Curve templates are used to define expected deviations of forecasted inventory, could be an expected increase or decrease of inventory. The specific distribution curve templates can be set on all the media tree elements: Media, Zone and Leaf. You can choose to either Boost or Decrease inventory for specific Weekdays, Weeks or Hours in day
Curves definition is done in the Administration (Main) -> Ad Configurations -> Deviation Curve Templates.
You can create as many curves as you want, and then assign them to the media zones.
Boost/Decrease of inventory for specific week, day in week or hour of day. Use this to increase or decrease inventory for specific weeks, day in week or hour of day. When defining a curve, modify the value of the “Boost %” column, and the percentage will be calculated automatically.
Note! for WeekinYear templates the Point Name starts with 0, so row 0 is the same as week 1 in the Year.
If you want to decrease inventory you can add a negative value in Boost % column.
Point value column (optional)
Normally it should be enough to just use the Boost column to increase or decrease your future inventory. The “Point Value” column can be used to alter inventory relative to each other, so an increase in one row would decrease inventory for the other rows. When defining a curve, modify the “Point Value” column, and the percentage will be calculated automatically for all rows. We did it this way to make sure the percentages always add up to 100%.
For example, in DayInWeek curve, you define how many weekly impressions happen during each day. Let’s say, you have 77,000 impressions a week on a node/zone; between Monday and Friday it is 10,000 impressions a day, on Saturday: 12,000, and 15,000 on Sunday.
The curve configuration would then look like this:
After curve(s) is/are defined and saved, they can be used in Inventory Config of a zone.
Uncheck automatic curves calculations, and choose a desired curve from a dropdown list. DayInWeek and HourInDay curves are used only if the forecasting is based on counters; if it’s set to use manually entered values, those remain unchanged. Don’t forget to Save after adding curve templates.
Curves in action
After curves are assigned to the zone, the inventory uses it to forecast number of impressions per each hour.
The calculation is done using the following algorithm:
- Using counters 5 weeks weighted average, we calculate expected impressions per week for the given zone
- This number is then distributed into 7 days according to the distribution specified by the DayInWeek curve
- Then each day value is distributed into 24 hours according to the distribution specified by the HourInDay curve
- The process above gives us a week template, 168 values with number of impressions forecasted for each hour of the week
- Week template values are increased according to Yearly Increase Percentage values (in a way that values one year ahead will be greater than next week forecast by that percentage)
- Those values are then updated according to the WeekInYear curve, proportionally to the distribution of that curve.