Supported features overview

Adobe Target’s server-side SDKs give developers the flexibility to choose between performance and freshness of data for decisions. In other words, if delivering the most relevant and engaging personalized content via machine learning is most important to you, a live server call should be made. But when performance is more critical, an on-device decision should be made. For on-device decisioning to work, please refer to the following list of supported features:

Activity types

The following table indicates which activity types created using the Form-based Experience Composer are supported or not supported for on-device decisioning.

Audience targeting

The following table indicates which audience rules are supported or not supported for on-device decisioning.

Geo targeting for on-device decisioning

In order to maintain near-zero latency for on-device decisioning activities with geo-based audiences, Adobe recommends you provide the geo values yourself in the call to getOffers. Do this by setting the Geo object in the Context of the request. This means your server will need a way to determine the location of each end user. For example, your server may perform an IP-to-Geo lookup, using a service you configure. Some hosting providers, such as Google Cloud, provide this functionality via custom headers in each HttpServletRequest.

const CONFIG = {
    client: "acmeclient",
    organizationId: "1234567890@AdobeOrg",
    decisioningMethod: "on-device"
};

const targetClient = TargetClient.create(CONFIG);

targetClient.getOffers({
    request: {
        context: {
            geo: {
                city: "SAN FRANCISCO",
                countryCode: "US",
                stateCode: "CA",
                latitude: 37.75,
                longitude: -122.4
            }
        },
        execute: {
            pageLoad: {}
        }
    }
})

public class TargetRequestUtils {

    public static Context getContext(HttpServletRequest request) {
        Context context = new Context()
            .geo(ipToGeoLookup(request.getRemoteAddr()))
            .channel(ChannelType.WEB)
            .timeOffsetInMinutes(330.0)
            .address(getAddress(request));
        return context;
    }

    public static Geo ipToGeoLookup(String ip) {
        GeoResult geoResult = geoLookupService.lookup(ip);
        return new Geo()
            .city(geoResult.getCity())
            .stateCode(geoResult.getStateCode())
            .countryCode(geoResult.getCountryCode());
    }

}

However, if you do not have the ability to perform IP-to-Geo lookups on your server, but you still want to perform on-device decisioning for getOffers requests that contain geo-based audiences, this is also supported. The downside of this approach is that it will use a remote IP-to-Geo lookup, which will add latency to each getOffers call. This latency should be lower than a remote getOffers call, since it hits a CDN that is located close to your server. You must only provide the ipAddress field in the Geo object in the Context of your request, in order for the SDK to retrieve the geo-location of your user’s IP address. If any other field in addition to the ipAddress is provided, the Target SDK will not fetch the geo-location metadata for resolution.

const CONFIG = {
    client: "acmeclient",
    organizationId: "1234567890@AdobeOrg",
    decisioningMethod: "on-device"
};

const targetClient = TargetClient.create(CONFIG);

targetClient.getOffers({
    request: {
        context: {
            geo: {
                ipAddress: "127.0.0.1"
            }
        },
        execute: {
            pageLoad: {}
        }
    }
})

public class TargetRequestUtils {

    public static Context getContext(HttpServletRequest request) {
        Context context = new Context()
            .geo(new Geo().ipAddress(request.getRemoteAddr()))
            .channel(ChannelType.WEB)
            .timeOffsetInMinutes(330.0)
            .address(getAddress(request));
        return context;
    }

}

Allocation method

The following table indicates which allocation methods are supported or not supported for on-device decisioning.