able to optimize for actions, clicks or cost per view (in video).
Break out the strategies by recency windows. Test retargeting in minutes, and days, and tier the bids according to the performance of the strategies.
Try to find out the best number of ads show to user in one day. Normally the reports of Frequency cap are limited per user / per day. To test different frequency caps you may need to create control groups of an audience with different frequency to compare the results.
If you have a reduced budget, you should run it only in the best hours of the day, excluding the worst hours. You can also breakout the budget per the best hours and the worst hours if you have enough budget for that, and also increase the bid for the best hours and reduce for the worst hours.
Geo (Country / City / Postal Codes)
Run the budget through the best locations, and bid higher on the locations with the best performance. If you have enough history, exclude the worst locations based on the kpis you are looking for.
In Programmatic the inventory is labeled as a) above the fold, b) below the fold or c) unknown. Above the fold means the ad is visible before any scrolling happens. Below the fold is after the scroll, and unknown means that the position was not declared by the publisher. Have in mind that globally, around 50% of the inventory in programmatic is not declared, being 37% above the fold and 12% below the fold. According to your main kpi, could be differences between targeting the ad position.
If you have a performance campaign, this can be one of the optimization levers with with more impact, as many times, the good size of the creative combined with a website, or an audience, can produce much better results according to your kpi goals.
Consider to spend the budget only with the top performing sizes, if you have a limited budget. If you have a large budget create a specific campaign for the worst sizes, and exclude the sites that don´t perform. Then you do an exploration with these worst sizes, but you end up not quitting on them.