Methodology · rubric 2026-09-28
An audit you can't argue with is one whose rules you can read. Every check the software runs, every threshold it holds you to, and every formula behind a dollar figure is on this page. The same rubric runs on every account, which is the point: it can't be charmed, rushed, or paid to look the other way.
The score
Flows carry the most weight because a working set of flows is the machine; everything else is how well the machine is fed and measured.
| Pillar | Points | What earns them |
|---|---|---|
| Flows | 30 | Each of the twelve standard flows earns its points when live and sending. Live but zero recipients in 90 days earns a quarter; draft or manual earns half; missing earns none. Points are weighted by the flow's typical revenue share (checkout and post-purchase 5, welcome 5, win-back 5, browse and cart 2, and so on). |
| Campaigns | 20 | Cadence between 0.75 and 4 emails a week (6). Fewer than 20% of sends to a raw list rather than a segment (5). Unsubscribe rate at or under 0.2% per delivered message (4). At least one A/B test in 90 days (3). Smart Sending on for 80%+ of sends (2). No campaigns in 90 days scores zero. |
| List & segments | 15 | An engaged segment exists, read from its definition (5). An unengaged or sunset segment (3). A one-time-buyer segment (4). A repeat-buyer or VIP segment (3). |
| Deliverability | 15 | Recipient-weighted across every flow and campaign message in 90 days. Bounce rate at or under 0.5% (5), under 1% (2). Spam complaints at or under 0.03% (5), under 0.08% (2). Unsubscribes at or under 0.3% (5), under 0.6% (2). |
| Acquisition | 10 | A live sign-up form (4). Form submit rate at or above 3% of unique views (3), above 1.5% (1). A form under A/B test (3). |
| Measurement | 10 | A holdout or control group anywhere in the account, by segment or flow name (5). UTM tracking on every campaign (2). Placed Order arriving from a store integration rather than a custom event (3). |
Grades: A at 85 and above, B at 70, C at 55, D at 40, F below. The thresholds are deliberately strict; a B is a well-run account.
The standard flows
Each live flow's trigger metric is read from its first action. Flows that share a trigger (post-purchase and win-back both fire on Placed Order) are told apart by their names and delays. Anything that can't be classified is listed under "other" and still counted.
| Flow | Points | Recognised by | Why it's on the list |
|---|---|---|---|
| Welcome series | 5 | Added-to-list trigger, or a subscribed/form-submitted metric | The first message a new subscriber gets, and the one that decides whether they buy at all. |
| Abandoned checkout | 5 | Started Checkout metric | The highest revenue per recipient of any flow. |
| Abandoned cart | 2 | Added to Cart metric | Fires earlier than checkout and catches people who never reach it. |
| Browse abandonment | 2 | Viewed Product or Active on Site metric | The widest net you have. |
| Post-purchase | 5 | Placed, Fulfilled or Ordered Product metric | Where the second order is decided. |
| Win-back | 5 | Name: win-back, lapsed, we miss you, reactivation | Customers who bought and went quiet. |
| Sunset | 2 | Name: sunset, unengaged, re-engagement, clean-up | Protects deliverability for everyone else. |
| Back in stock | 1 | Back in Stock metric or name | Demand already captured. |
| Review request | 1 | Name or a reviews integration metric | Feeds the next customer's first order. |
| Replenishment / cross-sell | 2 | Name: replenish, reorder, refill, cross-sell, upsell | Timed to when the product runs out. |
| VIP / loyalty | 1 | Name: VIP, loyalty, tier, birthday | Your best customers should hear from you differently. |
| SMS welcome | 2 | Subscribed to SMS metric or name | Consent is expensive; a silent number wastes it. |
"Sending to nobody" means live, created more than 45 days ago, and zero recipients in the last 90 days. "Untouched" means a live flow not updated in 365 days.
The dollars
5% × buyers (12 months) × average order value. What five points of repeat rate are worth on your own buyers and your own average order. Not a forecast; a unit of measure, so that every fix below it can be read against the same yardstick.
Klaviyo's aggregate API returns unique buyers per month, not per year. Yearly buyers are estimated from the monthly series bounded by orders per buyer; the estimate is labelled on the report. One-time share assumes repeat buyers average 2.6 orders a year, which is stated wherever it's used.
started checkouts (90 days) × 365/90 × 3% × AOV. A 3% recovery rate is a conservative floor for a three-message flow. Your own holdout replaces it.
buyers × 2 points × AOV. Two points of second-order lift from a post-purchase series that fires on the customer's real clock.
buyers × one-time share × 3% × AOV. Three percent of one-and-done customers returning once.
10% × revenue (12 months) − flow revenue (12 months), at Klaviyo's attribution. The gap to a 10% flow share, which is the floor a working flow set usually clears.
Revenue shares are Klaviyo's attribution and the report says so every time. Attribution credits a message for customers who would have bought anyway; the only number that survives that objection comes from a holdout, which is why "measured against a holdout" is a check in its own right.
The placement read
Not an inbox-placement test (that needs seed addresses and a send). A model of the signals that decide where a campaign is filed, read from the account, scored as risk out of 100 and paired with the fix. What it's worth is a holdout question, never an open-rate one.
| Component | Risk, max | Read from | Points |
|---|---|---|---|
| Identity and authentication | 25 | DNS of the campaigns' from-domain | No dedicated Klaviyo sending domain +12; no DMARC +8 (p=none +4); no SPF +5; unreadable +8. |
| Who the sends go to | 25 | Campaign audiences, 90 days | 25 × the share of email campaigns sent to a whole list rather than a segment. |
| How recipients respond | 25 | Campaign report, 90 days | Spam complaints over 0.08% +15 (over 0.03% +8); unsubscribes over 0.5% +6 (over 0.3% +3); click rate under 1% +4. |
| What the subject lines say | 25 | Subject lines, 90 days | 25 × the share of subjects that read as a promotion (percent off, sale, last chance, free shipping, today only, and so on); +3 for capitals, +3 for stacked punctuation or emoji, +5 over four sends a week. |
Low under 35, medium to 59, high from 60. Three findings can follow: the shared sending reputation, DMARC missing or unenforced, and a subject-line mix that reads as promotions.
Before Black Friday
Fewer than six passing raises a finding with the order to fix them in: the list, the domain, then the flows that carry the season.
Every finding
Each one has a trigger condition, a severity, and a fix written for the person who has to do it.
Limits, stated
Klaviyo's API gives aggregates, not customer histories, so repeat rate is estimated rather than counted. Report endpoints allow a couple of pulls a minute, so very large accounts take longer and the software waits rather than skipping. A key missing a scope produces a report with that section marked unread, not a failure. Flow actions are opened for up to 80 live flows; beyond that they are counted but not opened. Every one of these appears in the report's "What couldn't be read" list when it applies.
The rubric is versioned (this one is 2026-09-28) and the version is printed on every report, so two reports a quarter apart can be compared honestly.