Is Your Marketing Agency Actually Working? Separate Activity, Output, and Outcome

Start by taking four things off the table, because each one already has a home and none of them is what this article is about. The month-by-month curve of when search work legitimately begins to show: how long SEO takes to work. What a retainer of this kind ought to cost in the first place: how much SEO costs for a small business. Choosing a replacement, if it ends up coming to that: how to choose a social media agency. And the reason your dashboard and your bank account disagree even when every party is competent: our report on the attribution blind spot.

This piece answers the question that sits before all of those, and it is the one owners find hardest to ask out loud: I am paying this every month, the reports look busy, and I cannot tell whether anything is happening. That feeling is usually correct, and it is usually diagnosable in about an hour.

The Wrong Question Is “Are They Working Hard”

Almost every unhappy agency conversation gets stuck in the same place. The client suspects nothing is happening, the agency produces evidence of considerable effort, and both parties leave the call frustrated because they were answering different questions. The client wanted to know what changed. The agency answered what was done.

Effort is not the test and it is not even useful evidence. An agency can run a genuinely busy month — audits, research, meetings, revisions, a strategy document — and end it having changed nothing about your business that a stranger could detect. That is not fraud. It is the normal consequence of a working relationship that never agreed on what a month is supposed to leave behind.

The question that resolves it is dull and specific: what exists now that did not exist before, and where can I see it? Everything below is an expansion of that one sentence.

Three Ledgers, and Nearly Every Report Only Keeps One

Split what you are paying for into three ledgers. They behave differently, they run on different clocks, and almost all the confusion in this relationship comes from treating them as one thing.

Activity: what was done

Hours worked, keywords researched, audits performed, meetings held, competitors reviewed, posts scheduled. This is what the overwhelming majority of monthly reports contain, because it is the easiest ledger to fill and the only one entirely within the agency’s control. Activity is real work and it is a genuine input. It is not a deliverable, and it should never be the headline of anything.

Output: what now exists

Pages published. Pages rewritten. Structured data added so machines can read the business. A page that used to take nine seconds now taking two. An enquiry that used to land in an unattended inbox now routed somewhere with a response rule. Creative assets you hold copies of. Output is the ledger nobody reports and the one that matters most, because it is the only one that persists. Activity evaporates at the end of the month. Output is still there in three years, working, whether or not the agency is still engaged.

Outcome: what changed in the business

Enquiries, calls, booked jobs, revenue. The only ledger you actually care about and the slowest and least attributable of the three. Outcome is a lagging function of output, which is a lagging function of activity, and each step takes time. An agency that is behind on outcome may be fine. An agency that is behind on output cannot possibly be fine, because there is nothing in the pipeline for outcome to come from.

The Report Looks Busiest Exactly When the Output Is Lowest

This is the mechanism behind the sensation of paying for fog, and it is worth understanding because it is not usually anybody’s bad intention.

A monthly report is written by someone who has to demonstrate value. In a month where a lot was published, that is easy: here are the pages, here is the work. In a month where little was published — a month lost to research, internal staffing, an approval that never came back, or simply drift — the activity ledger is the only one with anything in it, so the report expands to fill the space. More slides, more charts, more keyword tables, more commentary on impressions.

The result is an inverse relationship most owners never notice: the fattest reports tend to arrive in the thinnest months. Once you have seen it you cannot unsee it, and it turns the report’s length into a signal rather than reassurance. It also explains why asking for a better report never helps. The report is not the problem; it is a faithful description of a month in which the output ledger stayed empty.

A second-order version of the same effect shows up in the metrics themselves. Impressions climb when more pages exist, regardless of whether anyone useful saw them. Tracked-keyword counts climb when more keywords are tracked. Engagement climbs with posting frequency. All three can rise every month for a year while qualified enquiries sit flat, which is the precise situation diagnosed in my website gets traffic but no leads.

The Leave-Behind Test

Here is the single test that settles most of these situations, and it takes one question.

If the engagement ended today, what would still exist twelve months from now, and would it still be working?

Some engagements leave a great deal behind. Pages that go on ranking. Structured data that goes on being read. A site that is still fast. Content assets you own and can reuse. A business profile that is accurate and complete. That agency was building something that belongs to you, and the retainer was buying an asset.

Other engagements leave behind a folder of reports and a gap in the calendar. Everything stops the day the payments stop, because nothing was ever deposited anywhere permanent. That is not automatically wrong — paid advertising legitimately works this way, and so does community management, where you are buying presence rather than an asset and both sides should say so plainly. It becomes a problem only when an engagement was sold as asset-building and is in fact rented attention.

