Ad fatigue in Meta Ads: when to refresh your creative
Keep, replace or pause? Frequency alone cannot diagnose ad fatigue. A five-row diagnostic table, and a model where cost per order doubles three different ways.
Field notes
Practical guides to acquisition, margin and retention
Guides to acquisition, retention and measurement: POAS, CAC payback, contribution margin and free-shipping thresholds worked through with numbers, plus reviews of the Loyalz source code. Hypothetical calculations are kept apart from observations.
- 17articles
- 4based on source-code review
- 7clusters
Verified implementation or data checked, with the method recorded. Sourced linked sources or explicit arithmetic. Opinion argument, no outcome numbers. Read the assumptions and limits alongside each result.
17 articles
Every article
Break-even ROAS calculator
Enter your margin, get break-even ROAS: 1 ÷ margin, so 3.33 at 30%. Then see what your current ROAS earns or costs per 1,000 of spend.
What is POAS, and how it differs from ROAS
Three illustrative campaigns: ROAS 4.8 gives POAS 0.43 at a 9% gross margin, ROAS 2.1 gives 1.30. Define the profit numerator before reading break-even.
Contribution margin for e-commerce: what to subtract
Contribution margin subtracts every cost that grows with an order. A worked 240 order: 45% gross margin, 29.6% contribution, break-even ROAS 3.4, not 2.2.
WooCommerce conversion tracking: why it breaks and how to test it
Trace test orders through checkout, event receipt and attribution. Check hooks, duplicates, consent and the limits of server-side capture.
Consent Mode v2 without losing the visits that did consent
Separate Consent Mode from first-party tracking. Test the timing of acceptance and review the consent exceptions in the plugin source.
CAC payback period: the number an LTV:CAC ratio hides
CAC payback from cumulative cohort contribution: in the worked cohort, a CAC of 180 is covered in month 5. Why CAC ÷ monthly margin only approximates it.
How to calculate customer lifetime value: 4 methods compared
Four common CLV methods run on one store's data give answers nearly three times apart. Why they disagree, which input drives it, and which to use for CAC.
Connect Meta Ads and Google Ads to Claude, with your store
Via MCP, Claude gets 11 Loyalz Insights tools and ad spend joined to store revenue per merchant and day, not per ad. Five other write tools skip the Insights log.
Should AI agents have write access to ad accounts?
Read-only tokens by default, rules that can only pause, a spend floor and an audit row per change: the limits Loyalz enforces in code, and the five places they still fail.
Loyalty programs for business: choosing one you can measure
What to budget for each loyalty program type, why member-versus-non-member spending does not isolate the effect, and a worked sample size: 3,528 customers per group.
Loyalty campaigns: how to tell if yours did anything
Plan a loyalty offer with a comparison group, explicit sample-size assumptions and a contribution calculation that includes reward costs.
Customer retention, and the three metrics people confuse for it
Retention rate, repeat purchase rate and revenue retention: define the cohort, window and denominator before comparing results.
Repeat purchase rate vs retention rate: which to use
Repeat rate and retention rate answer different questions. The formulas, the benchmark that states its sample, and why growth alone can drop your repeat rate.
Loyalty program: what it costs and what it must return
Loyalty programs are a discount with a delay. The four types, what each costs as a share of revenue, and the repeat-purchase lift needed to break even.
Customer loyalty programs: the time-to-first-reward test
Most customer loyalty programs fail one test: how long until the first reward. The formula, three worked stores, and the five decisions to make in order.
How to increase average order value without losing money
Free shipping at 220 lifts average order value to 183.60 in a worked example but cuts contribution by 5,420. Count orders that already qualify, then test.
What the dot on each card means
The labels describe the basis of an article. A source-code review checks an implementation; it does not prove client outcomes. Worked examples show arithmetic under stated assumptions.
- Verified · source checked
- Based on inspected implementation or measured data, with the scope and method stated.
- Sourced · cited
- Based on linked sources or a worked example. An illustrative result is not a client result or a forecast.
- Opinion
- Argument, explicitly not evidence. May not contain a numeric claim about outcomes at all.