Forecasting

Introduction

This guide explains the different ways of using forecasting in the Google Ad Manager API.

Traffic Forecast

A traffic forecast lets you read historical traffic alongside future impression opportunities to find better ways to segment and package inventory.

Java

// Create the date range. Include the previous and next 7 days.
Interval interval =
    new Interval(
        Instant.now().plus(Duration.standardDays(-7)),
        Instant.now().plus(Duration.standardDays(7)));
DateRange dateRange = new DateRange();
dateRange.setStartDate(DateTimes.toDateTime(interval.getStart()).getDate());
dateRange.setEndDate(DateTimes.toDateTime(interval.getEnd()).getDate());

// Request the traffic data.
TrafficDataRequest trafficDataRequest = new TrafficDataRequest();
trafficDataRequest.setRequestedDateRange(dateRange);
trafficDataRequest.setTargeting(targeting);
TrafficDataResponse trafficData = forecastService.getTrafficData(trafficDataRequest);
    

View on GitHub

Python

# Create a start date that's 7 days in the past and an end date that's 7 days
# in the future.
today = datetime.date.today()
start_date = today - datetime.timedelta(days=7)
end_date = today + datetime.timedelta(days=7)

# Create targeting.
targeting = {
    'inventoryTargeting': {
        'targetedAdUnits': [
            {
                'includeDescendants': True,
                'adUnitId': root_ad_unit_id,
            }
        ]
    }
}

# Request the traffic forecast data.
traffic_data = forecast_service.getTrafficData({
    'targeting': targeting,
    'requestedDateRange': {
        'startDate': start_date,
        'endDate': end_date
    }
})
    

View on GitHub

PHP

// Create the date range. Include the previous and next 7 days.
$startDate = AdManagerDateTimes::fromDateTime(new DateTime('-7 day'))
    ->getDate();
$endDate = AdManagerDateTimes::fromDateTime(new DateTime('+7 day'))
    ->getDate();
$dateRange = new DateRange();
$dateRange->setStartDate($startDate);
$dateRange->setEndDate($endDate);

// Request the traffic forecast data.
$trafficDataRequest = new TrafficDataRequest();
$trafficDataRequest->setRequestedDateRange($dateRange);
$trafficDataRequest->setTargeting($targeting);
$trafficData = $forecastService->getTrafficData($trafficDataRequest);
    

View on GitHub

C#

TrafficDataRequest trafficDataRequest = new TrafficDataRequest() {
    requestedDateRange = new DateRange() {
        startDate =
            DateTimeUtilities.FromDateTime(startDate, "America/New_York").date,
        endDate = DateTimeUtilities.FromDateTime(endDate, "America/New_York").date
    },
    targeting = new Targeting() {
        inventoryTargeting = new InventoryTargeting()
        {
            targetedAdUnits = new AdUnitTargeting[] {
                new AdUnitTargeting() {
                 adUnitId = rootAdUnitId,
                 includeDescendants = true
                }
            }
        }
    }
};
TrafficDataResponse trafficData =
    forecastService.getTrafficData(trafficDataRequest);
    

View on GitHub

Ruby

# Create a start date that's 7 days in the past and an end date that's 7 days
# in the future.
today = ad_manager.today
start_date = today - 7
end_date = today + 7

# Create targeting.
targeting = {
  :inventory_targeting => {
    :targeted_ad_units => [
      {
        :include_descendants => true,
        :ad_unit_id => root_ad_unit_id
      }
    ]
  }
}

# Request the traffic forecast data.
traffic_data = forecast_service.get_traffic_data({
  :targeting => targeting,
  :requested_date_range => {
    :start_date => start_date.to_h, :end_date => end_date.to_h
  }
})
    

View on GitHub

Availability forecasts

An AvailabilityForecast reports the maximum number of available units with which the line item can be booked. This forecast is analogous to the check inventory feature in the UI.

The forecast includes the available, matched, possible, delivered, and reserved units. It can also include all the contending line items and the available units for each targeting breakdown depending on the options set in the AvailabilityForecastOptions. By default, neither is included in the forecast.