Two points make this test sharper than it first appears. The first is that it is answerable today, without waiting for another quarter of data, which is what makes it so much more useful than outcome metrics. The second is that assets only count if you can actually take them, which is a separate question from whether they exist — the five things to check are in do I own my website. An asset built inside an account the agency controls is not yet yours, and the day you leave is a poor day to discover it.

The Six-Month Inventory: One Hour, No New Software

This is the practical instrument. Do not ask the agency for it first — do it yourself, then ask, and compare the two.

Take the last six months of invoices, then open your own website in a private browser window as a stranger would, and build a list with three columns: what exists now, when it appeared, and how you verified it. Verification means you looked at the thing, not that it was mentioned in a report.

Count the pages that are new or genuinely rewritten

Walk the navigation and the sitemap. A page with a changed headline is not a rewritten page. You are counting pages that answer a question they did not answer six months ago. Zero is a finding. Two is a finding. The number matters less than the fact that you now have one.

Check whether the site tells machines what the business is

Structured data is the part of search and answer engine optimization work that is invisible to humans and completely visible to anyone who looks. Run two or three of your important pages through any public structured-data testing tool. Finding nothing at all after six months of a search retainer is one of the clearest signals available, because it is a standard deliverable that takes an afternoon.

Submit your own enquiry form and time the reply

Use a personal address, at nine in the evening, with a realistic question. Then wait. This single test has ended more agency relationships in our experience than any ranking report, and it occasionally exonerates the agency entirely by proving the leak is internal. Either way you learn something no dashboard was going to tell you.

Ask who holds each account

Analytics, business profile, advertising, the site itself. Not whether you have access — whether you are the owner of record. Access is granted and access is revoked; ownership is not.

Four Honest Reasons a Competent Agency Has Nothing to Show

Before anyone gets fired, the fair cases, because three of these four are not the agency’s fault and one of them is probably yours.

The foundation had to be repaired first

An inherited site with broken structure, duplicate pages or an unusable content system can absorb two or three months before anything new can be published. This is legitimate and it is verifiable: the repairs themselves are output, and an agency doing this work can show you exactly what was fixed. An agency claiming it without artefacts is claiming something else.

Approvals are sitting with you

The most common cause of a stalled engagement in small businesses is drafts waiting on an owner who has a business to run. If a content calendar has been in your inbox for five weeks, the output ledger is empty because of the approval queue, and changing agencies changes nothing.

The market is harder than the pitch assumed

Output can be strong and outcome still absent because the terms being targeted are dominated by national platforms or by competitors a decade ahead. This is a strategy conversation, not a competence one, and the honest version of it usually ends with a narrower target rather than a new vendor.

Everything works and the leak is downstream

Rankings improved, traffic improved, enquiries arrived — and nobody answered them within a day, or the phone went to voicemail at lunchtime. The agency did its job and the business lost the result afterwards. The enquiry test above is what surfaces this, and the fix is an intake process rather than a marketing change.

Three Questions to Send Before You Decide Anything

Put these in writing, send them in one email, and give a week. The response is more informative than the answers.

One: list everything produced in the last ninety days that still exists today, with links. Two: which of those do we own outright, and what happens to each if we stop. Three: what is the one metric you are managing toward this quarter, and what is your own honest read on it right now.

A capable agency returns the first list quickly because it already exists internally. A capable agency answers the second crisply and without defensiveness. And on the third, the tell is volunteered bad news: partners who are comfortable saying which part is not working yet are the ones who will still be useful in a year. An agency that cannot produce the first list has answered all three questions at once, and you got there in a week rather than in another two quarters of retainer.

What Good Looks Like, and Our Position in Saying So

A healthy engagement is legible. Every month leaves something behind that you could show a stranger. The reporting leads with what now exists, then what it produced, and keeps the activity list short because it is the least interesting part. Ownership sits with you from the beginning rather than being transferred during an awkward exit. And the agency tells you which parts are not working before you ask.

We should be plain about the obvious: we sell search and website work, so an article teaching owners to audit their current agency is not a neutral act. Two things are worth knowing about that. The tests above are designed to be run without us, and their most common result is a repaired relationship rather than a switch — particularly where the inventory shows the approval queue or the enquiry response is the real constraint. And our own bias is visible in the framework itself: we build with Claude Code, which means our output ledger fills faster than our activity ledger, so a standard that judges agencies on what exists rather than on effort is a standard that suits us. Read it with that in mind, and then apply it to us on the same terms. A related version of this problem — software you pay for that never becomes something the business depends on — is in our report on the small business AI stack.

The last point is the one worth keeping. The expensive mistake in this situation is almost never firing a good agency. It is spending another two quarters unable to tell, because the only evidence anyone offered was effort and effort is the one thing that never accumulates.