Metrics that matter (and the ones that don't)

You will be asked whether it's working

Sooner or later someone asks whether the marketing is working — a boss, a co-founder, or your own bank statement. "We're getting good engagement" is not an answer. It's what people say when they don't know.

The wrong numbers are worse than no numbers, because they make you defend activity instead of results. A report full of follower counts and impressions proves you were busy. It cannot prove you were useful, and everyone in the room quietly knows it.

Here is the alternative, and it is smaller than you'd expect: five numbers, one per funnel stage. That is enough to run an SME's marketing honestly. This lesson is about choosing the right five.

The decision test

A metric earns its place by one criterion only: some value of it would change what you do next.

Apply the test like this. Name two plausible values the metric could take next month, and the two different actions you'd take in response. If every value leads to the same behaviour, the metric is decoration.

Try it on Ostara Ceramics. Mateus tracks Instagram followers: 14,200 today. If it reads 15,000 next month, what does Ana do differently? Nothing. If it reads 13,900? Also nothing. The number can move in either direction without a single decision changing — so it is a vanity metric, however good it feels to watch.

Now try orders per month: 91 today. At 70, Ana investigates the same day — is it the site, the ads, the season? At 120 before Christmas, she checks kiln capacity and shipping cut-offs. Different values, different actions. That's an actionable metric.

Vanity metrics tend to share three properties: they only ever go up (cumulative totals like "lifetime website visits" cannot fall, so they can't warn you), they flatter, and they imply no decision. Actionable metrics — orders per month, conversion rate, cost per lead — can go down, and it means something when they do.

One number per stage

You mapped your funnel in M1.L2: Awareness → Consideration → Conversion → Retention → Advocacy. Give each stage one honest number. Here is what that looks like across all three cast businesses, using only what they can actually count today:

Look at the gaps before you look at the numbers. Ostara's Shopify records add-to-carts; nobody has ever opened that report. Neither Storkflow nor Ostara counts advocacy at all. Writing "unknown" in a cell is not failure — it is a finding, and usually a cheap one to fix. A forty-metric dashboard with no gaps is almost always hiding more than this table reveals.

The rate between the stages

The five numbers are connected by conversion rates — the share of people who take the next step. This is the connective tissue of all funnel arithmetic:

next stage = this stage × conversion rate

Ostara: 6,500 sessions × 1.4% = 91 orders. That one line is most of funnel analysis. It tells you there are exactly two ways to get more orders — more sessions, or a better rate — and in M1.L6 you'll use it to work out which one your business needs.

Leading and lagging

Revenue is a lagging indicator: by the time it moves, the cause is weeks or months old. At Storkflow, a slow month for trial starts shows up in revenue a month or two later, after trials run and decisions land. Trial starts are the leading indicator — they predict what revenue will do.

The working habit: watch leading indicators, report lagging ones. Your boss cares about revenue; your steering wheel is leads, enquiries and conversion rates, because those are the ones you can still do something about.

When a measure becomes a target

In 1975 the economist Charles Goodhart observed what is now called Goodhart's Law, usually phrased as: when a measure becomes a target, it ceases to be a good measure.

The mechanism is ordinary, not sinister. The moment a number decides bonuses, budgets or praise, people optimise the number — and the cheapest way to move a number is rarely the way that moves the business.

Concretely: suppose Ana ties Mateus's bonus to the ROAS his Meta ads report. The fastest way to raise ROAS is not to find new customers. It is to shift the whole budget into retargeting people who were already about to buy, so the ads claim credit for sales Instagram had made anyway. ROAS climbs. New customers don't. The measure improved; the business didn't.

Notice that this has already half-happened at Ostara without anyone being paid for it: the reported ROAS of 3.1 is last-click and mostly retargeting. Goodhart's Law doesn't need bad intent — a dashboard is enough.

Two defences. First, pair every target with the number it could be gamed against: ROAS with new-customer count, trial starts with trial-to-paid rate. Second, keep asking what the number is for. A metric is an instrument for making decisions, not a score for winning.

Measure before you touch anything

Record your baseline before you change anything. It's the only way you will ever be able to say "that worked" and mean it — which is why measurement sits here in Module 1, not in Module 10 where analytics gets its full treatment.

The counting infrastructure you need today is minimal: site analytics (Plausible or GA4), some way to count enquiries by source — asking "how did you hear about us?" and keeping a tally counts — and your sales records. Honest ticks in a spreadsheet qualify.

Mateus's report, rebuilt

Mateus's old monthly report to Ana: followers, reach, likes per post, ROAS. Four numbers, zero decisions.

The new one is the middle column of the table above: 6,500 sessions → add-to-cart unknown → 91 orders at 1.4% → 19% returning → advocacy uncounted. Now a real conversation exists. The number to interrogate is 1.4%, the two unknowns become this month's homework, and follower count stops being the headline.

One line on Storkflow, because the principle travels: for a nine-person B2B firm, "trial starts per month" beats "website traffic" as the headline metric every time. Traffic is three steps removed from money; trials are one.

Now choose your five.

---