Sew Heart Felt · 10 September 2026
A working agenda, not a report. Almost all of this is a question or something I want to go and test. Two items are decisions I've already made in the ad account, and they're marked as such.
To be clear about scope: I am not pausing the Evergreen campaign. It is the engine and it is working. I'm cleaning up inside it.
Being straight about the upside: Meta's algorithm already ignores most of these, so this doesn't free up budget. It's housekeeping, and it stops a 2023 ad waking up and taking delivery at a bad moment in November.
Your team's gift box test has a clear winner after eight days, and it's beating the Evergreen average. I want to promote it before the Q4 ramp rather than leave it in a £50/day test.
| Ad | Spend | Purchases | Return |
|---|---|---|---|
| Gift Box Reel V2 | £210 | 18 | 2.92x |
| Gift Box Carousel | £38 | 5 | 3.46x |
| Gift Box Reel | £69 | 2 | 0.79x |
| Gift Box Carousel V2 | £32 | 0 | 0.00x |
Reel V2 goes across. I'm taking the Carousel too rather than killing it on five sales, because it has the cheapest reach cost in the set and it's a different format.
The test ad set is named "New Customers Only" but it has no audience exclusions set on it at all, so it's been showing to recent buyers too. Every other test ad set excludes them. I'll fix it as part of the move.
We found what broke the emails, and I'd rather show you than describe it.
On 10 July a single bulk edit added the same rule to 26 of your 38 live flows: only send if the person's Shopify Tags do not contain "b2b". The catch is that a subscriber who has never ordered has no Shopify Tags field at all, so they fail the test and never enter the flow. That's precisely the audience these flows exist to convert.
15 out of 15 people who did get a welcome email since 20 July have a Shopify Tags property. 15 out of 15 who joined the list in late August and got nothing don't. All consented, none suppressed. Deliverability, templates, DNS and demand were all checked and ruled out. Campaigns over the same window were up 18%.
The list is being burned while this runs. Around 5,200 people joined during the outage and only 393 got a welcome, and the unengaged sunset flow has gone from 1,426 sends last August to 6,541 this August. Those people are being written off before they were ever spoken to.
The fix is small. Either allow the rule to pass when the field isn't set, or drop the property test and exclude the B2B and Trade lists instead. The real question is who makes the change, and whether the previous agency still has edit access, because I don't want to protect Q4 in an account anyone can bulk-edit.
Also worth your view: there are 103 flows in there, 38 live and 63 drafts, and eight of the live ones are duplicate pairs competing for the same trigger. Is there an appetite to consolidate to one flow per trigger before Q4?
Right now nothing syncs. Every customer audience in the ad account is a manual spreadsheet upload, and the newest one is from January.
| Audience in Meta | Size | Last updated |
|---|---|---|
| All Customers | 53,400 | 12 January 2026 |
| Opted In Email Marketing List | 39,900 | October 2024 |
| Shopify Customer List.csv | 15,800 | November 2023 |
| Website Subscribers | 2,300 | November 2020, from Mailchimp |
Two things follow from that. The newest list holds 53,400 people against 119,017 actual customers, so it was already missing half the base the day it went up. And because that January list is what we use to exclude existing customers from tests, every "new customer" test since January has been running against eight months of people who had already bought.
Klaviyo has a native Meta integration that pushes a segment and keeps it live. It's a connection on your side. Are you happy for us to set that up?
I went back over 286 email sends across 24 months to see which offers actually moved money. Some of it surprised me and I'd like your read before I suggest changing anything.
| What was offered | Revenue per recipient |
|---|---|
| Black Friday | +85% |
| 20% off | +48% |
| Christmas occasion | +42% |
| New product launches with no discount at all | +16% |
| Gift with purchase, sent to the whole list | −32% |
| Free shipping | −12% |
| Small calendar days, e.g. National Dog Day | −27% |
Same list, same idea, two weeks apart. "Find your dog" with no discount earned £0.035 per recipient. "10% OFF International Dog Day" earned £0.019. The hook was never the problem. Launches with no discount are the single biggest earning category we have, at zero cost to margin.
Which makes me want to ask about three things in your 2026 plan, not tell you they're wrong:
The real question: if the 20% sitewide is already locked for Black Friday, I wouldn't argue with it, it's the proven play. But October is the only window left to test the mechanic underneath it. Is there anything you'd be willing to put in that slot?
This is the chart I most wanted to show you. Every bar is real orders, UK only, over twelve months.
Share of UK orders by basket value
Swipe the chart sideways to see it all.
