Floodlight only records conversions for users who have previously seen or clicked on a DFA ad within a period of time that you specify, called a lookback window. When a user loads a webpage containing Floodlight tags, the ad server accesses the user's DoubleClick cookie. DFA then runs an overnight process for each Floodlight impression, checking to see whether the ID is a match for any previous impressions or clicks on DFA ads within the lookback window. If no match for the cookie ID is found, the record of the activity is discarded.
The lookback window for any given user impression or click is based on the value set for the placement where it took place. If a user has clicked or viewed ads in multiple placements, DFA uses the most recent click that falls within the lookback window of the placement where the ad was served. If no click is available, the most recent impression is used.
Bob's Widgets is running a campaign across Placement A and Placement B. A user sees the ad called "Love the Lemur" in Placement A. A few minutes later, the user sees the ad called "Pet the Panda" in Placement B and decides to click because pandas are adorable. Ten days later, the user visits the advertiser's site and makes a purchase, triggering a Floodlight impression for a sales activity.
Placement A's lookback window for impressions is 14 days. Placement B's lookback windows for clicks is 5 days, so the click can't be used for calculating Floodlight data. However, Placement B's lookback window for impressions is 14 days. As such, Floodlight records the impression as a sales activity and associates it with the impression of the ad "Pet the Panda" served to Placement B.
Note that for publisher tags, the lookback windows for the advertiser are used, not the lookback windows for the placement. To avoid discrepancies, we recommend that you set lookback windows at the advertiser level and use default values for placements.
Lookback windows and publisher tags
Floodlight uses advertiser lookback windows, not placement lookback windows, to decide whether to serve publisher tags. That means that if you've got lookback window overrides in a placement, you might serve the publisher tags more or less often than you count a conversion.
For example, the lookback window for clicks is 3 days at the advertiser level, but 7 days at the placement level. A user clicks on an ad on Site A, then visits a Floodlight-tagged webpage 5 days later. A conversion is counted and attributed to the Site A based on the placement lookback window. However, based on the advertiser lookback window, no publisher tag is served.
This discrepancy is particularly significant if you use publisher tags to serve tracking tags for your sites, because it could lead to a disagreement over the number of conversions the site has provided.
This is only an issue if you use overrides in your placement lookback windows. To avoid this problem, don't use overrides.
Effects of changing placement lookback windows
You can change lookback windows to make them longer or shorter.
Keep in mind that DFA stores lookback data only for the length of time set in the lookback window. That means that when you expand the window, it will take time for DFA to collect data for the full length of time. For example, on January 15, you expand the lookback window for a placement from 15 to 30 days. On that day, DFA has only 15 days of data, so the effective lookback window is still only 15 days. On January 16, another day's data is added, for a total of 16 days of data. Day by day, more data is added. On January 30, a full 30 days of data will be available.
If you reduce the lookback window, the result is simpler: DFA simply discards data from days that are now outside of the new, reduced lookback window. So if you change the lookback window for a placement from 30 days to 14, only the most recent 14 days of data will be kept.
Changes to lookback windows are processed overnight.