Targeting breakdowns include the matched and available units for each targeting criteria. These breakdown entries are automatically generated according to the line item targeting. For example, a line item targeted at ad unit ID 123456 would include a breakdown similar to the following:

<targetingCriteriaBreakdowns>
  <targetingDimension>AD_UNIT</targetingDimension>
  <targetingCriteriaId>123456</targetingCriteriaId>
  <targetingCriteriaName>My Ad Unit Name</targetingCriteriaName>
   <excluded>false</excluded>
   <availableUnits>1000</availableUnits>
   <matchedUnits>2300</matchedUnits>
</targetingCriteriaBreakdowns>

You can run availability forecasts on either an existing or prospective line item.

Existing LineItem

You can run an availability forecast for an existing line item by using its ID.

Java

  // Get the ForecastService.
  ForecastServiceInterface forecastService =
      adManagerServices.get(session, ForecastServiceInterface.class);

  // Get forecast for line item.
  AvailabilityForecastOptions options = new AvailabilityForecastOptions();
  options.setIncludeContendingLineItems(true);
  options.setIncludeTargetingCriteriaBreakdown(true);
  AvailabilityForecast forecast =
      forecastService.getAvailabilityForecastById(lineItemId, options);

  long matched = forecast.getMatchedUnits();
  double availablePercent = (forecast.getAvailableUnits() / (matched * 1.0)) * 100;
  String unitType = forecast.getUnitType().toString().toLowerCase();

  System.out.printf("%d %s matched.%n", matched, unitType);
  System.out.printf("%.2f%% %s available.%n", availablePercent, unitType);

  if (forecast.getPossibleUnits() != null) {
    double possiblePercent = (forecast.getPossibleUnits() / (matched * 1.0)) * 100;
    System.out.printf("%.2f%% %s possible.%n", possiblePercent, unitType);
  }

  System.out.printf(
      "%d contending line items.%n",
      forecast.getContendingLineItems() == null ? 0 : forecast.getContendingLineItems().length);
    

View on GitHub

Python

  # Initialize appropriate service.
  forecast_service = client.GetService('ForecastService', version='v202408')

  # Set forecasting options.
  forecast_options = {
      'includeContendingLineItems': True,
      'includeTargetingCriteriaBreakdown': True,
  }

  # Get forecast for line item.
  forecast = forecast_service.getAvailabilityForecastById(
      line_item_id, forecast_options)
  matched = int(forecast['matchedUnits'])
  available_units = int(forecast['availableUnits'])

  if matched > 0:
    available_percent = (float(available_units) / matched) * 100.
  else:
    available_percent = 0

  contending_line_items = getattr(forecast, 'contentingLineItems', [])

  # Display results.
  print('%s %s matched.' % (matched, forecast['unitType'].lower()))
  print('%s%% %s available.' % (
      available_percent, forecast['unitType'].lower()))
  print('%d contending line items.' % len(contending_line_items))

  if 'possibleUnits' in forecast and matched:
    possible_percent = (int(forecast['possibleUnits'])/float(matched)) * 100.
    print('%s%% %s possible' % (possible_percent, forecast['unitType'].lower()))
    

View on GitHub

PHP

      $forecastService = $serviceFactory->createForecastService($session);

      // Get forecast for line item.
      $options = new AvailabilityForecastOptions();
      $options->setIncludeContendingLineItems(true);
      $options->setIncludeTargetingCriteriaBreakdown(true);
      $forecast = $forecastService->getAvailabilityForecastById(
          $lineItemId,
          $options
      );

      // Print out forecast results.
      $matchedUnits = $forecast->getMatchedUnits();
      $unitType = strtolower($forecast->getUnitType());
      printf("%d %s matched.%s", $matchedUnits, $unitType, PHP_EOL);

      if ($matchedUnits > 0) {
          $availableUnits = $forecast->getAvailableUnits();
          $percentAvailableUnits = $availableUnits / $matchedUnits * 100;
          $possibleUnits = $forecast->getPossibleUnits();
          $percentPossibleUnits = $possibleUnits / $matchedUnits * 100;
          printf(
              "%.2d%% %s available.%s",
              $percentAvailableUnits,
              $unitType,
              PHP_EOL
          );
          printf(
              "%.2d%% %s possible.%s",
              $percentPossibleUnits,
              $unitType,
              PHP_EOL
          );
      }

      printf(
          "%d contending line items.%s",
          count($forecast->getContendingLineItems()),
          PHP_EOL
      );
    

