1. The filter is gone, everywhere

Every one of the account's 103 flows was pulled with its full definition and searched for b2b and Shopify Tags. Zero matches remain, with no read errors. The fix was not applied only to the headline welcome flow, it cleared the whole account.

Timestamps show a single editing session on 10 September between 14:43 and 15:04 UTC, touching 27 flows of which 26 are live. That is the fix. Nothing has been edited since, so the numbers below are a clean read on one change.

2. Flow performance, four matched 9-day windows

All windows are nine days long and measured the same way, so they are directly comparable. The broken window is the nine days immediately before the fix; the post-fix window is the nine full days after it.

Swipe the table sideways to see every column.

Excludes TH | Sunset Unengaged, a one-off list-hygiene batch that sent 3,294 emails inside the broken window and none after. Leaving it in would understate the recovery in send volume, because it inflates the pre-fix baseline with mail that has nothing to do with the filter.

• Revenue per recipient is the honest headline. +78% against the broken window, and it has already passed the pre-outage July level of £0.251.
• Still 13% below last year on revenue per recipient. The gap is closing but not closed.
• Orders from flows went 24 → 73, roughly a tripling.

3. The welcome flow, which took the worst of it

Swipe the table sideways to see every column.

From 91 recipients and £53 across nine broken days to 992 recipients and £1,414 across nine days after. That is the single clearest piece of evidence that the filter, and not demand, was the problem.

It is at 42% of its pre-outage July volume. It should keep rising: the flow has multi-day delays between emails, so each day since the fix adds another cohort's second, third and fourth message. The people who subscribed during the two-month outage were skipped permanently and cannot be back-filled by the fix alone.

4. Daily welcome-flow sends

Red bars are the broken period, grey is the day of the fix itself (partly broken, partly live), blue is post-fix. The line rises every day from 10 September onward and reaches 177 on 19 September without flattening.

5. It was never only the welcome flow

The same filter sat in browse abandonment, abandoned cart, abandoned checkout, site abandonment, post-purchase, winback and the Okendo review requests. 16 flows more than doubled their send volume after the fix.

Swipe the table sideways to see every column.

Transactional flows (fulfilled order, out for delivery, order delivered) show lower volume post-fix. That tracks order and despatch volume in the period, not the filter, and those flows never carried it.

6. What this does not yet show

• Nine days is a short read. The direction is unambiguous and the mechanism is verified, but the level will keep moving while the flow refills.
• Revenue is Klaviyo-attributed against Shopify Placed Order, so it carries Klaviyo's attribution window. Treat it as directional against the store's own numbers, not as incremental profit.
• The lost cohort is still lost. Roughly two months of subscribers never received a welcome sequence. A one-off catch-up send to those profiles is the obvious next move and is not part of these numbers.
• Campaigns are unaffected either way, as expected: £3,476 before against £3,523 after. The filter only ever sat on flows.

7. What to do next

1. Back-fill the skipped subscribers. Everyone who joined between 10 July and 10 September and never got the welcome series. One segmented send, welcome offer included.
2. Re-read this in three weeks, once the welcome flow has fully refilled, for the true steady-state number.
3. Replace the wholesale exclusion properly. The original intent was to keep B2B buyers out of retail flows. Do that with a Klaviyo list or profile property that every profile actually has, never a Shopify field that only exists on profiles with a Shopify customer record.