Curious about what an Adobe Target Community Q&A Coffee Break looks like? Be sure to check out our past 2020 Adobe Target Coffee Breaks, covering a WIDE variety of Adobe Target topics, and our latest thread from our 2/24/21 Adobe Target Coffee Break with @Rob_Hornick
Topics help categorize Community content and increase your ability to discover relevant content.
Views
Replies
Total Likes
I have a question about Adobe Target Auto Personalization campaigns. Is there a way to specify which parameters the algorithm uses - or a way to weight some parameters? We have added custom profile parameters using info we know about the users past history on our site, but those custom profile values do not see to be used (or at least not weighted) in our AP campaign.
@JerrodStrid wrote:I have a question about Adobe Target Auto Personalization campaigns. Is there a way to specify which parameters the algorithm uses - or a way to weight some parameters? We have added custom profile parameters using info we know about the users past history on our site, but those custom profile values do not see to be used (or at least not weighted) in our AP campaign.
There isn't a way to weight parameters that the model uses. A few issues I've seen with custom data and why it may not be showing up include:
- It isn't predictive of the visitor behavior (/ your optimization goal) the model is seeing for that specific activity
- Your custom data isn't associated with enough profiles in the activity to be predictive for the overall visitor set in the activity (in other words, it's sparse)
- There isn't enough variation in the values for it to be predictive
This article explains the process of how the model selects attributes: https://experienceleague.adobe.com/docs/target/using/activities/automated-personalization/algo-rando...
@ShannonHamiltonPM wrote:
@JerrodStrid wrote:I have a question about Adobe Target Auto Personalization campaigns. Is there a way to specify which parameters the algorithm uses - or a way to weight some parameters? We have added custom profile parameters using info we know about the users past history on our site, but those custom profile values do not see to be used (or at least not weighted) in our AP campaign.
There isn't a way to weight parameters that the model uses. A few issues I've seen with custom data and why it may not be showing up include:
- It isn't predictive of the visitor behavior (/ your optimization goal) the model is seeing for that specific activity
- Your custom data isn't associated with enough profiles in the activity to be predictive for the overall visitor set in the activity (in other words, it's sparse)
- There isn't enough variation in the values for it to be predictive
This article explains the process of how the model selects attributes: https://experienceleague.adobe.com/docs/target/using/activities/automated-personalization/algo-rando...
@JerrodStrid,
We can see what Target deems as the "Important Attributes" if we look in the important attributes report in the UI. (Please note this report will show you only the top 10 visitor attributes sorted by importance score.) The second place is from the Important Attributes Report by downloading the CSV. (This one will show you the top 100 visitor attributes.)
If your attribute is not present in either it is likely due to one of the reasons @ShannonHamiltonPM mentioned above.
In regards to how everything is put together Personalization Insights is created using an Adobe patent-pending technique called MAGIX (Model Agnostic Globally Interpretable Explanations). You can learn more about MAGIX in the paper linked below.
Here are some supporting documents:
KB article on Important Attributes Report
Paper on Model Agnostic Globally Interpretable Explanations
KB article on Random Forrest Algorithm
@JerrodStrid wrote:I have a question about Adobe Target Auto Personalization campaigns. Is there a way to specify which parameters the algorithm uses - or a way to weight some parameters? We have added custom profile parameters using info we know about the users past history on our site, but those custom profile values do not see to be used (or at least not weighted) in our AP campaign.
There isn't a way to weight parameters that the model uses. A few issues I've seen with custom data and why it may not be showing up include:
- It isn't predictive of the visitor behavior (/ your optimization goal) the model is seeing for that specific activity
- Your custom data isn't associated with enough profiles in the activity to be predictive for the overall visitor set in the activity (in other words, it's sparse)
- There isn't enough variation in the values for it to be predictive
This article explains the process of how the model selects attributes: https://experienceleague.adobe.com/docs/target/using/activities/automated-personalization/algo-rando...
