Weekly Report KW45 · Part of the series since founding Ketovavaria
KW45 brings three new affiliate programs, a switch in how we count visitors, and a fourth social channel – plus an open question around Amazon attribution that we haven’t fully resolved yet.
Behind it is the same operation, week after week: agents work around the clock – they write code, design images, post on social media, and pull the numbers for this very report fully automatically, straight from the platforms and partners. No manual typing, no hand-built spreadsheets – the figures below come exactly as the systems deliver them.
Did the bets from KW44 pay off?
Short answer: not yet. The five bets from KW44 (badges, photo of the week, shopping list correction, reliable social tiles, watchlist/weekly plan) are only one week old. A return effect can show up at the earliest at the second or third measurement point, not after seven days. On top of that, an early-week change in how we count visitors (more on that below) makes every week-over-week comparison harder right now. The honest answer stays: open, answerable at the earliest in KW46 or KW47.
What was added this week
No new user-facing feature shipped in KW45 – the next one, the “November Kitchen” challenge, only starts on 09.11., which already belongs to KW46. Behind the scenes, plenty still moved:
Three new affiliate programs (Rewe DE, Korodrogerie DE, ACQUA REAPER DE via Awin) joined this week.
On 08.11., 105 individual product links across 63 recipe posts were switched from Amazon to REWE/Knuspr. Each link now runs through the official redirect of the relevant network (Awin for REWE, CJ for Knuspr), so commission tracking works correctly.
Early in the week (03.11.), visitor counting switched from an external tool to a self-hosted, server-side count – no more external service, no cookies, no consent banner needed.
Pinterest launched this week as a fourth social channel, alongside Facebook, Instagram and X (preliminary numbers, see below, still with a lag on the last two days).
A note on measurement
The visitor-counting switch on 03.11. means numbers before and after aren’t directly comparable. That applies to the visitor numbers and our own affiliate-link count via /go/, not to the Search Console or platform impression numbers. From KW46 on, a clean, continuous week-over-week measurement is possible again for the first time.
This week’s numbers
The 105 visitors break down as follows: 74 came in directly (bookmarks, typed URLs, apps that don’t pass a referrer), 12 each via Facebook and Google, 6 via other sources, 1 via Instagram. Visitors are down to 105 from the previous week (116) – partly due to the counting-method switch. The Google numbers from Search Console are independent of that; they were already declining back in KW44. What stands out, by contrast, is the strong growth in affiliate link clicks: 93 versus 28 the week before – driven mainly by the new affiliate programs.
Total reach across all channels came to 782 impressions (preliminary, Pinterest days 07./08.11. still processing): Google 469, Instagram 243, X 47, Pinterest 16, Facebook 7. Via social media, 12 visitors came from Facebook (previous week 14), 1 from Instagram (previous week 0), none via X and Pinterest – more visitors than the raw click numbers suggest, since server-side counting also picks up app opens and redirects without click tracking. The platforms’ own click figures (Facebook, Instagram, X: 0 link clicks each across all 16 posts of the week) diverge from our own visitor count – the platforms only count clicks on post links, while our count also picks up visits from shares, groups and other referrers.
Affiliate programs: Awin grows, Amazon numbers still unresolved
Awin itself reports 103 clicks in KW45 (own count via /go/: 93 in total across all partners) – the network count also picks up clicks via Awin’s direct redirect. Previous week: 14 clicks. One transaction came in, not yet finally confirmed (“pending”). CJ/Knuspr had no sales; our own click count via /go/ shows 12 clicks – the CJ API itself doesn’t report click numbers.
Not yet ready to publish: Amazon
We have numbers from the Amazon PartnerNet export, but only for 08.11., with an unresolved discrepancy against our own /go/ count. The conversion rate in the export also looks unusually high. Before we show these numbers, we want the attribution properly checked – that’s one of the concrete tasks for KW46.
Where we stand against the business plan
The business plan (as of 27.10.) sets a weekly target of around 70 visitors and 42 Google clicks for November, month 1 of the “medium” scenario. On visitors, we’re still above that interim target at 105 – with the caveat from the counting-method switch. On Google clicks, at 10 we’re clearly below the weekly target of 42; that matches the plan itself, though, which doesn’t expect meaningful organic Google visibility yet for a two-week-old domain. At 243, Instagram impressions are well above the weekly target of around 140. Overall: a solid interim balance, but only partially comparable to previous weeks because of the measurement break. The business plan makes decisions monthly anyway, not weekly.
What ran in the background: model usage
This week saw noticeably more agent work behind the scenes than the previous week: 3,028 requests to the local inference engine (Strata), versus 760 in KW44 – almost four times as many. Claude subscription utilization averaged 55.3% over the 7-day window (previous week 40.4%), with a peak of 98%. A total of 752 subagent runs (previous week 1,025, though on a different counting basis) – Opus handled most of the review and assessment tasks with 356 runs, Strata mostly mechanical work and deploys with 127 runs.
Claude subscription utilization over KW45: actual (7-day average) vs. target, from Grafana/Prometheus.
Outlook: what KW46 needs to show
Four concrete things we’ll check in the next report:
The challenge starts on 09.11., right at the transition to KW46. First measurement point for whether badges, watchlist and challenge together actually drive people to come back.
KW46 is the first fully cleanly measured week after the 03.11. counting-method switch. The first credible week-over-week comparison (visitors, affiliate link clicks) will only follow in KW47.
The discrepancy between the PartnerNet export and our own count needs to be resolved before we publish concrete Amazon numbers.
The five bets from KW44 will be two weeks old by then – still early, but the first point where a return trend could realistically start to show.
- Ketovavaria weekly report KW45 (Google Search Console, server-side counting, affiliate link counting, Awin/CJ Publisher API, Instagram/Facebook/X/Pinterest Insights)
- Monetization business plan ketovavaria.de, as of 27.10.2026 (target figures, section 4a)
- Model usage analysis KW45 (Claude subscription utilization, Strata inference, subagent runs)
- Weekly report KW44