“My marketing agency isn’t delivering” — how to tell before you fire them

Judge a marketing agency on booked outcomes across a full 90 days, not on one bad month. Google Ads recommends evaluating Smart Bidding performance only over a period carrying at least 30 conversions, so paid below that count is not readable yet. If your channel’s realistic time-to-result has not yet elapsed, the problem is the timeline, not the agency.

• Most common cause: the channel has not had time. SEO and cold outbound are measured in months; paid is measured in conversion counts, not days.
• Second most common: the agency counts leads, you count closed deals, and nobody counts booked calls.
• Genuine underperformance: the readable window has passed, every leading indicator is flat, and a written brief change produced no response.
• Free first move, tonight: pull leads, contacted, booked, showed and closed for 90 days out of your own CRM, not their dashboard.
• The decision: three failed gates, switch. Two, give them another quarter with a written scoreboard. One, the leak is on your side.

“My marketing agency isn’t delivering results” — what is actually happening

Almost every version of this sits in one of three states, and they need opposite responses. Too early: the engagement is 6 or 10 weeks old and the channel has not reached the point where its numbers mean anything. Measurement mismatch: something is working, but the agency reports impressions, clicks and “leads” while your constraint is booked calls that show up — both sides honest, both frustrated. Genuine underperformance: the window elapsed, the upstream numbers never moved, and nothing changed when you asked.

How long before an agency should show results, by channel?

Check this before anything else. “Readable” means the number has enough volume behind it to be signal rather than noise.

Two rows have published sources rather than our opinion behind them. Google Ads “recommends measuring performance over longer time periods that have at least 30 conversions, such as a month or longer (50 conversions for Target ROAS)”, and Google’s SEO starter guide is blunter: “Some changes might take effect in a few hours, others could take several months.” We apply that 30-conversion floor to paid social as a working minimum rather than quoting the learning-phase figure that circulates on ad blogs, because we could not verify it on Meta’s own help pages. The outbound row is ours — week by week in how long AI outbound takes to ramp — and the reactivation row is our database reactivation record of 4.4% average and 8.9% peak on dormant lists, our result rather than an industry benchmark.

The quotable version: an agency hired in March for SEO cannot be fired for underperformance in May, because May is inside the window where nobody could tell.

Is it as urgent as it feels?

Usually less urgent than it feels. If you book 20 calls a month, one week is about five calls, and ordinary variation moves five by one or two with nothing wrong. At that volume a quiet week carries no information.

The readable unit is a rolling four weeks against the previous four, and it takes two consecutive declining four-week windows before you have a trend. One is noise. Check the same month last year in your own CRM first: a lot of what presents as an agency problem is seasonality both parties forgot to expect.

The genuinely urgent version is cash: if the business runs out of runway before the readable window closes, the decision is not about the agency at all, and waiting does not fix it.

What to do in the next 24 hours

All five are free, take about an hour, and none involve an agency, including us.

1. Count it yourself. From your CRM and calendar, not their dashboard: leads created, contacted, calls booked, calls showed, deals closed. Last 90 days, and the 90 before.
2. Time-stamp your own follow-up. Measure the gap between the last 20 leads arriving and your first outbound attempt. This is the number most often blamed on an agency that the agency does not control.
3. Re-read what you bought. Deliverables or outcomes? Most disputes are a contract promising activity being judged against a hope of revenue.
4. List what changed on your side. Price, offer, landing page, broken form, tracking or consent change, a salesperson leaving. Write the dates down.
5. Send one email. Ask for the number the engagement was bought to move, monthly, for six months. An agency that produces that in a day is usually not the problem; one that cannot has been reporting activity.

The quotable version: the fastest way to find out whether an agency is failing is to count what they count, in your own system, for 90 days.

What to do in the next 7 days: the three-gate test

The rule we use before recommending anyone change supplier, including away from us. Run the gates in order.

The decision rule: three failed gates and you have evidence rather than a feeling — switch. Two failed gates and the honest answer is another quarter with a written scoreboard, because a switch costs 6 to 10 weeks of restart to learn what waiting would have told you. One failed gate and the problem is on your side of the line.

When it really is the agency, and when it is your side of the line

The dividing line is the moment a lead is created: above it is the agency’s work, below it is yours. Substitute your own numbers.