That makes sense. On our last AP campaign we were running it on the homepage and allowing all users into the test. That means a large percentage were "unknown" users so they would not have had values in these custom profile parameters. We may try to run this again but limit the audience to known users.
@ShannonHamiltonPM How our Users adding Taeget to their aps
I think you might by asking about implementing Adobe Target? If so, this link might be helpful: https://experienceleague.adobe.com/docs/target/using/implement-target/client-side/at-js-implementati...
We have several implementation methods you can use depending on the channel and if you are planning on integrating with Target on the client or server-side.
@ShannonHamiltonPM : How can i exclude users from a campaign activity after x number of exposition ?
@ShannonHamiltonPM What is in the development pipeline for Adobe Target AI?
@lanar1 wrote:@ShannonHamiltonPM What is in the development pipeline for Adobe Target AI?
We have some great features we are working on! Two that I'm personally excited for are:
- A new Recommendations algorithm that looks at items a customer has added to their cart
- Enhancements to the A4T reporting we released last year for Auto Allocate and Auto Target
@ShannonHamiltonPM how can I log into the session? Any URL?
@JaninS wrote:@ShannonHamiltonPM how can I log into the session? Any URL?
Hi @JaninS the Q&A happens here in the forum! No need to log into anything else.
@ShannonHamiltonPM Question: Any plans how it will Offer Decisioning service will be integrated with Adobe Target?
@dinorastoder wrote:@ShannonHamiltonPM Question: Any plans how it will Offer Decisioning service will be integrated with Adobe Target?
Great question! We are going to be creating an exciting integration where you'll be able to use offer decisioning offers/ content within Adobe Target activities. The team is actively working on it as we speak.
@dinorastoder wrote:@ShannonHamiltonPM Question: Any plans how it will Offer Decisioning service will be integrated with Adobe Target?
You can learn more about our roadmap and product announcements on this blog post, including Target with Adobe Journey Optimizer's offer decisioning capability: https://blog.adobe.com/en/publish/2021/04/27/adobe-target-updates-at-adobe-summit-delivering-persona...
@ShannonHamiltonPM there's a limitation that Activities created using the APIs cannot be edited in AT UI. Is fixing this in your roadmap?
@Rajneesh_Gautam_ wrote:@ShannonHamiltonPM there's a limitation that Activities created using the APIs cannot be edited in AT UI. Is fixing this in your roadmap?
We are definitely aware of this limitation today and are actively looking at how it ranks against other projects we are considering. Keep watching the Adobe Target release notes for all feature updates.
@ShannonHamiltonPM when will at.js v2.5.0 be released and how will on-device decisioning shape the future of Target activities?
@davidmailey wrote:@ShannonHamiltonPM when will at.js v2.5.0 be released and how will on-device decisioning shape the future of Target activities?
The new at.js version will be released in about a week!
We are very excited about offering on-device decisioning. We see it as another option to make sure we are enabling customers to use Adobe Target however and wherever you want to. On-device decisioning provides customers who want to make a tradeoff of lower latency -- but less "fresh" information -- for specific activities they are running through Target.
@davidmailey wrote:@ShannonHamiltonPM when will at.js v2.5.0 be released and how will on-device decisioning shape the future of Target activities?
We are putting the final touches on AT.JS 2.5 and hope to release it shortly. You can stay up to date on the exact release date from our release notes page: https://experienceleague.adobe.com/docs/target/using/release-notes/release-notes.html?lang=en#releas...
On-Device offers a ton of new ultra-low latency use cases in Target like feature flagging and parameterized release rollouts - all without a call to Adobe servers. We've seen latency go down to near zero (1-5ms). Very exciting! Check out our developer docs and videos here: https://experienceleague.adobe.com/docs/target/using/implement-target/server-side/on-device-decision...
Views
Like
Replies
Views
Like
Replies
Views
Likes
Replies
Views
Likes
Replies