Urumi.AI
One business. Several moving parts.
Valentte’s sales had fallen through the summer. Paid acquisition had been cut, a store migration had taken place, and a new theme was about to launch. Email, repeat purchases and social traffic were all part of the picture. Looking at any one dashboard in isolation could tell the wrong story.
Urumi brought WooCommerce order data together with Bloomreach engagement data to understand what was changing, why it mattered, and where the team should look next. The 29 September review traced the decline to reductions in paid acquisition. The October follow-up tested that explanation against the recovery.
Was the storefront losing customers, or were fewer new customers reaching it?
Put the changes in order.
- 9 September
The store migrates
A GeneratePress child theme goes live as part of the migration.
- 29 September
Follow the acquisition signal
Urumi’s review with Valentte traces the sales decline to cuts in paid acquisition.
- 30 September
A new theme and a new recovery flow
The Valentte theme replaces GeneratePress. Luke creates an abandoned-checkout email and SMS flow in Bloomreach.
- 8–11 October
New creatives start converting
35 new Meta ads bring in their first orders across four days. Meta orders and first-time customer orders rise.
- 9–12 October
Connect the rest of the activity
A three-day Direct traffic spike overlaps with social activity. Welcome, post-purchase and abandonment flows are edited, and a Black Friday direct-mail draft is created.
Recovery in a week.
A clearer view of what drove it.
Sales in the week of 5–11 October were up 10% on the previous week, and 24% above the low in the week of 14 September. The year-on-year gap narrowed to 3%, from 14% in the week of 28 September.
The improvement was spread across channels. Comparing two-week averages before and after the review made that contribution visible.
About 18% more per week, indexed to the two weeks before the review. Sales include VAT and shipping. The review was on 29 September; these periods show association, not a causal estimate.
Rounded estimates; other sources were approximately flat. Direct includes some traffic from Facebook and Instagram’s in-app browsers.
Direct contributed the largest increase, but “Direct” did not mean the absence of social influence. During 9–11 October, around 40% of Direct orders came through Facebook or Instagram’s in-app browser. Tagged Meta orders still carried their tracking identifiers, pointing away from broken ad tags as the explanation.
This gave the team specific follow-ups: identify the social activity behind the bundle-offer traffic, and check tracking on the new email and SMS cart-recovery links.
Meta and Google paid sales were still well below spring levels. The direction had changed; the gap had not disappeared.
A theme change,
without a conversion setback.
A sales increase after a theme launch can look like evidence that the theme caused it. Urumi checked conversion, orders and tracked sessions together before drawing that conclusion.
The new theme converted at approximately 6.3% over its first nine days, similar to the preceding fortnight on GeneratePress. Statistical testing found no new conversion break at the 30 September launch.
The last two periods overlap. Orders are aligned to Bloomreach’s US Pacific day. Rates reflect changes in traffic mix as well as storefront behaviour.
The earlier increases in conversion aligned with reductions in sessions after paid acquisition cuts. Fewer low-converting paid visitors meant a higher overall rate, even while order volumes fell. Traffic mix explained more than a simple before-and-after theme comparison could.
The September migration also left a measurement question to verify: did session tracking change? The review recommended a cross-check against bot-filtered Cloudflare requests.
Today’s acquisition.
Tomorrow’s repeat purchases.
The analysis followed new customers beyond the first order. Paid traffic feeds email sign-ups; those subscribers can become buyers weeks later. Cutting acquisition can therefore affect revenue long after the ad budget changes.
Week of 5 October, up from 4.2k the week before and roughly 3k per week since August.
The second week after the review, following 756 the week before. The figure had fallen to 624 by the September review.
Meta orders were already rising before the review. By 8–14 October they averaged around 114 per day, versus 81 during 17–28 September. The burst of new creatives coincided with stronger acquisition, although budget data was still needed to assess efficiency.
Hover, tap or focus a point to inspect it. First-time buyers rose from 1,863 in the week of 21 September to 2,547 in the week of 5 October.