View on GitHub

C#

using (ForecastService forecastService = user.GetService<ForecastService>())
{
// Get forecast for line item.
AvailabilityForecastOptions options = new AvailabilityForecastOptions();
options.includeContendingLineItems = true;
options.includeTargetingCriteriaBreakdown = true;
AvailabilityForecast forecast =
    forecastService.getAvailabilityForecastById(lineItemId, options);

// Display results.
long matched = forecast.matchedUnits;
double availablePercent =
    (double) (forecast.availableUnits / (matched * 1.0)) * 100;
String unitType = forecast.unitType.ToString().ToLower();

Console.WriteLine("{0} {1} matched.\n{2} % {3} available.", matched, unitType,
    availablePercent, unitType);
if (forecast.possibleUnitsSpecified)
{
    double possiblePercent =
        (double) (forecast.possibleUnits / (matched * 1.0)) * 100;
    Console.WriteLine(possiblePercent + "% " + unitType + " possible.\n");
}

Console.WriteLine("{0} contending line items.",
    (forecast.contendingLineItems != null)
        ? forecast.contendingLineItems.Length
        : 0);
    

View on GitHub

Ruby

  # Get the ForecastService.
  forecast_service = ad_manager.service(:ForecastService, API_VERSION)
  # Set forecasting options.
  forecast_options = {
    :include_contending_line_items => True,
    :include_targeting_criteria_breakdown => True,
  }

  # Get forecast for the line item.
  forecast = forecast_service.get_availability_forecast_by_id(
      line_item_id, forecast_options
  )

  unless forecast.nil?
    # Display results.
    matched = forecast[:matched_units]
    available_percent = forecast[:available_units] * 100.0 / matched
    unit_type = forecast[:unit_type].to_s.downcase
    puts '%.2f %s matched.' % [matched, unit_type]
    puts '%.2f%% of %s available.' % [available_percent, unit_type]
    puts '%d contending line items.' % forecast[:contending_line_items].size
    unless forecast[:possible_units].nil?
      possible_percent = forecast[:possible_units] * 100.0 / matched
      puts '%.2f%% of %s possible.' % [possible_percent, unit_type]
    end
  end
    

View on GitHub

This example will have an output similar to the following:

100 clicks matched.
2 contending line items.

Prospective LineItem

Alternatively, you can instantiate a prospective line item and forecast without persisting it. To do so, instantiate a local line item object and set it on the ProspectiveLineItem. If you set an advertiser ID, the forecast also takes unified blocking rules into account.

Java

  // Get forecast for prospective line item.
  ProspectiveLineItem prospectiveLineItem = new ProspectiveLineItem();
  prospectiveLineItem.setAdvertiserId(advertiserId);
  prospectiveLineItem.setLineItem(lineItem);
  AvailabilityForecastOptions options = new AvailabilityForecastOptions();
  options.setIncludeContendingLineItems(true);
  options.setIncludeTargetingCriteriaBreakdown(true);

  AvailabilityForecast forecast =
      forecastService.getAvailabilityForecast(prospectiveLineItem, options);
    

View on GitHub

Python

  prospective_line_item = {
      'lineItem': line_item,
      'advertiserId': advertiser_id
  }

