The client would fill almost all the available inventory, but because of the high traffic fluctuations, 35% - 65% of the contracts were regularly left unfulfilled. The traffic was strongly influenced by external events such as film premieres and even the weather.
The situation was complexified by the fact that 55% of impressions were sold by the auction model, often at the first priority. Moreover, there were buyers with First-Look access who could buy the impression even before the auction started. It impeded the assessment of the inventory available for direct deals.
Inventale Forecasting solved the problem of delayed campaigns.
We implemented a system for the client that solved three tasks at once
The system predicted traffic spikes taking into account the features of the video inventory. The number of pre-rolls, mid-rolls, and post-rolls depends on the content duration, the content tagging, the amount of ads already viewed by the user and how long ago they were viewed, etc. To take all this into account in the calculations, we practically replicated inventory placement algorithms.
Since part of the RTB inventory had been sold at the first priority, we integrated with the client’s custom ad server, simulated the ad server’s delivery logic, and virtually run an auction for every impression with a 90%+ accuracy. Traffic managers could now estimate the result of each ad campaign delivery and adjust its parameters to prevent delays already at the very planning stage.
The algorithms recommended changes in the ad campaign’s settings, based on delivery forecasts. The system monitored the delivery over the network 24/7 and adjusted the recommendations when needed. To top it off, the Inventale platform generated detailed reports.