61.3% of all orders are a single Tiny Animal. Orders pile up at £35–40, which is two animals, then collapse by three quarters in the very next band. The free shipping threshold sits just past that collapse.
Two animals at £19.99 comes to £39.98. It cannot reach £45 no matter what the customer does, so she'd need a third item or an add-on, and there isn't much sitting in that gap to add.
Dropping to £39.95 gives free shipping to roughly 3,542 orders a year that already spend £35–40. At £3.96 a parcel that's about £14,000 a year given away. To break even we'd need to pull about 5% of single-item orders up one rung. That's why I want to run it as a proper test in October with a clean before and after, rather than just switch it on.
Two things I need from you first: what do the new bag tags retail at, and what's the price ladder on the Christmas decorations? If there's something good in the £5–10 range, that's the bridge between two animals and the threshold, and it changes the whole answer.
I can see the Zipify upsell firing in the order data, but only the accepted ones, so I can't tell whether the offer is weak or whether it just isn't being shown very often. That distinction changes the fix completely.
The pattern that stands out: acceptance drops when volume peaks. January runs at 2.99%, but February falls to 1.28% and March to 1.36%, and those are your two biggest months of the year. My guess is that a heart-themed display dome is a brilliant Valentine's and Mother's Day offer being shown to a birthday or Christmas buyer the rest of the time, but that's a guess until I can see the rules.
What I'd like: the funnel rules, and the app's own views versus conversions. Also, what are the best new products to point it at? The bag tags and the Christmas decorations both look like natural candidates.
I want your view before I form one, because the data pulls in two directions and I don't want to push you at the wrong problem.
So on the raw numbers, order value looks like the bigger lever than repeat rate. But I'd genuinely rather hear what you've tried, because two things in the data suggest retention has never had a fair run:
Repeat rate depends heavily on what somebody first bought for. Halloween buyers come back at 17.3% and Christmas buyers at 15.2%, with a £45 first order. Mother's Day buyers come back at 6.6% on a £29 first order. March 2026 brought in 11,265 customers, your biggest month, repeating at 4.6%. It isn't a failure, it just means March's job is volume and Q4's job is value.
Also: the Collectors Club link in the site navigation currently 404s. Is that mid-build?
One wedding ad has carried that whole ad set since April, and it's tiring. I'd rather refresh it than pause it, because the product works.
The same wedding ad, April to September
Swipe the chart sideways to see it all.
Click-through has fallen 56% from its June peak while frequency has nearly doubled. Over its life it has reached 470,181 people and been seen 3.3 times each. That's saturation, not a copy problem.
Return has slipped from 2.81x in May to 1.92x this month, and it's still taking around £2,500 a month. The good news is your own plan solves it: the new wedding collection lands on 12 October with seven new designs behind it. Could we get stills and a short video of those as they arrive, rather than commissioning a separate shoot?
Worth saying the timing is natural anyway. Bride and Groom Mice was your number one product every month from April to August, so its season is ending regardless. This is handing over to Q4, not a rescue.
This is the one deadline on the page. Working backwards from your own promotions plan, the concept sign-off is what everything else waits on.
Assets have to exist by late October to make that. Which means concepts need signing off now.
I've got the autumn and Christmas concept sets ready for you to mark up, and the autumn ad copy ready to send across separately. What's the quickest way to get your yes or no on these, and does anyone else need to see them first?
Across last autumn and the last 90 days, the split is consistent: naming one person and one thing they love beats describing the range. Sandy the Rock Pooling Mouse returned 9.34x, the Scotland Footballer got a 4.40% click-through. The Hobby Collection catalogue ads sat at 1.4–1.9x with click-through under 0.9%. Introducing a new character isn't itself a reason to buy, but naming who it's for is.
Fair warning on that 9.34x: it's from £284 of spend. Directional, not proven.
Not for this call, but it's the next thing I want to dig into, and I think it might be the biggest one on the page.
Every single active Evergreen ad currently sends people to a collection page, either Best Sellers or Tiny Animals, used more or less interchangeably. There isn't a purpose-built landing page anywhere in the account.
We pulled customer journeys for the first time last week, and they say something that sits awkwardly with that:
She isn't browsing and she isn't comparing. The animal she lands on is very close to the animal she buys. And yet we're dropping her onto a grid of a hundred animals and asking her to browse.
What I'd like to do: a proper conversion review of the site, and a plan of A/B tests off the back of it, including whether a purpose-built landing page beats the collection page. Are you running anything on the site at the moment I should know about first?
Prepared by Kova Commerce, 10 September 2026. Every figure here is from the live ad account, Klaviyo, or Shopify order data. Private page, not indexed.