Domain
The business, what we measure, and the language we use.
Business
Alaskan Fisherman sells premium fish/seafood products in Germany via a direct-to-consumer web checkout, including a subscription/recurring-order component. The active product surface is the new React frontend (AlaskanFishermanFrontend/).
Acquisition is driven by paid Meta advertising. Other sources (Google, direct, organic) exist but Meta is the primary spend channel and the focus of optimization.
DE acquisition mix currently includes a creator-economy partnership: paid-per-order with an influencer at roughly 500k followers. Order counts and CPA in periods touched by the partnership reflect both paid Meta and creator-driven orders — segment by source before drawing organic-vs-paid conclusions.
Goal
Make German paid acquisition economically viable by improving conversion economics on the new web.
Concrete macro targets:
- Paid conversion rate ladder: minimum >0.60%, target >2.00%, ambitious >4.00%
- Reduce CPA from a baseline around 220 EUR toward 100 EUR
Two operational sub-goals at the funnel-step level:
- Stabilize conversion rates where they are currently volatile day-to-day
- Grow conversion rates at funnel steps where pass-through is currently low
The goal is reached through faster diagnosis, hypothesis generation, and page/creative/offer iteration — not by a fixed list of tactics. See method.md for the workflow.
Primary KPIs
Conversion rate
paid orders / unique visitors
The primary business KPI. Use paid orders for final reporting because paid orders represent completed business value.
CPA
marketing spend / paid orders
Acquisition cost per paying customer. Calculate against paid orders when possible.
Supporting metrics
| Metric | Formula | Use |
|---|---|---|
| Created orders | orders initiated | Fast operational reads when payment lag distorts paid totals |
| Paid orders | orders confirmed paid | Final CR / CPA |
| Unique visitors | distinct visitors in window | Denominator for CR; source depends on the question (Plausible, GA4, PostHog, or Data Master) |
| Cost per unique visitor | marketing spend / unique visitors | Acquisition efficiency signal, not the goal |
| CPM | cost per 1,000 impressions | Targeting health heuristic |
CPM heuristic: roughly 17 EUR is healthy, below ~10 EUR may signal broader or lower-quality traffic. Heuristic, not a hard rule.
Funnel pass-through targets
Repeated targets across the source material:
Choose your fish → product / order intent: ~40% desired pass-through- Unpaid order rate: target <20%; >30% is critical
Specific event names and the full funnel definition live in semantic.md.
Glossary
Plain-language terms for stakeholders and LLM context.
Conversion rate — How many visitors became paying customers. Formula: paid orders / unique visitors.
CPA — Cost to acquire one paying customer. Formula: marketing spend / paid orders.
CPM — Cost per thousand ad impressions. Heuristic for targeting quality.
Created order — An order that was started or created but may not yet be paid. Useful for fast operational reads.
Paid order — An order that has been paid. Final basis for CR and CPA.
Funnel — The sequence from visit to purchase. Each step shows how many users continue and where they drop off.
Drop-off — Users who reach one step but do not continue to the next.
Product intent — Evidence that a visitor moved beyond the landing page and looked at or selected a product.
Meta — The Facebook + Instagram advertising ecosystem. In analytics, FB and Instagram clicks may collapse into a single “Meta” source depending on referrer, in-app browser, UTM, and privacy behaviour.
Creative — The ad content a user saw before clicking: video, image, UGC, copy, hook, or concept.
UTM — URL parameters identifying source, campaign, ad, or creative.
PostHog replay — A recorded user session showing scrolling, clicking, hesitation, and abandonment.
Clarity — Microsoft UX analytics. Useful for aggregate friction signals (rage clicks, dead clicks); not a complete user-journey source.
Hypothesis — A testable explanation for a conversion problem. Example: “Facebook mobile users leave because the landing page does not match the promise made by the creative.”
Test — A controlled change made to learn whether a hypothesis is true.