Hiring an SEO agency after getting burned: the 30-day test

You paid an SEO agency for a year. The reports showed traffic going up every month, and your sales team still can't name a single deal it produced. You still need a channel besides paid ads, so you're shopping again.

Hire the next agency on terms that would catch the same failure within 30 days. Pay for the strategy month as its own project, and ask for a forecast that ties the planned pages to leads and revenue. Sign a contract you can leave with 30 days' notice. At the end of month 1, judge the plan before anyone writes a page.

When a founder tells me an agency let them down, the cause usually fits 1 of 4 patterns, and each one shows up in that first month. Below is how to spot each one in your old agency's reports, and the questions that catch them before you sign another contract.

Why did my last SEO agency fail?

Usually the old agency either chased traffic instead of revenue, sank its time into technical fixes, never promoted the work, or took shortcuts that Google later punished.

Agencies usually guess wrong about why they got fired. In Setup's 2025 Marketing Relationship Survey (a small sample, about 100 responses, so treat it as directional), 61% of clients who ended an agency relationship cited dissatisfaction with delivery, and 61% cited value. In Setup's 2020 survey, 75% of agencies believed budget cuts were the leading reason they lost clients, while only 24% of clients said budget cuts were the reason.

Here are the 4 patterns I see most when we inherit someone else's work.

1. They chased traffic instead of revenue. The reports showed clicks, impressions, and rankings. To grow those numbers, the agency went after high-volume, top-of-funnel searches with no real way to turn a reader into a lead. You end up paying for a library of content that brings in a lot of eyeballs and no return.
2. They got stuck in technical SEO. Month after month of tickets for changes worth a fraction of a percent, eating your developers' time along with the agency's. A site needs a cleanup so it isn't shooting itself in the foot. After that, the work should move to the pages that bring in business.
3. They never promoted anything. All the work happened on your website. Nobody earned links or press for it, shared it on social, or put paid behind it, so the content just sat there, waiting for Google to come along and bless it with rankings.
4. They took shortcuts. Thousands of programmatic pages, AI content with no human in the loop, paid syndication, and links from private blog networks. These tricks work for a while, until Google catches on and the site pays for it. Around 2013 I cleaned up more manual link penalties than I can count. It was gross.

Before you talk to another agency, pull the last 6 months of reports from the old one and write down which of these 4 you see. That's the first thing to screen for.

What does it look like when an agency chases the wrong numbers?

It looks like a site full of pages that bring in visitors who were never going to buy. One of our clients, a home improvement financing marketplace, came to us right after Google's helpful content update hit them, and that's what we found.

Their previous SEO vendor did mostly good work. But it also built hundreds of near-duplicate pages, one for nearly every keyword variant it could find, whether or not the topic fit the business. Over time the site grew to thousands of pages, covering loan types that had little to do with what the company does.

For a while it paid off. The traffic was big enough that the company built a few new product lines to monetize it. Then the helpful content update landed, and those pages have lost ground with almost every Google update since.

So we pushed in the other direction. We kept recommending fewer pages chasing loose keywords and more depth on the core business lines, and they've acted on most of it. They're letting the old pages bring in what they can, and our job is to grow the core business. A big part of that is content we write for the contractors on their platform and send by email, which opens at about 27–28%.

Their team asks every new contractor how they found the company and logs the answer on every sales call. Lately, close to half of new contractors say on those calls that they found it through ChatGPT or another AI assistant.

You can check your own company the same way. Ask ChatGPT and Google's AI Mode which companies they'd recommend in your category, and see whether yours comes up. If you'd rather have us look, we do that in a free teardown, along with your site and your lead path.

If your last agency left you with a page factory, don't tear it all down on day 1. Ask the next agency to sort your pages into two groups (pages that tie to your core business, and pages that only bring traffic) and to put the budget behind the first group.

Can I start with a short commitment and prove it out first?

Yes, and after a bad agency you should. Buy the strategy month on its own, with a 30-day out, before you commit to production.

Full disclosure, this is how we sell our own work at Narrative Bent. Every contract runs 30 days at a time, and month 1 is strategy only. At the end of it, a client can hand the plan to their in-house team, take it to another agency, or start production with us in month 2. Either way they've paid for the strategy, and they own it. Month 1 is billed like any other retainer month, with no onboarding fee, and our plans start around $10K a month.

There's a tradeoff for us too, by the way:

Our revenue is less predictable, and some clients have used the 30-day out. Most stick around, and our typical engagement has run 9 months or more. I'll take that trade, because long contracts hide bad work. If a client can leave next month, we have to earn next month, and that's the only accountability I trust.

This idea is older than SEO agencies: In 2012 the White House budget office told federal IT buyers to break big contracts into smaller pieces, because a smaller piece "can be terminated with fewer sunk costs, capping the risk exposure to the agency when priorities change, a technology decision doesn't work or the contractor's performance doesn't deliver results" (OMB guidance on modular contracting, p. 3). "Agency" there means a government agency, but the logic carries over.

Ask every agency on your list whether you can buy the strategy without the retainer. If they won't sell it on its own, ask what they're worried you'll find out.

What should I have in hand after 30 days?

By day 30 you should have a plan you could hand to anyone, and a baseline to measure it against. If you don't have both, more months won't fix it.

Here's what a good month 1 should produce, using ours as the example:

1. Week 1: kickoff, plus access to your analytics, Search Console, and CRM.
2. Weeks 1–2: interviews with your team and your customers, plus keyword, audience, and competitor research.
3. Weeks 2–3: a technical SEO and AI search audit, and a check of what AI assistants say about your brand, with every fix ranked and prioritized based on the potential impact on the business' bottom line.
4. Week 4: the strategy, an editorial calendar, a prioritized roadmap, and a dashboard that ties Search Console, analytics, AI visibility, and your CRM together, with a starting baseline. You should be able to see the entire funnel from what someone searches all the way through to whether or not they became a qualified lead.

