The weekly lifecycle review, when you're the only person who'd read it

You sent two emails this week. You have a vague sense that one did better than the other. Your email tool shows you a screen with some percentages on it. No third party will ever ask you about them. This is the moment most solo lifecycle work stops being measured at all — not from laziness, but because the reporting formats everyone publishes were designed for a team presenting to a stakeholder, and performing that ritual alone in a Google Doc feels as ridiculous as it is.

Write the report anyway. Write a different one. The thing worth producing at your size is a note of about five lines, appended to a single running file, that takes ten minutes to fill and is addressed to a specific reader: you, in November, trying to reconstruct what happened. Pulling the numbers is the easy half. The note is where the value is, and the reason why is a bit of arithmetic most reporting advice skips.

Say you have 2,000 reachable people and your click rate — the share of delivered emails where someone clicked something — sits around 2.5%. That's 50 clicks. Next week the same send gets 40. Written as a percentage, click rate "fell 20%", which sounds like a finding worth a meeting. Written as people, ten humans who clicked last time didn't this time. At least one of them was on a plane.

That is the trap in solo reporting. You write "clicks down 20%, likely subject-line fatigue". Eight weeks later you read your own note and treat it as evidence. You now believe a thing about your audience that never happened. The reporting produced a false memory, which is a worse outcome than not reporting at all.

So invert the purpose. The report is not a detection instrument — it can't be one, the sample won't support it. It's a record: what you sent, what the numbers were at the time and what you decided. Trends emerge from reading eight of these in a row, not from any single one. Which means the format has to be identical every week and the file has to be one file, because a log you can't read in sequence isn't a log.

You need four inputs and they take ten minutes to collect. Named quickly, because the numbers are the ingredients here, not the dish.

1. Reached and responded. Delivered count and click count for each send this week, as raw numbers rather than percentages. Both are in your email tool's per-campaign view or its CSV export. Not open rate — Apple's Mail Privacy Protection has been inflating opens with machine image-loads since 2021, at a ratio that varies by audience and that you cannot measure. It's a fine A/B proxy and a terrible number to write down weekly.

2. The safety line. Spam complaints, hard bounces and unsubscribes for the week, as counts. Google's bulk sender guidelines put the complaint line at 0.3% and the target under 0.1%,Source · GoogleEmail sender guidelinesGoogle's published sender requirements, including the 0.3% spam-rate limit and the 0.1% target.support.google.com/a/answer/81126 so at your volume this is a line you read as "zero, one, or a problem". Google Postmaster Tools is free and reports your domain reputation directly, though it needs a decent daily volume of Gmail recipients before it will show you anything at all — the Postmaster walkthrough covers setting it up and reading it. Until it has data, your own complaint and bounce counts are the proxy.

3. One outcome for the one thing you're working on. If this month is about trial-to-paid, it's conversions from the trial cohort. If it's onboarding, it's the count who hit your activation event. One number, from your product's own database, tied to the single program you're actually running. Not five outcome metrics — one, because you only have the bandwidth to move one thing at a time and a report that pretends otherwise is decoration.

4. Reachable list size, and the net change. How many people you're allowed to email, up or down on last week. This is the denominator under everything else and it's the number that explains half of what looks like performance change.

Create lifecycle-log.md somewhere you'll find it in three months. Every Friday, append one entry in exactly this shape. Here it is filled in, with invented numbers so you can see the format rather than borrow the figures:

## Week of 8 Sep 2026

SENT: Welcome email (auto, 84 new signups). Feature announce
broadcast, 1,940 delivered, Wed 10am.

NUMBERS: 1,940 delivered / 61 clicks. 0 complaints, 4 hard
bounces, 9 unsubs. 6 trial-to-paid this week. List 2,024,
up 71.

WHAT I THINK HAPPENED: No idea. Clicks sit in the same band
as the previous four weeks. The 4 hard bounces all trace to
the March import.

WHAT I CHANGED: Rewrote the welcome email CTA from "Explore
the dashboard" to "Connect your first integration". Live
from Thursday.

WATCHING: Welcome-email clicks, from next week. Needs to be
clearly above 15% of delivered for two weeks before I believe
the CTA change did anything.

Five labels, always the same, in that order. The details are yours; the shape is not negotiable, because the entire value of the log comes from being able to scan the same line position down twelve weeks.

Two rules govern what goes in. First: no cause without a threshold. Nothing gets an explanation unless it moved more than about 20% relative and more than 20 events. Below that, the entry says "no idea" or "same band as last month". Writing that down is not a failure — it's the single most valuable habit in this whole practice, because it's what stops the log filling with confident fiction. Second: every entry names one change or explicitly says none. "WHAT I CHANGED: nothing" is a legitimate and quite loud entry. Three of those in a row is a finding about you, not about the emails.

The WATCHING line is the one people drop first and it's the one doing the real work. It converts a vague intention into a falsifiable claim with a date on it, which means next week's entry has something specific to answer. Without it, every entry is a fresh start and the log never accumulates into anything.

Ten minutes of clicking around campaign screens on a Friday afternoon is exactly the kind of chore that gets skipped in week three, so hand the assembly over. Export the week's campaign CSV from your email tool, paste it into a Claude session with the last three entries of your log, and ask for the next entry in the same format. The request that matters is the constraint, not the summarisation:

Here's this week's campaign export and my last three log
entries. Write the next entry in exactly the same five-line
format.

Rules:
- Use raw counts, not percentages, in the NUMBERS line.
- Do not attribute a cause to any movement smaller than 20%
  relative or 20 events. Write "same band" instead.
- If nothing in the data supports an explanation, WHAT I
  THINK HAPPENED is "no idea". Do not fill it with a
  plausible-sounding reason.
- Leave WHAT I CHANGED blank for me to complete. You weren't
  there.

That fourth rule is the important one. A model reading a CSV can produce the numbers and can compare them to your previous entries honestly. It cannot know that you rewrote the welcome CTA on Thursday, and if you let it guess at the change line it will confabulate something reasonable. Keep the two halves separate: the machine does the counting and the comparison, you do the memory.

The third rule earns its place too. Left unconstrained, any competent model will find a narrative in noise, because a narrative is what you asked for and refusing is not usually rewarded. Telling it the size of movement that deserves an explanation is the difference between a log you can trust in November and a set of confident sentences you'll have to relearn to disbelieve.

The Orbit Lifecycle Reporting Framework skillcovers the instrumentation side — getting the outcome number out of your product reliably, so line three of the entry doesn't depend on someone remembering to run a query.

One entry tells you almost nothing, and that's the deal you're accepting when you start. Eight entries answer questions that no dashboard can: what did we change in the two weeks before the unsubscribes climbed, how long did the welcome rewrite actually take to show up, what were we doing in the month the trial conversions moved. Those are causal questions, and the only way a one-person team can answer them is by having written down the causes at the time. A metrics tool records what happened. Nothing except you records what you did.

It compounds in a second way, too. When you eventually hire someone, or bring in help, the log is the handover document you would otherwise have spent a week writing from memory — and it's more honest than the one you'd write, because it contains the weeks where the answer was "no idea" and the weeks where you changed nothing.

The call, and it is genuinely a twenty-minute job: put a recurring Friday block in the calendar, create the file and write this week's entry now from whatever you have in front of you. If this week's entry reads "SENT: nothing. WHAT I CHANGED: nothing", write it exactly like that and file it. Three of those in a row is the most useful thing this practice will ever tell you. You will only see it if it's written down.