The problem this solves
Launch retros are where marketing's memory gets written, and most of them are written from vibes. The launch felt big, the channel was loud, a screenshot of a traffic spike circulates, and the record hardens into we crushed it before anyone separates the launch's effect from seasonality, an unrelated ranking shift, or a campaign that happened to overlap. The next launch then gets planned against a memory instead of a measurement, and the errors compound quarter over quarter.
Separating signal from noise after a launch is real analytical work across four surfaces that rarely get read together: organic movement, paid performance, how AI systems picked up the story, and what actually reached the pipeline. Each has its own baseline and its own lag. Nobody staffs that analysis in the week after a launch, because that week belongs to the next thing, so the honest retro loses to the celebratory one by default.
A reconstruction with lineage settles it. Every claimed effect is measured against its baseline, every figure traces to the pull that produced it, and the write-up says plainly what moved, what did not, and what cannot be attributed either way. Posted where the team talks and archived where the team plans, it becomes the version of events the next launch is built on.
How the mission runs
- Establish the baselines. Before measuring the launch, the agent establishes what normal looked like: pre-launch trends from Search Console, typical paid performance, the AI-visibility baseline, and the pipeline's ordinary rhythm in HubSpot. Every impact claim that follows is a comparison against these baselines rather than against zero.
- Measure the organic and paid response. Search Console shows what the launch did to queries, clicks, and positions against trend; Ad Libraries adds the competitive advertising context, including whether rivals responded during the window. Movements that track pre-existing trend lines get attributed to the trend, which is where retro honesty usually dies first.
- Check what the AI layer absorbed. AI Visibility measures whether the launch changed how AI-generated answers describe and cite you: new mentions, changed framing, or silence. This surface moves on its own clock, and knowing whether the story landed there is increasingly part of what a launch was for.
- Trace the pipeline effect. HubSpot shows what reached revenue: contacts, deals, and pipeline touched during the window, compared against the ordinary rhythm and tagged by plausible launch connection. Where attribution is genuinely uncertain, the retro says so, because an honest unknown beats a confident fiction in the record.
- Post the retro and archive the analysis. A concise verdict posts to #launches in Slack: what moved, what did not, what remains unattributable, and the two or three lessons with evidence. The full analysis, baselines and all, lands in Notion as the permanent record the next launch plan will actually cite.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
An honest launch record in two forms: a Slack post in #launches with the verdict - what moved against baseline, what did not, what cannot be attributed - and the lessons that follow, plus the full analysis archived in Notion with baselines, per-surface measurements, and every figure traceable to its source pull. The launch enters institutional memory as a measurement, and the next launch plan inherits evidence instead of folklore.
Make it yours
- Run it twice, at two weeks and at six, since organic and AI-layer effects lag paid ones; the second pass catches what the first was too early to see.
- Add a costs-against-outcomes section when the launch had dedicated spend, so the retro doubles as the budget justification for the next one.
- Standardize it as the closing step of every launch, with the Notion archive accumulating into a comparable launch library the team can actually learn from.
Frequently asked questions
What stops this from being a celebratory retro with charts?
The baseline discipline. Every effect is measured against the pre-launch trend rather than against zero, movements that track existing trends get attributed to the trend, and unattributable changes are labeled as exactly that. The structure makes inflation visible, which is why the honest version survives review.
When should the retro run?
Two weeks after launch is the earliest sensible single pass; paid effects are readable by then while organic and AI-layer movement is just emerging. The two-pass variation, at two and six weeks, fits launches where the organic story matters most, with the Notion page updated in place.
Can the analysis be challenged after the fact?
That is what the Notion archive is for. Baselines, definitions, and every figure's source pull are preserved with the write-up, so a disputed claim gets checked against its evidence in minutes. Retro arguments become factual rather than rhetorical, which is the entire cultural point.