  # Set forecasting options.
  forecast_options = {
      'includeContendingLineItems': True,
      # The field includeTargetingCriteriaBreakdown can only be set if
      # breakdowns are not manually specified.
      # 'includeTargetingCriteriaBreakdown': True,
      'breakdown': {
          'timeWindows': [
              now_datetime,
              now_datetime + datetime.timedelta(days=1),
              now_datetime + datetime.timedelta(days=2),
              now_datetime + datetime.timedelta(days=3),
              now_datetime + datetime.timedelta(days=4),
              end_datetime
          ],
          'targets': [
              {
                  # Optional, the name field is only used to identify this
                  # breakdown in the response.
                  'name': 'United States',
                  'targeting': {
                      'inventoryTargeting': {
                          'targetedAdUnits': [
                              {
                                  'includeDescendants': True,
                                  'adUnitId': root_ad_unit_id,
                              }
                          ]
                      },
                      'geoTargeting': {
                          'targetedLocations': [
                              {
                                  'id': '2840',
                                  'displayName': 'US'
                              }
                          ]
                      }
                  }
              },
              {
                  # Optional, the name field is only used to identify this
                  # breakdown in the response.
                  'name': 'Geneva',
                  'targeting': {
                      'inventoryTargeting': {
                          'targetedAdUnits': [
                              {
                                  'includeDescendants': True,
                                  'adUnitId': root_ad_unit_id,
                              }
                          ]
                      },
                      'geoTargeting': {
                          'targetedLocations': [
                              {
                                  'id': '20133',
                                  'displayName': 'Geneva'
                              }
                          ]
                      }
                  }
              }
          ]
      }
  }

  # Get forecast.
  forecast = forecast_service.getAvailabilityForecast(
      prospective_line_item, forecast_options)
    

View on GitHub

PHP

      // Get forecast for prospective line item.
      $prospectiveLineItem = new ProspectiveLineItem();
      $prospectiveLineItem->setAdvertiserId($advertiserId);
      $prospectiveLineItem->setLineItem($lineItem);
      $options = new AvailabilityForecastOptions();
      $options->setIncludeContendingLineItems(true);
      $options->setIncludeTargetingCriteriaBreakdown(true);

      $forecast = $forecastService->getAvailabilityForecast(
          $prospectiveLineItem,
          $options
      );
    

View on GitHub

C#

// Get availability forecast.
AvailabilityForecastOptions options = new AvailabilityForecastOptions()
{
    includeContendingLineItems = true,
    // Targeting criteria breakdown can only be included if breakdowns
    // are not speficied.
    includeTargetingCriteriaBreakdown = false,
    breakdown = new ForecastBreakdownOptions
    {
        timeWindows = new DateTime[] {
            lineItem.startDateTime,
            DateTimeUtilities.FromDateTime(tomorrow.AddDays(1),
                "America/New_York"),
            DateTimeUtilities.FromDateTime(tomorrow.AddDays(2),
                "America/New_York"),
            DateTimeUtilities.FromDateTime(tomorrow.AddDays(3),
                "America/New_York"),
            DateTimeUtilities.FromDateTime(tomorrow.AddDays(4),
                "America/New_York"),
            lineItem.endDateTime
        },
        targets = new ForecastBreakdownTarget[] {
            new ForecastBreakdownTarget()
            {
                // Optional name field to identify this breakdown
                // in the response.
                name = "United States",
                targeting = new Targeting()
                {
                    inventoryTargeting = new InventoryTargeting()
                    {
                        targetedAdUnits = new AdUnitTargeting[] {
                            new AdUnitTargeting()
                            {
                                adUnitId = rootAdUnitId,
                                includeDescendants = true
                            }
                        }
                    },
                    geoTargeting = new GeoTargeting()
                    {
                        targetedLocations = new Location[] {
                            new Location() { id = 2840L }
                        }
                    }
                }
            }, new ForecastBreakdownTarget()
            {
                // Optional name field to identify this breakdown
                // in the response.
                name = "Geneva",
                targeting = new Targeting()
                {
                    inventoryTargeting = new InventoryTargeting()
                    {
                        targetedAdUnits = new AdUnitTargeting[] {
                            new AdUnitTargeting()
                            {
                                adUnitId = rootAdUnitId,
                                includeDescendants = true
                            }
                        }
                    },
                    geoTargeting = new GeoTargeting()
                    {
                        targetedLocations = new Location[] {
                            new Location () { id = 20133L }
                        }
                    }
                }
            }
        }
    }
};
ProspectiveLineItem prospectiveLineItem = new ProspectiveLineItem()
{
    advertiserId = advertiserId,
    lineItem = lineItem
};
AvailabilityForecast forecast =
  forecastService.getAvailabilityForecast(prospectiveLineItem, options);
    

View on GitHub

Ruby

  prospective_line_item = {
    :advertiser_id => advertiser_id,
    :line_item => line_item
  }