Then you decide whether to continue, change the scope, or walk away.

Each item on that list answers one of the 4 failure patterns. The forecast and the CRM baseline catch an agency that would chase traffic. The ranked audit shows whether technical work will stay in proportion. The plan should say how content gets promoted, and nothing in it should depend on page counts or links from sites your buyers never read.

Caroline Feeney, who runs content and client services at Narrative Bent, puts it this way: "By the end of month 1, your CEO should be able to open the dashboard, see where leads come from today, and point to the first pages we'll build and why each one should bring in business. If they can't, we haven't done our job yet."

Google's own page on hiring an SEO suggests asking, "What kind of results do you expect to see, and in what timeframe?" Month 1 is when you should get that answer in writing.

How will we tie the work to leads and revenue?

Ask for a forecast before you sign. It should run from the searches the agency would target, to the clicks you can expect, to leads at your conversion rate, to revenue at your close rate and deal size, and you should be able to change any number and watch the revenue move.

Look, I'm not going to lie to you, marketing attribution is an absolute mess right now and plenty of marketing teams have this exact same problem. In the Content Marketing Institute's 2025 B2B research (980 respondents), 56% said attributing ROI to content was a challenge. In The CMO Survey 2025 of 281 senior marketers, 64% named showing marketing's impact on financial outcomes as their top challenge.

We build the forecast at the start of every strategy and it's how we decide what to write first, and it gives you a number to hold us to.

Some smart people think forecasting is a mistake though. Take Eli Schwartz, author of Product-Led SEO, who wrote this year that "keyword-based SEO forecasting is basically fiction." He has a point, since search volumes are estimates and every forecast will be off somewhere.

Where I land is that the forecast works as a North Star, and you revise it as real lead data comes in. Kevin Indig, who writes the Growth Memo newsletter, put it this way: "The key is managing expectations with leadership that forecasts might be off and must be adjusted on a regular basis."

An agency that won't forecast at all is asking you to judge it on rankings and traffic and that's how bad pattern #1 starts.

Ask the agency to walk you through the forecast math for the first page they'd build, using your own deal value and close rate.

When should I expect results?

Plan for 6 to 12 months before content pays for itself, and judge the direction at month 3. That's why I tell clients to be ready for a 12-month commitment, even though we work month to month.

The 30-day out protects you from a bad agency. Cutting a good program at month 6 is a different mistake. I've watched founders look at month 6, see a line that's barely moved, and cut it. The client burned 6 months and a pile of money, and the agency takes the blame for a channel that never got a fair shot.

By month 3 you should see early signals: rankings moving on the first pages built, AI visibility tracked against your month 1 baseline, and a reporting dashboard you actually read. Leads come after that.

In a 2017 video on hiring an SEO, Google's Maile Ohye said, "In most cases, the SEO will need four months to a year to help your business first implement improvements and then see potential benefit" (as reported by The SEM Post). The video is old, but I haven't seen anything since that changes the range.

Write down what you expect to see at months 3, 6, and 12, and get the agency to agree to it before production starts.

Which answers should end the sales call?

End the call if an agency guarantees rankings or AI mentions, sells you special files for AI, reports traffic as the only win, or can't tell you who edits AI-assisted content.

Let me make this extremely clear:

Anyone guaranteeing results in this business (especially right now when AI is disrupting EVERYTHING) is lying.

• Guaranteed rankings. Google's page on hiring an SEO says it plainly: "No one can guarantee a #1 ranking on Google."
• Guaranteed AI mentions. This is the same pitch in a new channel. OpenAI's help page on ChatGPT search says, "Placement is not guaranteed." Anyone selling you a guaranteed number of AI mentions is selling a number they don't control.
• Special files and AI-only pages. Google's guide to its generative AI features says, "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them." If an agency's AI plan is mostly llms.txt files, you're paying for something Google says it ignores.
• Links with nothing behind them. The links that hold up are contextual, useful to the reader, and point to something only you have, like your own data or research. Content-led PR is slow, but it's about the only link building I still believe in, because a journalist choosing to cite your data is the whole point.
• Traffic as the headline. Traffic, rankings, and citations are inputs. Leads, pipeline, and revenue are the score. Nothing matters if it isn't growing the business.
• AI content with no editor. Google released spam updates in August and September 2026 (Search Engine Roundtable). Glenn Gabe, who studies algorithm updates at GSQi, wrote that the August update "was fierce for those impacted, and there were some huge drops," and his case studies included programmatic, AI-generated, and scaled-content sites.

A quick early win doesn't settle any of this. Lily Ray, who writes about SEO and GEO, warned that "Seeing something work well in SEO/GEO for 3-6 months is not a guarantee that it will work long-term." Our home improvement financing client learned that the slow way.

The 30-day test: what to ask before you sign

Take this table into your next sales call. Each row matches one of the 4 ways agencies fail, plus the contract terms that let you act on what you find.

One red-flag answer from any agency (including ours) is enough reason to keep looking.

Want a second set of eyes before you hire anyone?

Book a free teardown with us. We'll look at your site, how a visitor becomes a lead, and what AI assistants say about your company. We'll tell you where we think the path from search to a sales conversation breaks, and send you 2–3 recommendations you can act on.

They're yours to use with your own team, with another agency, or with us. If we're one of the agencies you're vetting, use the table above on us too.

Book a free teardown