Notes
This solution is only available after consultation with a Jacquard Solutions Architect.
It consists of two workflows.
These workflows are beta versions.
In this documentation, campaign ID and experiment ID are interchangeable.
Advantages of the solution
The delivery of messages is controlled by the workflow. By default, the workflow will send 5% of the audience at minute 0 and another 5% at minute 60. Then at minute 120, it splits the rest of the audience up into batch sizes that will ensure the remainder of the emails deploy in 15 minute intervals for the rest of the optimisation period.
This ensures the workflow is not held up by the delivery processing or send times.
The optimisation period is pre-configured based on total audience volume.
If the results queries fail, the failure has no effect on the delivery workflow.
Overview of workflows
1. Master control workflow
Creates lists and deliveries.
Spawns a delivery workflow for each delivery created.
Creates a single spawn of the results workflow for each Jacquard experiment ID.
2. Results workflow
One created per Jacquard experiment ID used in the Master workflow.
The link between the Jacquard experiment ID and the Results workflow created for that experiment is stored in the Script tab in the Properties of the Master workflow.
Setup
1. Import the workflow templates into Adobe.
Broadcast-DO-Master
Broadcast-DO-ExternalResults
2. Create a lists folder for Jacquard. (e.g. JacquardExperiments)
3. The following information is needed to complete the configuration of Broadcast-DO-Master template.
Jacquard security token
Jacquard account ID
Primary Key / ID of the Broadcast-DO-ExternalResults template
Primary Key of the Jacquard lists folder
Adobe Campaign datasource
4. Add the above information to the Optimization settings JavaScript activity of the Broadcast-DO-Master template.
5. (Optionally) Add Jacquard SeedAddressIDs to the ProofingCode JavaScript activity of the Broadcast-DO-Master template (see here for further info on phraseeProofing).
Execution
1. Enter the Jacquard experiment details in the Optimization settings activity.
Here you can also set phraseeProofing to true if you are using the proofing method above.
The only parameters required for the phraseeCampaign are the experiment_id and project_id. Everything else should be left as the defaults.
2. Configure the delivery activity as you normally would.
3. If using phraseeProofing, disable the Wait node for the proofing run and then re-enable after approving the proofs (and set phraseeProofing to false in Optimization settings).
4. Schedule and run the workflow.
Walkthrough
The following process happens when the workflow is executed:
One subscriber is split from the audience and sent the delivery with the subject line configured in the delivery activity. This is done so the workflow can make copies of the delivery for subsequent batches.
The workflow notifies Jacquard the experiment is about to start.
A calculation is made on how many batches are required and what size they should be.
A results workflow is created for the campaign.
A delivery is created to send to the first batch of subscribers.
The first batch of subscribers is selected from the audience and assigned subject line variants.
The workflow creates a list of the first batch of subscribers.
It starts an instance of the delivery workflow telling that instance to send the delivery to the list.
The delivery workflow created is named using the ID of the delivery and the current workflow batch numbers.
This workflow will send the top level results to Jacquard so Jacquard can decide how to optimise the next batch to maximise the subscriber response to the email.
Every 15 minutes, the workflow restarts the results workflow, requests the latest variant proportions from Jacquard and repeats steps 5-8 using the batch information determined in step 3.
Once the workflow has finished processing the whole audience, the results workflow will continue to send results to Jacquard once a day for 3* days. (*This is configurable depending on Adobe system/database performance and data availability.)