  # Set forecasting options.
  forecast_options = {
    :include_contending_line_items => true,
    # The field includeTargetingCriteriaBreakdown can only be set if breakdowns
    # are not mannually specified.
    # :include_targeting_criteria_breakdown => true,
    :breakdown => {
      # Break down forecast by day from start_time to end_time
      :time_windows => time_windows,
      # Break down forecast by any targeting configuration
      :targets => [
        {
          # Optional, the name field is only used to identify this breakdown in
          # the response.
          :name => 'United States',
          :targeting => {
            :inventory_targeting => targeting[:inventory_targeting],
            :geo_targeting => {
              :targeted_locations => [
                {
                  :id => '2840',
                  :display_name => 'US'
                }
              ]
            }
          }
        },
        {
          # Optional, the name field is only used to identify this breakdown in
          # the response.
          :name => 'Geneva',
          :targeting => {
            :inventory_targeting => targeting[:inventory_targeting],
            :geo_targeting => {
              :targeted_locations => [
                {
                  :id => '20133',
                  :display_name => 'Geneva'
                }
              ]
            }
          }
        }
      ]
    }
  }

  # Get forecast for the line item.
  forecast = forecast_service.get_availability_forecast(
      prospective_line_item, forecast_options)
    

View on GitHub

Delivery forecasts

If you want to simulate the delivery of multiple competing line items, you can do so with a DeliveryForecast.

Note: A delivery forecast is not equivalent to running multiple availability forecasts. Delivery forecasts consider displacement between the forecasted line items.

Existing LineItems

You can run a delivery forecast for existing line items by using their IDs.

Java

  // Get the ForecastService.
  ForecastServiceInterface forecastService =
      adManagerServices.get(session, ForecastServiceInterface.class);

  DeliveryForecastOptions options = new DeliveryForecastOptions();

  DeliveryForecast forecast =
      forecastService.getDeliveryForecastByIds(Longs.toArray(lineItemIds), options);

  for (LineItemDeliveryForecast lineItemForecast : forecast.getLineItemDeliveryForecasts()) {
    String unitType = lineItemForecast.getUnitType().toString().toLowerCase();
    System.out.printf("Forecast for line item %d:%n", lineItemForecast.getLineItemId());
    System.out.printf("\t%d %s matched%n", lineItemForecast.getMatchedUnits(), unitType);
    System.out.printf("\t%d %s delivered%n", lineItemForecast.getDeliveredUnits(), unitType);
    System.out.printf(
        "\t%d %s predicted%n", lineItemForecast.getPredictedDeliveryUnits(), unitType);
  }
    

View on GitHub

Python

  # Initialize appropriate service.
  forecast_service = client.GetService('ForecastService', version='v202408')

  # Get forecast for line item.
  forecast = forecast_service.getDeliveryForecastByIds(
      [line_item_id1, line_item_id2], {'ignoredLineItemIds': []})

  for single_forecast in forecast['lineItemDeliveryForecasts']:
    unit_type = single_forecast['unitType']
    print('Forecast for line item %d:\n\t%d %s matched\n\t%d %s delivered\n\t'
          '%d %s predicted\n' % (
              single_forecast['lineItemId'], single_forecast['matchedUnits'],
              unit_type, single_forecast['deliveredUnits'], unit_type,
              single_forecast['predictedDeliveryUnits'], unit_type))


if __name__ == '__main__':
  # Initialize client object.
  ad_manager_client = ad_manager.AdManagerClient.LoadFromStorage()
  main(ad_manager_client, LINE_ITEM_ID_1, LINE_ITEM_ID_2)

    

View on GitHub

PHP

      $forecastService = $serviceFactory->createForecastService($session);

      // Get forecast for the line items with no options set.
      $forecast = $forecastService->getDeliveryForecastByIds(
          [$lineItemId1, $lineItemId2],
          new DeliveryForecastOptions()
      );

