Best Practices for Ranking Linked Posts Aligned with Your Brand | Clay

Best Practices for Ranking Linked Posts Aligned with Your Brand

Hi, Question, I have an overview of linked posts in the rows that I want to rank and check which one is the most aligned with my brand, so I have to make an agent or write to another table to read all the rows in 1 prompt? What would be the best practice?

16 comments

Louis V.
·
·
I think this should be already possible, just don't know how to do it
Daniela D.
·
·
Hey Louis! Thanks for reaching out. Happy to help✨. Could you share your table url? You should be able to do this with the AI integration. In the prompt field, press the / icon to map the post results array ad provide an instruction for this to be analyzed to determine which post best aligns with your brand. You'll also have to provide some context on your brand and what the model should base its analysis on.
Louis V.
·
·
yes, but if the posts are below each other
Louis V.
·
·
https://app.clay.com/workspaces/328258/workbooks/wb_M2SyE9ognNki/tables/t_qmQH3HJ2FXNA/views/gv_vmoKbjztu6Xg
Louis V.
·
·
For example like this, can I consolidate all info from different rows into 1?
Daniela D.
·
·
Hey Louis! Thanks for the reply. To confirm, are you looking to run a comparison/ranking for all comments found in the table? That is, are you looking to use AI to analyze all comments exported to the destination table in a single request, or to run the analysis per profile to determine which comment or post data best aligns with your brand?
Louis V.
·
·
yess, exactly, all different comments from the different rows should be consolidated into 1 response.
Louis V.
·
·
??
Daniela D.
·
·
Hey Louis! Apologies for the delay. Thanks for confirming. We'll review this and get back with an update.
Louis V.
·
·
thank you!
Daniela D.
·
·
Hey Louis! Thanks for your patience. It is possible to achieve a merge of multiple rows to run an analysis with AI, but the lookup row integration is limited to 100 records (max count) per cell due to the cell size limits. If the goal is to run this analysis for records that have ICPFit as checked (206 rows), you can add tags to split the records, and add 2/3 Lookup row columns in the table to run the analysis. I made a quick video showing an example here: https://www.loom.com/share/38f814aed7644eb0829a565443056dd2 Let me know if this helps!
Abdul S.
·
·
Yes, the best practice is to batch the posts into one prompt if the number is small. If you have a lot of rows, it’s better to create an agent or write them to another table and process them in chunks. That way, you avoid token limits and keep context clear.
Louis V.
·
·
Okay, thanks so much for this video! Tbh, might make a post on how you guys take care of customers on this thing alone!
Louis V.
·
·
Okay, I understand the solution, let me work on it and see how I can improve. KOL stands for Key Opinion Leader btw. So we have that workbook to scrape posts from KOL and to use it to send messages to people engaging with the kols. But now I want to indeed analyze their posts and use them as an inspiration to write our own content. Just to give you a bit of an idea of my goals
Daniela D.
·
·
Happy to help! Thanks for clarifying🙏🏾. Got it. In that case, you can use AI in the main table to analyze each post and generate content based on that. This will be the best flow if you want each post to be used as inspiration to write content. Let us know if you run into any issues.
1
Louis V.
·
·
no, all good!