# Seed public datasets

## cold-email-to-shopify-brands-2026.csv / cold-email-to-shopify-brands-2026-by-stage.csv

**Source:** Seed's own self-hosted cold-email outreach engine (`outreach_sends` /
`outreach_contacts` in Seed's production database), pulled 2026-09-14. Seed is a Shopify app
for creator gifting (seedinfluencers.com). These are the real, aggregate send logs from
outreach Seed ran to Shopify DTC brand founders/operators between 2026-06 and 2026-09 (partial).

**Definitions:**
- `sent` — an email with `sent_at` set (left the SMTP server; excludes queued/failed sends).
- `opened` — the tracking pixel fired (`opened_at` set). **Inflated by Apple Mail's image
  pre-fetch** — treat as a soft signal, not a trustworthy open rate.
- `clicked` — a tracked link was clicked (`clicked_at` set). Genuinely 0 across this entire
  dataset; click tracking was unused, not a real 0% CTR.
- `replied` — a human reply was matched via IMAP (`replied_at` set). Auto-replies,
  out-of-office, and unsubscribe acknowledgements are explicitly excluded before this flag is
  set — see the methodology section of the companion blog post.
- `bounced` — the send hard-bounced (`status='bounced'`), detected via a mailer-daemon /
  postmaster reply matched back to the original send.
- `unsubscribed` (monthly file) — distinct opt-out email addresses, by the month each address
  was first seen unsubscribing. Deduplicated from a noisier raw event log (some recipients'
  mail clients re-fire the same unsubscribe acknowledgement many times).
- `reply_rate` — `replied / sent` for that row.
- `stage` / `label` (by-stage file) — the chronological position of a touch in a contact's
  send sequence: 0 = the first ("cold") email, 1 = the first follow-up, 2 = the second
  follow-up, and so on. A contact stops receiving further touches the moment they reply.

**Date:** pulled and published 2026-09-14. Full methodology, caveats, and every number's `n`
are on the companion post: `/blog/cold-email-to-shopify-brands-benchmarks` on
seedinfluencers.com.

**Licence:** CC BY 4.0. Attribution: "Seed (seedinfluencers.com)". Cite or link back to the
blog post above when you use these numbers.

**No personal data.** These files contain only monthly/stage aggregates — no recipient
emails, names, or company-identifying information.

## gifting-app-reviews-coded-2026.csv

**Source:** 56 public Shopify App Store reviews of a competing influencer gifting app,
Influencer Gift Form (IGF), read in full and coded by hand-written, deterministic keyword/
phrase rules (not an AI model). Coded 2026-09-14. These are reviews of a competitor, not of
Seed; this dataset is about what merchants say they want from the category, not an endorsement
or a critique of any specific brand.

**One row per review, containing only:**
- `review_id` — sequential, 1 through 56. Not the original App Store review order.
- `category` — the merchant's product vertical, where captured (blank for 13 of 56 rows).
- `size_band` — a bucketed LinkedIn employee-headcount proxy for brand size (`1-10`, `11-50`,
  `51-200`, `201+`, or `unknown`). Present for only 11 of 56 rows; `unknown` otherwise. This is
  a company headcount figure, not a Shopify order volume or revenue figure.
- `rating` — left blank for every row. The source file carries no star-rating column.
- `word_count` — word count of the original review text (not included in this file).
- One `true`/`false` column per coded theme: `saves_time_manual_orders`,
  `address_size_collection`, `draft_order_shopify_admin`, `support_responsiveness`,
  `ease_of_setup`, `branding_custom_form`, `creator_experience`,
  `tracking_fulfilment_visibility`, `pricing_value`, `feature_requests`, `bugs_complaints`. A
  review can carry more than one `true`.

**No review text, no brand name, and no person's name appears anywhere in this file.**

**Methodology, the full codebook, the precision hand-check, and every limitation** (single-
competitor corpus, 5-star review-selection skew, sparse category/size metadata) are on the
companion post: `/blog/what-merchants-want-from-an-influencer-gifting-app` on
seedinfluencers.com. The coding script itself, reproducible end to end, is
`docs/marketing/benchmarks/igf-reviews-coding.mjs` in Seed's repo.

**Licence:** CC BY 4.0. Attribution: "Seed (seedinfluencers.com)". Cite or link back to the
blog post above when you use these numbers.