      // Print out forecast results.
      $lineItemDeliveryForecasts = $forecast->getLineItemDeliveryForecasts();
      foreach ($lineItemDeliveryForecasts as $lineItemForecast) {
          $unitType = strtolower($lineItemForecast->getUnitType());
          printf(
              "Forecast for line item ID %d:%s",
              $lineItemForecast->getLineItemId(),
              PHP_EOL
          );
          printf(
              "    %d %s matched%s",
              $lineItemForecast->getMatchedUnits(),
              $unitType,
              PHP_EOL
          );
          printf(
              "    %d %s delivered%s",
              $lineItemForecast->getDeliveredUnits(),
              $unitType,
              PHP_EOL
          );
          printf(
              "    %d %s predicted%s",
              $lineItemForecast->getPredictedDeliveryUnits(),
              $unitType,
              PHP_EOL
          );
      }
    

View on GitHub

C#

using (ForecastService forecastService = user.GetService<ForecastService>())
{
// Get a delivery forecast for the line items.
DeliveryForecastOptions options = new DeliveryForecastOptions();
options.ignoredLineItemIds = new long[]
{
};
DeliveryForecast forecast = forecastService.getDeliveryForecastByIds(new long[]
{
    lineItemId1,
    lineItemId2
}, options);

// Display results.
foreach (LineItemDeliveryForecast lineItemForecast in forecast
    .lineItemDeliveryForecasts)
{
    String unitType = lineItemForecast.unitType.GetType().Name.ToLower();
    Console.WriteLine("Forecast for line item {0}:",
        lineItemForecast.lineItemId);
    Console.WriteLine("\t{0} {1} matched", lineItemForecast.matchedUnits,
        unitType);
    Console.WriteLine("\t{0} {1} delivered", lineItemForecast.deliveredUnits,
        unitType);
    Console.WriteLine("\t{0} {1} predicted",
        lineItemForecast.predictedDeliveryUnits, unitType);
}

    

View on GitHub

Ruby

  # Get the ForecastService.
  forecast_service = ad_manager.service(:ForecastService, API_VERSION)
  # Get forecast for the line item.
  forecast = forecast_service.get_delivery_forecast_by_ids(
      [line_item_id1, line_item_id2], nil)

  unless forecast.nil? || forecast[:line_item_delivery_forecasts].nil?
    forecast[:line_item_delivery_forecasts].each do |single_forecast|
      # Display results.
      unit_type = single_forecast[:unit_type]
      puts ('Forecast for line item %d:\n\t%d %s matched\n\t%d %s ' +
          'delivered\n\t%d %s predicted\n') % [single_forecast[:line_item_id],
          single_forecast[:matched_units], unit_type,
          single_forecast[:delivered_units], unit_type,
          single_forecast[:predicted_delivery_units], unit_type]
    end
  end
    

View on GitHub

This example will have an output similar to the following:

Forecast for line item 14678:
    100 clicks matched
    0 clicks delivered
    98 clicks predicted

If you want to exclude any line items from the delivery simulation, you can do so by setting their IDs in the DeliveryForecastOptions.

Prospective LineItems

Similar to the availability forecasts, you can run delivery forecasts on line items that are not persisted. To do so, use ProspectiveLineItem objects in the ForecastService.getDeliveryForecast method.

Future sell-through reports

The future sell-through report predicts availability of impressions for your Google Ad Manager network so you can make sure you're maximizing revenue and sell-through rate. Future sell-through reports can be run through the ReportService.

Note: There are significant differences between sell-through reports and availability forecasts.

FAQ

I have a lot of line items for which I want to forecast availability. Can I run multiple forecasts in one request?

No. You must make a separate availability forecast request for each line item or prospective line item.

I ran a number of forecasts and now I'm getting EXCEEDED_QUOTA errors. What gives?

Forecasts are computationally expensive, and the quota system ensures the service is reliable. You can safely wait and retry any forecasts that resulted in a quota error.

What causes NO_FORECAST_YET or NOT_ENOUGH_INVENTORY errors?

Forecasts are run based on your network's traffic history. Sometimes, there's not enough historical data to run a forecast. For more details on these errors, see the ForecastError documentation.

What's an AlternativeUnitTypeForecast?

Sometimes it's useful to know what other inventory is available. For example, when forecasting for a CPC line item, the alternative unit type forecasts will include information about the number of impressions.

I have more questions about Ad Manager forecasting in general.

See if they're covered in the product FAQ, or ask in our developer forum.