Skip to content

What is ATL

ATL (above the line) modelling is how we measure the impact of TV, Facebook, OOH (Out of Home; Billboards, Tube campaigns) in terms of traffic and conversions. This type of modelling is more difficult than usual because we do not know exactly which visitors saw a specific TV ad, billboard etc. Hence, ATL predictions are built upon us modelling a 'baseline' of traffic, where no ATL campaigns were running. We can then use this baseline to measure the incremental difference in traffic when ATL campaigns are running.

  1. Configuration

To configure ATL modelling for a client, use Account Configuration in Dash. We require a period in time where no 'ads' were running, so we can build our baseline. You can also set a region for your ATL ad, for example, if the client ran a TV campaign in a specific location.

Once the configuration has been added, we would expect the following three tables to have data:

Table Name
attrib_atl
attrib_atl_config
attrib_atl_region

  1. Training Model & Predictions

    Once the configuration steps above have been completed, we can train the atl model, and then, start predicting traffic and sales driven by ATL.

    • Training the ATL model (baseline predictions):

      • Command: update_fabric_train_atl_baseline
      • Output Table: agg_atl_base_traffic_model
      • When: daily from the crons

    • Predicting the incremental traffic and sales which are driven by ATL.

      • Command: update_fabric_report_atl_base_traffic_forecast
      • Output Table: report_atl_base_traffic_forecast
      • When: daily from the crons

The train script always needs to be run before the predict script, as the output of the model train feeds the predictions. The data will be reported in the ATL section within dash (Channels -> Insights -> ATL Performance).