Skip to content

Discogs Feedback - 2025-08-27

TL;DR

Feature validated as useful for "fine-tuning" forecasts, but current friction (no auto-categorization, no duplicate detection) makes it not worth the manual effort for their use case. Discogs is not cash-constrained, reducing urgency.

Overall Sentiment
Direction validated Partially
Stage shown Skateboard
Ready for next stage Needs changes
Blocker for them? No - nice to have

Source

  • Transcript: 2025-08-27-batch-upload-feedback.md
  • Date: 2025-08-27
  • Participants: Jamie McKinstry (Controller), Stephanie Burns (Senior Accounting)
  • Context: Prototype walkthrough - exported Ramp bills, attempted upload to Palm

ICP Fit Note

Discogs is not Palm's ideal customer profile. Their feedback should be weighted accordingly: - Not cash-constrained ("cash isn't too big of a concern... a week here, a week there isn't gonna make that big of a difference") - Smaller company without complex treasury needs - Less urgency around cash visibility and forecasting precision

Jamie acknowledged this themselves - a high-growth, cash-constrained company would see more value in tight forecasting.

"If I were a bigger business... where there's just a lot more change like a high growth size, then I think this would be more important because we're like... watching our cash closely" - Jamie

This Iteration: Validation

What they validated about the current design:

  • Upload flow works - Exported from Ramp, uploaded to Palm successfully
  • Mapping UI is clear - Understood the column mapping concept
  • Value proposition understood - "Fine-tuning" forecasts, not replacing ML predictions

Important Nuances

This is supplementary, not primary: - ML should handle recurring payments - Uploads are for confirming timing and one-offs - "The machine should be doing it right. But this is like, yes this payment is coming, but we're planning on paying it this week versus what you predicted."

Single-category limitation is acceptable IF auto-categorization works: - They'd manipulate one export into multiple uploads - But only worthwhile if categorization is automatic

Future Iterations: Suggestions

High Value (Blockers for adoption)

  • Automatic categorization - Read GL account/department from file, auto-assign Palm category. "If Palm could do its thing and reading my document and automatically categorizing it... it's able to do that with our other bank transactions."
  • Duplicate detection - Flag if vendor name + amount already exists (either from manual entry or ML prediction). "The vendor name... that's the most important detail to me."

Medium Value

  • Bank account dropdown - Select target account in UI instead of requiring it in file. "If I could just upload the data that I do have and then dropdown, that'd be quicker."
  • Vendor name field - Add proper vendor field to forecasts (not just free text note). Critical for duplicate detection and audit trail.

Longer Term

  • Upload from account page - Entry point that implicitly knows the bank account
  • ML + upload reconciliation - When uploading, show "this matches existing ML prediction" and allow updating amount/date

Questions Asked

  • "Would you prefer one file per account, or a dropdown on each line item?" → Dropdown preferred, or even better: auto-detect
  • "How often would you do this?" → Not frequently unless very easy; Discogs doesn't need to watch cash that closely
  • "What would you consider a duplicate?" → Vendor name is the key identifier

Raw Feedback Quotes

"I would view this upload more of like a fine-tuning the forecast... this is like yes this payment is coming, but we're planning on paying it this week versus what you probably predicted." - Jamie

"It definitely feels like if you have accurately got all of the big payments you're expecting and then you upload this file, it should be like 'Oh, there's that big payment' and match it and maybe update the amount to the actual amount." - Stephanie

"I don't know that it's worth Stephanie's time to put together the spreadsheet in the right format if we're just fine tuning like a week here, a week there" - Jamie

"In my perfect world, I feel like Palm would do its thing and reading my document and automatically categorizing it like it does with our other bank transactions." - Jamie

"A concern that I do have with it getting uploaded is potential duplication... the system would need to be able to flag that" - Jamie

"The vendor name... that's the most important detail to me. How do I know what you're adding is real if I can't tie it to a description of what it is." - Jamie

Action Items

  • [ ] Implement automatic categorization based on GL account/department
  • [ ] Add duplicate detection (vendor name + amount matching)
  • [ ] Add bank account dropdown to upload UI
  • [ ] Consider vendor name as proper field (not just free text)