Decide where your next small acquisition experiment belongs

Channel Experiments

If you run a business with an agent, the next marketing decision can be specific: which channel deserves your available time and cash, what should you test, and what evidence would justify continuing?

Channel Experiments is a downloadable Skill that gives your agent a structured way to work through that decision. Supply your product facts, audience evidence, goal and resource limits. The workflow guides a reasoned comparison and a bounded experiment plan, with unknowns kept explicit.

USD 9.90 — one-time purchase of the current version. No fixed expiry. No required bundle.

Part of Marketing & Sales Workflow Skills. Each Skill can be used on its own.

What you can work toward with your agent

The concrete deliverable is a channel decision brief and experiment card. The included instructions guide your agent to prepare:

• A relevant channel comparison: alternatives, evidence of audience fit, assumptions, exclusions and reasons for the proposed choice.
• A resource plan: cash, working time and delivery capacity, including relevant fulfillment costs and reserves. Cash spending and unpriced labor remain separate; sales receipts are not automatically treated as available funding.
• A complete small-test card: the hypothesis, audience, meaningful event, measurement method, observation window, decision threshold, owner, required permissions, and stopping or fallback conditions.
• A review decision: focus, retest, repair measurement, return to research, or stop—supported by the observations actually available. “No winner yet” is a valid result.

The reference material covers selection boundaries for 19 channel categories. It helps examine different acquisition mechanisms without requiring you to execute all 19 or buy other Skills. Website pages and video can form part of a channel's discovery or delivery path; they are not excluded because your business uses a different sales model.

A hand-constructed illustration

The following is a fictional, manually written illustration. It is not a recorded Skill or Agent execution, a customer case, or a measured business result. All figures and thresholds below are invented planning assumptions.

Imagine you sell a downloadable worksheet and want to learn whether a relevant audience will complete a free sample task and voluntarily ask for product details. You have $20 total cash and three hours for one small experiment. You are not accepting orders in this illustration.

An illustrative experiment card could then say:

• Question: Will readers from one relevant placement complete the sample task and choose to request the details?
• Owner and permission: You own the experiment and its record. The draft alone does not authorize buying the placement, sending invitations or publishing the sample; those steps must be explicitly permitted before execution.
• Resources: Treat a $12 placement as an unverified estimate, reserve the other $8 for necessary cash costs, and allow 1.5 hours for preparation, one hour for handling up to six voluntary sample participants, and half an hour for review. Confirm the actual total before committing; do not assume zero labor cost.
• Event and window: Within seven days of the authorized placement, count distinct, non-test participants who complete the defined sample task and voluntarily request details. Record arrivals and each later step separately. Do not count repeated visits as different people.
• Proposed decision rule: Two qualifying participants out of the small, capacity-limited group would justify discussing a further bounded test. This is an illustrative learning threshold, not statistical proof, paid demand or a forecast.
• Change or stop: If fewer than two people qualify after a complete observation window, do not expand; first distinguish lack of relevant reach from failure to complete the sample. If people arrive but cannot complete it, inspect the task and path before increasing reach. If attribution is missing, repair the record first. If audience fit, permission or the real cost is unresolved, do not book the placement. Stop accepting additional sample participants at the agreed capacity; do not start a larger campaign automatically.

The useful output is the decision and its conditions. This illustration contains no observed participants, purchases, income or promised acquisition result.

What is in the download

You receive readable Markdown instructions and relative reference files, blank JSON input/output templates, a project-rules template, and buyer, compatibility, licence and release notes. The method version is 0.1.1-candidate, packaged as commercial revision MS01. The development-version suffix is retained for traceability.

This is a method-file download, not a hosted analysis service, an autonomous campaign operator or a subscription. The files are readable; technical secrecy is not enforced. The package does not charge per local invocation. Your chosen agent or model may charge separately under its own terms.

Some instructions and references are Chinese or mixed Chinese/English. This English page does not mean every file is English. Ask your agent to explain or produce an output in your preferred language, then review it.

Bring the facts you already have

Start with your product and audience, the decision you need to make, your target event and timeframe, any existing channel observations, and your cash, time and delivery limits. Missing historical results do not prevent exploratory planning. Missing facts that change the decision should remain explicit questions or conditions.

You do not need the source book, the rest of the collection, our company records or our accounts. Keep private business data outside the method folder and separate different customer contexts. Use your host's supported local-file or import procedure and confirm that it can read the linked references.

Suitable when—and when to choose something else

Use this Skill when you want to compare acquisition routes, design the next small experiment, or review channel evidence before assigning more resources. It can be adapted to physical B2B, digital products, self-serve software or enterprise buying processes using the facts you supply.

If the channel is settled and you only need one email rewritten, a copywriting workflow is the more direct task. A live sales negotiation, advertising-account operation, statistical audit or current-platform policy check needs the appropriate separate capability. This Skill does not contact prospects, buy ads, post content, publish, collect payments or grant those permissions. It does not guarantee that a selected channel will acquire customers.

Method basis and verification limits

This is an original channel-experiment workflow informed by selected methods in Gabriel Weinberg and Justin Mares's Traction: A Startup Guide to Getting Customers, using the researched 2014 version. It combines that research with original experiment, resource, evidence and handoff controls. It is not a complete conversion of the book, an official product, an author endorsement or a claim of exclusive access to the methods. No source book or extracted book material is included.

Prior local validation is limited to Windows Codex installation/discovery and synthetic use. Other hosts remain unverified. There is no universal installer or guarantee of compatibility with every agent. Those checks do not establish buyer download/import success, real customer use, campaign performance, revenue or profit. Review your agent's output against your actual evidence and constraints.

Purchase, licence and help

Publisher and licensor: Jason Huang, an individual. A valid USD 9.90 purchase licenses the current version to you personally or to one institution identified for the purchase. Internal use and adaptation across multiple agents and devices are permitted, with no fixed expiry. Lawful outputs may be used commercially in your own or your clients' business.

The original Skill package may not be resold, publicly redistributed, sublicensed or passed to clients. The licence covers our original implementation, not rights in the source book. It does not promise lifetime updates, a service-level agreement or business results.

Support covers one installation issue reported within 30 days of purchase, for up to 15 minutes, by email at [email protected]. Refunds follow the actual applicable Capafy download-product rules for your order; there is no additional 14-day refund promise. Agent/model costs are separate from this download.

We build practical tools to help you succeed in your work and business.

Buy Channel Experiments — USD 9.90, one-time for the current version.