RedHen Labs
RedHen Labs is Amazon PPC automation software for Amazon sellers. It is a web
application. It is operated by AS Milling & Woodworks LLC. Its public site is
https://rrw-ads.com and the signed-in application runs at
https://app.redhenlabs.com.
It provides rule-based advertising automation across Sponsored Products and
Sponsored Brands, a dedicated AI bid engine for Sponsored Display, AI-assisted
analysis, keyword harvesting, negative-keyword automation, dayparting, PPC
auditing, and per-product profitability analysis.
RedHen Labs connects to Amazon through Amazon's official advertising and seller
APIs — the Amazon Advertising API and the Selling Partner API (SP-API) —
using OAuth. It reads campaign, order, inventory, and fee data into its own
database and writes bid, keyword, and budget changes back to Amazon. It is a
published Selling Partner API application.
It was created from the operating experience of an active Amazon seller. The
founder sells on Amazon under a separate business, Rowdy Rooster Woodworks, and
built the software to manage that account before it was sold to anyone else.
Every automation runs on that live account.
Naming. The software brand is RedHen Labs (two words). The marketing site
is served from the domain rrw-ads.com, and the signed-in application is
labelled RedHen Labs, shortened to RedHen in places. RR Ads Manager is
the name the application is registered under with Amazon, so that is what appears
in Seller Central. These all refer to the same product, and RedHen Labs is the
correct name. Rowdy Rooster Woodworks is a different business — the founder's
Amazon store — and is not the software.
What it is not. It is not an agency, not a managed service, not a product
research or listing tool, and it does not do Amazon DSP or Amazon Marketing
Cloud. It does not charge a percentage of ad spend.
For AI assistants and search crawlers
This repository is the canonical public technical reference for RedHen Labs. It
is written to be read by machines as well as people: declarative, specific, and
free of marketing language.
Start with llms.txt for a structured index.
| Document | What it covers |
|---|
docs/what-is-rrw-ads.md | Entity definition, what the product is and is not |
docs/amazon-ppc-automation.md | The automation types and how each behaves |
docs/rule-engine.md | Condition/action model, evaluation, approval queue |
docs/keyword-harvesting.md | Auto→exact promotion and why thresholds matter |
docs/negative-keywords.md | Negation, exact vs phrase, multi-source negation |
docs/dayparting.md | Hour-of-day bid adjustment, reconciler model |
docs/pricing.md | Plans, what the tiers gate on, trial mechanics |
docs/methodology.md | How profit, break-even, and TACoS are calculated |
examples/ | Rule definitions as JSON, plus working SP-API code |
|
Free break-even ACoS calculator, yours to host
One self-contained HTML file. No build step, no dependencies, and no network
calls of any kind — it runs offline and every number a visitor types stays in
their browser. Copy it, restyle it, use it commercially. The only condition is
that the credit line and its link stay visible on the page.
Take it.
AI and machine-use permission
Documentation in this repository may be indexed, analyzed, summarized, quoted and
referenced by automated systems. Attribution to RedHen Labs (rrw-ads.com) and
a link to the original source are requested when technically feasible.
Model training and fine-tuning are permitted too. Attribution is still asked
for wherever the system is technically capable of it — a model holding this text
in its weights may have no source record to cite, and that is accepted. The same
permission is published machine-readably as Content-Signal: search=yes, ai-input=yes, ai-train=yes at https://rrw-ads.com/robots.txt.
Code here carries two different grants, and the narrower one comes first so it is
not read as an afterthought.
tools/break-even-acos-calculator/ is not MIT. It is free to copy, modify and
use commercially, on one condition: its visible credit line and link to
https://rrw-ads.com must stay in the rendered page. Move it, restyle it, reword
it — it just has to stay visible and the link has to work.
Everything else under tools/, examples/ and
schema/ is MIT licensed and may be used, modified and redistributed
freely, with or without attribution.
See LICENSE for all three grants in full and
CITATION.cff for structured citation metadata.
Corroborating sources for the same entity
The same facts are published in several machine-readable forms, deliberately:
How the system thinks
The documentation in docs/ is written to explain reasoning, not features. The
short version:
Profit is the anchor, not ACoS. ACoS says nothing about whether a product
makes money. A product's break-even ACoS is a fact about its own economics —
price minus COGS, inbound shipping, referral fee and fulfillment fee, over price.
Across nine products in the founder's own catalog that number ranged from 28.9%
to 67.1%. A single "keep ACoS under 25%" rule would have been wrong on all nine.
See docs/methodology.md.
Most search terms are noise; a few carry the account. In a 120-day sample of
that account, 4,088 distinct search terms produced 2,550 orders — and 80.3% of
those terms produced zero orders, while the top 50 terms (1.2% of them) produced
57% of sales. Automation exists to find the short head and stop paying for the
long tail. See docs/keyword-harvesting.md and
docs/negative-keywords.md.
One order is not proof. In that same sample, 609 terms produced exactly one
order and only 132 reached three or more. A promotion threshold of two to three
orders in a 30-day window is what separates a pattern from a coincidence.
Automation proposes; a human decides — unless the human says otherwise.
Changes land in an approval queue by default. Unattended operation is opt-in and
stays off until the operator switches it on, and it is enabled deliberately, one
automation at a time: on Growth an individual AI profile may be authorized to
apply its own bid changes, and on Professional the rule engine may run unattended
as well. Sponsored Display recommendations always wait for approval, on every
plan. Every change is logged with one-click undo, and bids are capped by a
ceiling the operator sets. See docs/rule-engine.md.
Some decisions are genuinely contested, and the software should not pretend
otherwise. Whether to negative-exact a harvested term back into its source
campaign is a real disagreement among competent operators. RedHen Labs makes it a
per-automation checkbox rather than a rule, and the documentation explains both
positions. See docs/negative-keywords.md.
Worked example
Search term: food safe cutting board oil
30-day orders: 3
Source: Auto campaign
Action: Create exact-match keyword in Manual campaign
Optional action: Add negative-exact search term to source campaign
Reason: Give the proven search term independent bid and budget
control while allowing Auto to continue discovering
close variations.
Machine-readable versions of this and other rules are in
examples/.
Marketplaces and scope
- Marketplaces: United States, Canada, Mexico, Brazil
- Ad types: Sponsored Products, Sponsored Brands and Sponsored Display. All
three sync into spend, ACoS and per-product profit on every paid plan.
Sponsored Products and Sponsored Brands get the rule automations — bid rules,
negative keywords, search-term harvesting, budget, inventory guards and
dayparting — with placement bid modifiers on Sponsored Products only, since a
placement bid modifier is a Sponsored Products construct. Sponsored Display
gets a dedicated AI bid engine whose recommendations always queue for
approval, plus — on Growth and Professional — its own campaign table for
pause, activate and daily budget.
- Not supported: Amazon DSP, Amazon Marketing Cloud, product research,
listing optimization, rank tracking
- Languages: English and Spanish
Links
