GOAS stands for Goal-Oriented Activity Score. It's not a reporting template — it's a discipline: nothing gets judged until it's been assigned one of four jobs, and nothing gets called "healthy" until it's judged against that job, not a blended average.
A single blended ROAS, a single account-wide CAC target, a single team-wide productivity metric — all of these average together things that are trying to do completely different jobs. A new launch and a bestseller are not the same job. A test campaign and a proven channel are not the same job. When they're judged by the same number, one of them is always being judged unfairly — and usually nobody notices which one, because the average still looks fine.
There's a second problem underneath the first: ROAS, CTR, and CAC aren't your goals — they're proxies, borrowed from generic marketing reporting because they're easy to put in a dashboard. Nobody built a business to hit a target ROAS. GOAS starts from what you're actually trying to achieve — profit, time back, a sellable business, whatever it genuinely is — and scores everything against that instead.
Every job gets scored 1-10 against how well it's doing what it's supposed to be doing — not against a generic benchmark, and not the same scale for every job. A 6/10 Learn job and a 6/10 Scale job aren't measuring the same thing, because the two jobs have different definitions of success. But the bands mean the same thing everywhere:
Actively losing money, or the job can't even be measured right now. This gets fixed first, not scheduled for later.
Real, quantifiable inefficiency — but no acute crisis. Usually the biggest pool of recoverable profit on the whole scorecard.
Aligned to its job, minimal leakage. Worth checking periodically, not worth spending time on right now.
The score isn't a vibe. It's set from real evidence — the data behind it, the root cause, and the fix are always shown alongside the number, on every finding on this site.
Efficiently driving growth that hasn't been earned yet — new demand, new customers, new categories, new markets. Scale is judged on incremental economics, not blended ones: what does the next pound of spend, or the next unit of effort, actually return once you isolate it from what was already working.
This is the job most commonly misjudged, because platforms report blended averages that hide what the marginal spend is really doing.
See the full worked example →Protecting margin and revenue that already exists. Defend campaigns and decisions aren't there to grow anything — they're there to stop something you've already earned from leaking away. Because nothing looks "wrong" when a Defend job fails, it tends to go unnoticed for months: revenue is still coming in, it's just quietly costing more to keep than it should.
The tell is almost always the same: a decision or a campaign gets treated like a growth lever when its actual job was protection.
| Stage | Impact |
|---|---|
| Threshold raised £15 → £25, untested | Silent CVR decline |
| Root cause traced, threshold corrected | ≈£205,000 recovered |
Any policy change (shipping, returns, pricing) that went live without a before/after conversion comparison.
Restoring performance that has genuinely slipped from where it used to be. Recover is different from Defend because there's already evidence something has gone wrong — a decline, a drop, a channel that used to work and doesn't anymore. The job here is root-cause diagnosis first, fix second: the fix is usually obvious once the actual cause is found, but most businesses skip straight to guessing at fixes without confirming the cause.
| Before | After |
|---|---|
| −£1,000/mo net loss | +£2,500/mo profit |
| £42,000/yr swing — zero change to revenue | |
Platform or headcount cost sized for a bigger operation than the one actually being run today.
Generating a signal you can actually trust and act on. Not every activity is meant to pay for itself immediately — a test, a new channel, a new audience, a pilot store. The job of a Learn activity is producing a reliable answer, and it should be judged on whether it did that, not on same-day ROAS. The most common failure here isn't a bad test — it's killing a test before it's produced enough signal to mean anything, and then repeating the same untested assumption next quarter.
Tests paused inside two weeks with no statistically meaningful volume behind the decision to kill them.
GOAS isn't a marketing tool that happens to use business language — the four jobs apply to any commercial decision.
Is this role Scaling capacity the business has earned, or Defending against a gap that could be solved another way? A departing senior role doesn't automatically get re-bought at full cost — its job gets re-evaluated first.
Its job is Scale — but it only earns that job if the unit economics (rent, footfall, contribution) actually support growing the estate. Modelled before signing, not assumed.
Its job is to Defend or enable a specific outcome. If the operation it was sized for has shrunk, the cost should have shrunk with it — most don't, because nobody re-checks.
Its job is Learn — to be a reliable enough signal to plan against. A forecast built on last year's assumptions with no live recalibration isn't doing that job, however confident it sounds.
Assigning a job isn't enough on its own. If that job only gets pursued through one channel because it's the easiest one to switch on, every other channel that could have served the same job just sat idle.
A 20%-off sale on personalised products goes live to drive incremental revenue during a quiet week. The job is clearly Scale. The campaign gets built — but only online: an email send, a paid push, a homepage banner. That's genuinely well executed. It's also only one channel out of several that could have served the exact same job.
Eight physical stores, no window decals, no in-store signage, no staff briefed to mention the sale. Footfall that was already walking past the door got no signal the offer existed.
The email list can be segmented by proximity to a physical store. That segment — customers who could walk in today — got the same generic online send as everyone else, with no nudge toward the nearest store.
Underneath both gaps sits the same habit: a POS request from a store manager — "can you print us something for the window" — gets treated as a design brief. Nobody asks what uplift that store actually needs, or what mix of window decals, local email, staff push, or paid support would get there. The request gets answered. The job doesn't.
| Channel | Included in the plan? |
|---|---|
| Online — email, paid, homepage | Yes |
| Retail stores — window decals, staff briefing | No |
| Local-proximity email segment | No |
| Influencers / local awareness | Not considered |
| Illustrative missed in-store uplift, one promo week (8 stores × ~£3,000 avg weekly revenue × ~20% uplift from proper in-store support) | ≈£4,800 |
Everything above is about finding what's already gone quietly wrong. The same scoring works just as well on something you haven't committed to yet. Is this hire actually going to move profit, or just add cost? Should you sign that agency? Is that SaaS subscription going to deliver the goal it's being bought for, or become another line nobody revisits next year? A job can be assigned and scored before it exists, not only diagnosed after.
15 minutes. No pitch, just a look.
Book a free chat →