Take 400 leads over 90 days. A 32% contact rate is 128 conversations. A 19% conversation-to-booking rate gives 24 booked calls. A 61% show rate, 15 calls held. A 24% close rate, roughly 3.5 deals. Now change one number below the line: lift contact rate to 55% by answering faster, and the same 400 leads give 220 conversations, 42 bookings, 26 held calls and about 6 deals.

The quotable version: the same 400 leads produced either three and a half deals or six, and the agency’s work was identical in both versions.

In our own client work, speed to lead alone is typically worth about 3x on that chain and doubling contact rate about 2x. The honest wrinkle: they do not multiply. 3x and 2x is not 6x, because fixing speed to lead is part of how contact rate improves — overlapping descriptions of one repair — and anyone quoting stacked multipliers is selling. If the leak is below the line, the fix is a follow-up system rather than a new supplier, which is what our AI appointment setting service does; the arithmetic behind our published lift figures, including that the 7x average has a median closer to 4x, is on our methodology page.

Where we are not the answer

Four situations where hiring any agency, us included, is the wrong purchase.

• You want money back. A contract question for a lawyer, not a marketer; the answer is in the termination and performance clauses you signed.
• The claims were false, not merely unmet. A claim made with no reasonable grounds is a regulatory matter — the ACCC in Australia, the FTC in the United States. Both publish the standard, and it is about evidence held before the claim, not results after.
• You need to know what the write-off does to this year. Accountant.
• Nobody is selling. If held calls have converted below 10% for two quarters, more leads make the problem more expensive, not smaller.

The fix that stops this recurring

The structural fix takes an afternoon: agree a scoreboard before the next engagement starts, not after it disappoints. Four lines, signed by both sides.

1. One number the engagement exists to move, defined in your CRM, with the field named.
2. The readable window for the channel, from the table above, written as a date.
3. A monthly leading-indicator review of the upstream numbers, not revenue, because revenue moves last.
4. A written stop condition agreed on day one: the result that would mean this is not working, decided before either side is invested in it.

Stop conditions only work if people honour them. We published one of ours. Measured 19 August to 3 September 2026, the booking quiz embedded inline on our content pages drew 145 impressions, 1 start, 0 contacts and 0 bookings — against 126 impressions, 46 starts, 22 contacts and 8 bookings on our booking pages over the same window. We removed it from content pages rather than argue with the number. An agency willing to show you a test it stopped is telling you how it behaves in month four.

Stated plainly, once: we run on pay-per-result, so you pay on booked qualified appointments rather than retainers or seats, which means the readable window and the invoice move together. Since 2017 that model has produced 50,769+ AI-booked sales appointments across 1M+ leads worked. Whether you buy that structure or write outcomes into the contract you already have — trade-offs both ways in our comparison of pay-per-result versus retainer marketing agencies — the mechanism is the same: put the outcome on the invoice.

Frequently asked questions

How long should I give a marketing agency before firing them?

It depends on the channel. Paid is judged on data volume rather than time: Google Ads advises measuring Smart Bidding performance “over longer time periods that have at least 30 conversions, such as a month or longer (50 conversions for Target ROAS)” — see Google’s About Smart Bidding documentation. SEO runs 6–12 months to revenue, cold outbound 3–6 months, database reactivation 2–6 weeks. Firing inside the window buys a restart, not an improvement.

My agency sends reports every month but I still cannot tell if it is working. What should I ask for?

Ask for one number, monthly, for six months, defined as a field in your CRM rather than in their dashboard. Booked qualified calls is usually the right one, because it is the last metric both sides can genuinely influence. If the answer arrives as a deck of activity, that is the finding.

Is two bad months enough to say an agency is not performing?

Two consecutive declining four-week windows is a trend rather than noise, so it is enough to trigger the three-gate test, but not enough on its own to switch. Check seasonality against the same months last year, and whether contact rate and show rate on your side moved too.

Can I get a refund if the agency promised results it did not deliver?

That is a contract and consumer-law question, and it belongs with a lawyer. For the standard itself, the ACCC’s guidance on false or misleading claims says any claim a business makes must be “accurate, truthful and based on reasonable grounds”, and the FTC’s advertising FAQs for small business say “the law requires that advertisers have proof before the ad runs”. Neither turns a missed forecast into a breach.

Does changing agencies reset the clock?

On paid search and paid social, effectively yes — new accounts and campaigns relearn and you are back under the 30-conversion floor. On cold outbound, domain and number warm-up restarts, which is weeks. On SEO the published pages survive, which is why organic is the cheapest channel to switch supplier on and paid the most expensive.