Expand my Community achievements bar.

We are excited to introduce our latest innovation to enhance the Adobe Campaign user experience — the Adobe Campaign v8 Web User Interface!
SOLVED

Predictive Subject Line - Subject Line Training

Avatar

Level 6

I have had the subject line training workflow running for about a year now.  It has trained on 883 records and I am still seeing the message "The existing dataset is not enough to generate a model."  I do get an error in the workflow every time it runs "XML-110009 Unable to find the element 'runwf' of path '/' (document with schema 'serverConf')". 

Has anyone got this to working with workflow training running over your emails, not bring in a model?

John

1 Accepted Solution

Avatar

Correct answer by
Employee Advisor

Hi John,

Please find inputs that we received from Engineering for a similar issue reported by another customer.

Predictive subject line is a machine learning process where-in we take the subject lines, with their open rates, that you have uploaded and try to come up with predictions on the future subject line that you are going to be using.

For this, Campaign takes the subject lines as training data and tries to come up with a significant enough model to predict open rate with at least 95% confidence.

The 400 subject line number is mentioned with "at least" and is only indicative, the model generation will also depend on the quality of these subject lines.

It may happen that even 10k subject lines don't lead to a model generation if they are too similar.

It means that your existing set of 883 subject records is not enough for our predictive subject module to give a high confidence suggestion. Hence the error/warning message.

I would suggest to review your data that you are uploading as the similarity of the subject lines might be the issue here.

Regards,

Vipul

View solution in original post

6 Replies

Avatar

Correct answer by
Employee Advisor

Hi John,

Please find inputs that we received from Engineering for a similar issue reported by another customer.

Predictive subject line is a machine learning process where-in we take the subject lines, with their open rates, that you have uploaded and try to come up with predictions on the future subject line that you are going to be using.

For this, Campaign takes the subject lines as training data and tries to come up with a significant enough model to predict open rate with at least 95% confidence.

The 400 subject line number is mentioned with "at least" and is only indicative, the model generation will also depend on the quality of these subject lines.

It may happen that even 10k subject lines don't lead to a model generation if they are too similar.

It means that your existing set of 883 subject records is not enough for our predictive subject module to give a high confidence suggestion. Hence the error/warning message.

I would suggest to review your data that you are uploading as the similarity of the subject lines might be the issue here.

Regards,

Vipul

Avatar

Level 2

Hi Vipul

Thank you for the answer regarding John's question. I have the same issue as John and it make very good sense that the data for the ML process needs to be significant in order to make a good model for predictions.

But what does the following error mean? :

XML-110009 Unable to find the element 'runwf' of path '/' (document with schema 'serverConf')

I just want to make sure that this error doesn't affect the workflow.

Thank you in advance

Best regards

Sebastian

Avatar

Level 2

Hi Vipul

Thank you for the answer regarding John's question. I have the same issue as John and it make very good sense that the data for the ML process needs to be significant in order to make a good model for predictions.

But what does the following error mean? :

XML-110009 Unable to find the element 'runwf' of path '/' (document with schema 'serverConf')

I just want to make sure that this error doesn't affect the workflow.

Thank you in advance

Best regards

Sebastian

Avatar

Level 6

Hi John,

We have been using ACS for over 12 months now and the 2176 records still aren't enough to train a model apparently, so don't expect this feature to start working anytime soon!

Avatar

Level 6

Wow,  that is a lot.  Are you also getting an error message in your workflow like Vipul mentioned?  I am also getting that message and I want to be assured that its not causing an issue.

"XML-110009 Unable to find the element 'runwf' of path '/' "

Avatar

Level 6

Yes I also get the same error when the workflow tries to run.

1684737_pastedImage_1.png