---
name: coppica-backfill
description: Use when the user already has marketing copy (and ideally its conversion history) and wants Coppica to learn from it, instead of starting from zero. A one-time on-ramp that warms the taxonomy.
---

# Backfill existing copy

You do not have to start cold. Import copy the user already has so it can be tracked and learned
from alongside net-new generations. Historical winners are some of the best signal Coppica can
get.

## Steps

1. **Import each piece.** Send existing copy to `coppica_extract_copy` (REST: `POST /v1/extract`).
   It analyzes the copy, extracts persuasion techniques, and with `persist=true` (the default)
   creates an `output_id` you can report outcomes against.
2. **Attach history.** For any piece with known past conversions, report those outcomes against
   its new `output_id` with `coppica_report_outcome`. This is what makes backfill valuable rather
   than cosmetic: the taxonomy starts with real evidence.
3. **Track going forward.** If the piece is still live, attach a tracking link so future
   conversions keep flowing (load `coppica-close-the-loop`).

## Decision branches

- **Large library?** Batch the imports. Report what you've covered and what's left so nothing is
  silently dropped.
- **No conversion history, just the copy?** Still worth importing for technique extraction, but
  flag to the user that the loop can't learn from it until outcomes exist.

## Verify

`coppica_extract_copy` returns an `output_id`; outcomes you report reference it.

## Next

`coppica-optimization-loop` can now include these historical pieces in its ranking.