Issue with Contact Ranking Formulas: Seeking Assistance | Clay

Clay
APP
·
·
For ranking formulas that aren't working correctly, you can iterate on your prompts and adjust the wording to get better results. The preview feature helps ensure you're getting the right data before applying the formula. For counting and ranking contacts within companies, you can use the Lookup Multiple Rows integration to scan columns for company identifiers and count matching values within the same table. You can also use conditional formulas to perform actions on specific subsets based on company parameters. For filtering by rank, use the Filter feature above your table. Select your rank column, choose "less or equal to" and enter "3" to keep only ranks 1-3. What specific formula structure are you currently using for the ranking calculation?
Sources: AI Formulas + Conditional Runs [Automated Outbound] - Clay University Hide, Re-order, Sort, Filter, Merge Columns
Hari K.
·
·
Hey Jannis! For the second part, I believe it would be more effective to do a simple passthrough that returns "True" if the ranking is equal to 1,2, or 3 and false otherwise. For the first formula, I believe the issue here is that you can't accurately passthrough the "Relevance Score" within the array because there are differing lengths of the list. Have you tried using an AI column here? I believe it would be much more effective
Jannis E.
·
·
Hari K. good idea for the second part! For the first part: No I haven't - how would you set it up in this case? Is there anything in particular I need to pay attention to as it is an array?
Hari K.
·
·
Hey Jannis - You should be able to input the entire array into the AI column and it should be able to parse it. I would pay attention to make sure to use a "Create or Modify Content" use case instead of webscraping to prevent any AI hallucination
1
Jannis E.
·
·
I just created an AI column and it seems to work well but I need to do some more testing tomorrow
Loading attachments
Jannis E.
·
·
There is only one problem I'm running into now. If I e.g. have 10 contacts who all work at the same company and all of them have the same score, all of them would be ranked #1. Thus I'd reach out to all 10 if I solve it this way:
For the second part, I believe it would be more effective to do a simple passthrough that returns "True" if the ranking is equal to 1,2, or 3 and false otherwise.
Jannis E.
·
·
Loading attachments
Jannis E.
·
·
Here another example where eg 4 contacts were ranked as 1, 3, 3 and 3 even though it does not make sense looking at the scores
Loading attachments
Jannis E.
·
·
And this one also does not make sense
Loading attachments
Jannis E.
·
·
Hope you can point me in the right direction 🙏🏼
Luca P.
·
·
Hi there! 👋
I completely understand how frustrating formula issues can be, especially when they work sometimes but not consistently. The good news is that these kinds of ranking and filtering problems are definitely solvable! Based on what you've described, here are a few things that could help:
For the Ranking Formula Issues: AI Column Approach (as your teammate suggested): ** Your teammate's suggestion to use an AI column for the ranking formula is actually spot-on! AI columns are often more reliable for complex ranking logic because they can handle edge cases and inconsistencies better than traditional formulas ** You can prompt the AI to "Rank contacts within each company by relevance score from 1 to X, where 1 is the highest relevance" Formula Troubleshooting: ** When formulas assign wrong ranks or don't apply at all, it's often due to data inconsistencies or edge cases the formula isn't handling ** Try copying your current formula code and pasting it into Claude (or even Clay's chat) with examples of where it's failing - this can help identify exactly what's going wrong For the Filtering Formula: Simple Passthrough (also great advice from your teammate): ** A passthrough formula for filtering by rank is usually much more reliable than complex conditional logic ** Something like: "If rank ≤ 3, return 'Include', otherwise return 'Exclude'" General Debugging Tips: 1. Test with a small batch first - Run your formulas on 10-20 rows to identify patterns in the failures 2. Check your data consistency - Make sure the relevance scores are in a consistent format across all companies 3. Use conditional runs - Add "only run if" conditions to prevent the formula from running on rows where it might break Since you mentioned you'll be testing more tomorrow, I'd definitely recommend starting with the AI column approach for ranking - it tends to be much more forgiving with data inconsistencies and usually produces more reliable results. Let me know how this goes and if you have any other questions
Jannis E.
·
·
Luca P. thank you for your reply! I'm iterating on the prompt now and am testing more, but would love some more guidance on how to fix the prompt and also how to solve this. PS: Hari K. also thank you again for your initial help and pointing me in the right direction. That was super helpful. Maybe you have some more insights on how to solve the open questions?
Loading attachments
Jannis E.
·
·
Hey guys, I just took the last 3-4 hours to work on this table but am stuck with the following issue.
Context
I’ve rebuilt the scoring system that assigns relevance scores to contacts based on their job titles. The scoring works well:
Contacts with scores ≥70 are relevant.
Contacts with scores <70 are not relevant.
This part is stable and accurate.
Issue
The problem starts with the next step:
I only want to rank the relevant contacts (relevant = "Relevant for Sales outreach?" checkbox is checked) who work at the same company.
They should be sorted strictly by score, highest to lowest.
If multiple contacts have the same score, that’s fine, they can just be ordered arbitrarily — but each should still get a unique rank.
Once ranked, I need to cap the selection at 3 contacts per company (the top 3).
Right now, Clay/GPT often:
Ranks non-relevant contacts.
Produces wrong or inconsistent rankings.
Sometimes outputs “not found” or skips contacts.
Here also a loom video to explain it in more detail: https://www.loom.com/share/0835781172f74cf786bfe40901887687
Amit Kallakuri
APP
·
·
Hey Jannis! Just to clarify, you want to rank the relevant contacts at the same company based on their relevancy score(as long as it is > 70)? I think the best approach is the one you took with using lookup multiple rows and then comparing values within the records. Could I see your table and the prompts you're using?
Jannis E.
·
·
Yes correct