Reporting upwards: dashboards a boss actually reads

Invisible work gets cut

Your work is invisible until it is reported — and over-reported work looks like hiding. The person who approves your budget does not want fourteen charts. They want three numbers, a trend, and what you're doing about it. Report well and marketing conversations end in decisions; report badly and they end in shrugs, and the budget shrinks by shrug.

For the accidental marketer this is not a nice-to-have. Reporting upwards is the lesson that most directly protects your job, because the person deciding whether "the marketing thing" continues will never see your work — only your page about it.

Start from their questions, not your data

A report is not a tour of your dashboard. Your boss has three questions, and they are always the same:

Are we on target? · What changed, and why? · What do you need from me?

Every widget, every sentence, exists to answer one of those. This is the same decision test as M10.L1, applied to reporting: if a number answers no question the audience actually has, it comes off the page — however proud of it you are. That is how vanity metrics sneak back in: not as KPIs, but as decoration on reports.

The five-slot dashboard

One page. Five slots. Nothing else.

1. KPI vs target — the 2–4 numbers from your KPI tree, each beside its target. Big type.
2. Trend — twelve months of the primary KPI as a line. Trends carry meaning single numbers can't; your M1.L6 baseline is the line's starting point.
3. Best and worst — the top performer and the biggest problem, named. One of each; symmetry is the honesty.
4. One experiment or learning — what you tested or observed (M10.L4's design, or its Plan B verdict), and what it taught.
5. One decision needed — the ask. If nothing is needed: "no decision needed this month" — written, so the slot never silently disappears.

Annotations are the dashboard's memory. Every anomaly gets a dated, written cause on the chart itself. Ostara's chart carries "consent banner changed — measured sessions fell a third; orders unchanged" on the banner date, and "seasonal trough, recurs yearly" every January, because the pack's 34%-of-revenue November–December makes every January look like a catastrophe to anyone without the annotation. Tomas's first annotation: "Easter fortnight — bookings dip, school holidays, recovers unaided." Six months later nobody remembers causes; annotated charts do.

Cadence: weekly glance for you, monthly report for the boss, quarterly review for strategy. Never daily for decision-makers — daily numbers wobble (M10.L4 told you why), and a boss watching wobble learns either panic or contempt.

Situation → Cause → Action → Ask

The monthly report is four short paragraphs, in this order. Here is Lena's first one, now canon — sent to Storkflow's two founders:

Situation. Trials 41 against a target of 55; trial→paid steady at 22%. New MRR added this month: 9 firms. Cause. The gap to target is top-of-funnel: landing-page sessions grew, but Google Ads still delivers a trial at ≈€382 against our €340 target. The signup question says over half our trials start with colleagues or the trade show — channels the click reports can't see. Action. Ads stay pointed at the landing page; the welcome sequence keeps running; next month I'll test the intake-study CTA on the two top forum threads. Ask. Approve keeping the €400/month search spend one more quarter. If the landing page hasn't reached the 3.8% break-even conversion rate by then, I'll cut it myself.

Three minutes to read. And notice what it contains: a miss, reported at full size, with a deadline she imposed on herself. The founders approved — the first Storkflow marketing decision made on a page of numbers rather than a feeling.

How to report a failure

The ads number is a failure — €382 against a €340 target. Watch what Lena did with it, because this is the skill:

• Same prominence as a win. Not in a footnote, not averaged into a blend. Second sentence of the report.
• Cause attached. Not "underperformed" — why: cost per trial versus target, and what would have to change.
• Action and threshold attached. A failure plus a decision rule reads as control. A failure alone reads as confession; a hidden failure, later discovered, reads as the end of your credibility. You are the person who tells the boss bad news early and calibrated — that reputation is worth more than any single result.

The other two honesty rules, standing: show measured vs modelled where it matters ("measured sessions", per M10.L2's banner footnote), and never change a metric's definition without a footnote — a redefined number that improves is indistinguishable from a lie.

The spec outlives the tool

Write the dashboard as a specification — each widget's question, metric, visual, source, refresh — before choosing software. Looker Studio, a spreadsheet, your ESP's own screens: the tool is the last decision, not the first. Tomas's dashboard is a paper page written each Sunday — Check-up bookings from Tally, new patients and their stated source from the front desk, capacity at both clinics, review count. Audience: himself and the team. It satisfies every rule in this lesson at a cost of €0, and it would survive the death of every analytics vendor on earth.

One last thing before you build yours. Open your Module 1 funnel baseline — the page with the estimate boxes ticked and a hypothesis at the bottom. You now have measured numbers for most of it, and a verdict on the leak you predicted. Annotating that page with what measurement actually found is the final act of this module, and for most students the most satisfying moment in the Program so far: the fog you started in, dated and disproven, in your own handwriting.

---