---
name: "weekly-marketing-read"
description: "Use when someone asks for the weekly marketing report or why a marketing number changed, or when the Monday read comes due: fix the period, check which sources you actually have, find the biggest moves, trace each to its cause, and say plainly when the numbers cannot explain themselves."
triggers: ["weekly marketing report", "what moved last week", "why did this number change", "week over week", "marketing numbers", "read my marketing week"]
version: "1"
---
# Weekly marketing read
Use this for the weekly report, for any question about why a marketing figure
moved, and whenever the Monday read routine runs.
## Fix the period out loud
Before any number, write both windows with their dates: the week under review
and the week it is measured against. Both must be complete and equal. If an
export ends early, trim the other window to match or flag the missing days in
the headline, so a short week never passes for a down week.
## Check coverage before you compute anything
List the sources you have and the days each one covers. A source that came
back empty still appears, marked empty, so nobody reads silence as a zero.
With one source, build from it and carry the absent ones into the gaps.
## Compute the headline set from the columns you actually have
- **Where a spend column exists:** spend, impressions, clicks, CTR, cost per
click (CPC), conversions, cost per acquisition (CPA), and conversion value
plus return on ad spend (ROAS) where a value column exists.
- **Where email columns exist:** delivered, bounced, clicks, conversions or
replies, unsubscribes, spam complaints, and list growth. Opens are a
deliverability health check only.
- **Always:** content shipped against the editorial calendar and what is stuck
on it, plus pipeline linkage when the customer relationship management (CRM)
columns exist: marketing-qualified leads, sales-qualified leads,
visitor-to-lead rate.
Currency to the whole unit, percentages to one decimal place.
## Judge email against benchmarks, not vibes
Published all-industry averages from the big email providers (roughly a third
of recipients opening, a low single-digit click rate, and a fraction of a
percent unsubscribing, with wide spread by industry and drift year to year)
are a sanity check, not a target; cite the report and year if you quote one.
Mail privacy features preload images and inflate opens, so score clicks,
conversions and replies instead. Cold, warm and customer sends each get their
own target.
## Read goal against actual, by funnel stage
Each headline figure against its plan target: engagement at the top, conversion
in the middle, pipeline and revenue at the bottom. Check the four-week rolling
average before you call anything a trend. Attribution windows of 7 to 14 days
distort the most recent week of ROAS and CPA, so flag those figures as
preliminary.
## Find the biggest move, then decompose it
Put each figure beside last week's and show the change twice, as a raw
difference and as a percentage: a 40% jump on a tiny base can matter less than
a 5% slide in spend. Start from the largest move and take it apart with the
columns you hold. CPA equals CPC divided by the click-to-conversion rate, so a
CPA change came from one of those two or both; name which carried it and by
roughly how much. Then anchor it to an event you can point at: a campaign
switched on or off, a send that went out, a page that went live.
If your data cannot account for a move, say so in a sentence and name the
single input that would settle it. Seasonality and algorithm updates are not
explanations unless the numbers show them.
## Treat small samples as noise
Below roughly 30 conversions for a source in the period, CPA and conversion
rate swing on chance alone. Call the move noise and build no story on it.
## Monthly and quarterly
Roll the weeks into one read against plan: the trend per headline metric, what
compounded, and what to change. Read the quarter the same way.
## Rules
Every claim carries its number and that number's source. A suggestion to
raise or cut a budget names the metric behind it, and the named approver
decides. Ad platforms count conversions under their own attribution windows,
so never set one against another unless the windows match. A figure the
export lacks stays blank; never estimate it. Label every line FACT, INFERENCE
or UNKNOWN.
## What you hand back
The headline (the period and its biggest move), the table with last week
beside every figure, the top movers with their causes, what shipped and what
is stuck on the calendar, what to look at next, and the gaps. A quiet week gets
the headline, the table, one line saying so, and the stuck list. Save the read
dated, add this week to the metrics history, and compare against the previous
saved read. Nothing reaches a channel or an inbox until they say yes; stage it
as a draft.
## Fallbacks
- No prior week on file: this read is the baseline, and comparisons start
with the next one.
- No plan targets: skip goal against actual, report against last week and the
rolling average, and ask for targets once.