View chart data
| Week starting | First-time buyers |
|---|---|
| 1 Jun 2026 | 3,705 |
| 8 Jun 2026 | 5,080 |
| 15 Jun 2026 | 3,926 |
| 22 Jun 2026 | 3,554 |
| 29 Jun 2026 | 3,316 |
| 6 Jul 2026 | 1,884 |
| 13 Jul 2026 | 2,014 |
| 20 Jul 2026 | 2,665 |
| 27 Jul 2026 | 2,811 |
| 3 Aug 2026 | 2,889 |
| 10 Aug 2026 | 2,125 |
| 17 Aug 2026 | 1,916 |
| 24 Aug 2026 | 2,185 |
| 31 Aug 2026 | 2,076 |
| 7 Sep 2026 | 1,764 |
| 14 Sep 2026 | 1,914 |
| 21 Sep 2026 | 1,863 |
| 28 Sep 2026 | 2,161 |
| 5 Oct 2026 | 2,547 |
Repeat-purchase behaviour was relatively steady: around 7–9% of first-time buyers returned within four weeks, and 19–23% within twelve. The bigger issue was the size of the incoming cohorts, which had shrunk from roughly 3,500–5,000 new buyers a week in June to around 1,800–2,100 in September.
The follow-up made the priorities more concrete: rebuild the acquisition pipeline, measure the reach of the first-to-second-purchase flow, and evaluate mature cohorts before judging the newest ones.
A quiet issue,
found before it became background noise.
The same review surfaced an operational issue in the checkout lifecycle. Since the September migration, unpaid attempts were appearing as pending rather than failed. From 6 October, the automatic cancellation of unpaid orders stopped.
Pending orders more than one day old at the time of the review.
- 01Separate attempts from payments
Pending orders had no transaction ID. They were unpaid attempts, not confirmed payments stuck in processing.
- 02Check the cancellation process
Review the scheduled job and hold-stock behaviour that previously cleared unpaid orders.
- 03Reconcile the backlog
Confirm payment status before clearing stale orders and releasing any reserved stock.
These were findings and recommended actions in the October review. The report does not confirm that the issue had been resolved.
That is the value of connecting the business and operational context: a review of revenue also exposes work that affects the team, inventory and the quality of the underlying data.
From more dashboards
to a more useful next step.
For Valentte, the outcome of the review was a joined-up explanation. Acquisition changes, email activity, storefront performance and checkout operations could be understood as parts of the same business.
Explain the movement
Separate channel growth, traffic mix and site changes instead of attributing every change in sales to the latest release.
See the delayed effects
Connect today’s acquisition to the subscriber and repeat-customer pipeline it creates.
Know what to do next
Turn unclear trends into specific checks on ad spend, campaign tracking, lifecycle flows and unpaid orders.
That week is an observed business result during this period. Urumi’s contribution in this case is the connected analysis that helped the team understand the recovery and prioritise its next decisions.
Book a demoThe detail behind the story.
What period does this case study cover?
The October follow-up prepared for Valentte and Urumi’s 20 October 2026 review. Most underlying data covers 6 April–14 October 2026. The latest complete sales week is 5–11 October; weeks begin Monday in UK time.
How are sales and conversion measured?
Sales use WooCommerce paid order totals, including VAT and shipping, after discounts. Fully refunded orders are excluded; partial refunds, approximately 0.1%, are not deducted. Conversion divides paid orders by Bloomreach tracked sessions, with order timestamps aligned to Bloomreach’s US Pacific day. These gross sales totals are not directly comparable with the earlier review’s net totals.
Did Urumi or the new theme cause the 24% recovery?
The report does not establish that. Paid acquisition, creative activity, social traffic, email flows and the theme changed around the same period. The 24% figure compares the latest full week with the September low. The analysis found no distinct conversion break at the new theme’s launch.
How should the cohort data be read?
Customers are matched by email, including guest checkouts, which can slightly understate repeat purchasing. Recent cohorts have had less time to convert or buy again. The 741 and 756 email-buyer figures refer to subscribers first seen within the preceding 90 days, not the entire email customer base.
Where does the data come from?
WooCommerce orders via Urumi Analytics and Bloomreach Engagement events. Charts on this page reproduce selected values from the Valentte October Review. Rounded estimates are labelled, and pending-order findings are presented as follow-up work rather than completed fixes.
