What this is: the working checklist we run when we open a Klaviyo account, in the order we run it. It assumes the account has flows and asks whether they are built right; if you are evaluating an account you might hand to an agency, our pre-hiring audit guide covers that wider frame (deliverability history, attribution honesty, reporting hygiene). This one is the operator's pass: fifteen checks, roughly in descending order of how much missed revenue each typically hides.
Part 1: coverage, the five-minute census
1. The core flows exist. Welcome, abandoned checkout, abandoned cart, browse abandonment, post-purchase, winback, sunset, and for consumables replenishment and for subscriptions dunning and cancel-save. Most accounts we open are missing three or more, and the absent flow is always the cheapest fix in the audit: the replenishment flow and back-in-stock flow are the two most valuable no-shows in practice.
2. The events actually fire. Placed Order, Checkout Started, Added to Cart, Viewed Product: check the metrics tab for live event volume before auditing any flow that consumes them. A flawless browse flow on a broken onsite-tracking snippet is a parked car.
3. The transactional stack is owned or at least styled, per the owned-stack standard: branded, educational, and sending from the brand's domain.
Part 2: triggers and timing, where flows quietly misfire
4. Trigger events match intent tiers. Checkout recovery rides Checkout Started, cart nudges ride Added to Cart, and the two dedupe per the split architecture. One flow doing both jobs mis-times both audiences.
5. Consumable clocks use consumption math. Replenishment and winback thresholds derive from days-of-supply and median reorder gaps, quantity-adjusted, not a flat 90 days. The quantity-blind entry is the single most common timing bug we find.
6. Post-purchase counts from the delivered event, not order date, so education lands with the product in hand.
7. Seasonal brands season-gate. Winback, sunset, and even first-trip content check the season property before day-counting, per the seasonal architecture. The off-season winback and the 90-day sunset on an annual purchase cycle are the two classic self-inflicted suppressions.
Part 3: suppression and collision, the trust protectors
8. Purchase exits everything, at every send. The placed-order filter checked at each step, not just entry. The post-purchase pitch to someone who already bought is the most reported bug by customers themselves.
9. Subscribers are suppressed from acquisition-flavored sends: replenishment, reorder pitches, and discount offers never reach an active subscriber whose next box is already scheduled.
10. Flows do not stack on one profile. The intent ladder hands off upward (browse exits into cart, cart into checkout), frequency caps hold, and a test profile walked through the funnel receives one coherent story, not four parallel sequences.
11. The sunset pipeline exists and runs in order: replenishment, then winback, then sunset, with campaign audiences already excluding the unengaged window and the dormancy threshold set per the repeat-cycle math.
Part 4: offers and content, the margin protectors
12. The offer hierarchy holds. The welcome offer is the ceiling; no recovery, winback, or save flow beats it; escalating ladders do not exist; expiries are real. Where this fails, the account is training its list to abandon and lapse strategically, per the offer framework.
13. Recovery flows persuade before they pay. Checkout and cart sequences carry objection content (fit, taste, match, logistics per vertical) with any incentive held to the final send. A flow whose only argument is a code is leaving its full-margin recoveries unclaimed.
Part 5: measurement, the honesty layer
14. Success is judged on clicks and orders, never opens, with Apple MPP inflation acknowledged in every open-based segment and report, and flow conversion measured against published benchmarks (the 3.33% average and 7.69% top-decile cart marks being the calibration example).
15. Every flow has one owned metric (the fit-check response, the favorite-question rate, the subscription attach, the reactivation rate) reviewed on a calendar, because a flow nobody measures decays silently, and the decay is invisible in top-line revenue until the quarter it is not.
The fix order
When the audit is done, fix in this order: broken events first (nothing downstream matters until they fire), missing high-leverage flows second (back-in-stock and replenishment usually lead), suppression collisions third (they burn trust daily), timing math fourth, offers fifth, measurement last. This ordering front-loads recovered revenue and stops the daily damage before the optimization work begins.
Frequently asked questions
How long does a Klaviyo flow audit take?
The fifteen checks run in two to four hours on a typical account: the census and event checks are minutes, the collision-testing with walked test profiles is the slow middle, and the measurement review depends on how honest the existing reporting is.
What does a flow audit find most often?
Missing replenishment and back-in-stock flows, quantity-blind timing math, purchase filters checked only at entry, and offer hierarchies where a recovery discount beats the welcome offer.
What should be fixed first?
Broken event tracking, then the missing high-leverage flows, then suppression collisions. Revenue recovery front-loads and trust damage stops before fine-tuning begins.
How is this different from a full account audit?
This is the flows-only operator pass. A full account audit adds deliverability history, list hygiene, attribution settings, and campaign strategy, the frame covered in our pre-hiring audit guide.
Can I run this myself?
Yes, that is what the checklist is for, and every check links to the full build standard it tests against. If you would rather we run it, this is the audit we offer brands before the season.
Sources
- Klaviyo. Abandoned cart benchmarks.
- Klaviyo. Email marketing benchmarks.
- Mailflow Authority. Apple Mail Privacy Protection and open rate inflation.