Cost per conversion (or cost per lead, or cost per acquisition — same idea, slightly different metrics) is a useful number, but it is one input into a real budgeting decision, not the entire decision. A $400 lead for a $40,000 product is a wildly better outcome than a $90 lead for a $4,000 product, even though the per-lead number favors the second. An agency that optimises to lowest cost per conversion in the absence of revenue or margin signal is not running your strategy. They are running Google’s default heuristic, and your strategic priorities are silently losing the budget allocation argument every week.
It’s the easiest target to hit and the easiest to defend.
Cost per conversion is countable and immediate. It shows up in the platform within days. It is the number Google’s automated bidding strategies are designed to minimise out of the box. The agency can hit it without conversation, without context, without understanding what your business actually values. As long as the per-lead number trends down, the report looks good.
The strategic alternative — weighting conversions by product, by margin, by deal size, by sales-cycle stage, by lifetime value — requires three things most retainers don’t fund. It requires the client to tell the agency what each conversion is worth (which most clients haven’t formalised). It requires conversion-value tracking inside Google Ads, with values feeding through from your forms or your CRM (technical setup). And it requires the agency to optimise toward target ROAS or value-based bidding instead of the simpler cost-per-conversion targets, which means more variability in the early weeks and more explanation in the report meeting.
So the easy version wins. The agency hits cost-per-lead targets. Budget flows toward whichever product happens to be cheapest to lead-gen. Your strategy is a footnote on slide twelve.
Two questions, in this order.
First: “What conversion value is currently configured for each of our conversion actions in Google Ads, and how was that value derived?”
Second: “If I told you Product A is twice as strategically valuable to us as Product B, how would the campaigns reflect that — specifically, what changes? ”
It means your strategic priority is not encoded anywhere the algorithm can see. The agency is using one number — cost per lead — as a stand-in for the entire economic decision about where money should go. As long as that one number looks good in the deck, the underlying allocation is invisible. The agency has, without saying so out loud, decided that the auction’s view of efficiency outranks your view of strategy.
The fix is genuinely technical and requires both sides to do work. You owe the agency a written ranking of conversion value by product, line of business, or deal type, with rough numbers attached — even imperfect numbers, because $14,000 vs. $3,000 is a directional signal the algorithm can use. The agency owes you proper conversion-value tracking, the right bidding strategy, and a campaign structure that can be rebalanced when your priorities change. That conversation is what a real partnership looks like.
If the conversation produces shrugs on the agency side — “we just optimise to cost per lead, that’s our methodology” — you’ve learned where the ceiling is on what the agency is willing to do. Your portfolio strategy should not be downstream of an account manager’s methodology. If you want help putting numbers on this for your own account, that’s a good free question.
]]>A description is not insight. “Click-through rate is up 8% month over month, with strong performance across our consideration campaigns” is a sentence that contains zero information you couldn’t get by reading the dashboard yourself. A real report sentence is: “Click-through rate is up 8%, driven almost entirely by ad-copy variant C in the consideration campaign, which we shipped on April 11th. The variant emphasises ROI calculator language, and the next step is rolling that messaging into our prospecting campaign by the 25th.” The first one tells you the score. The second one tells you the game.
Interpretation is exposing. Description is safe.
Saying “X happened because of Y” commits the agency to a causal explanation. If that explanation is wrong, it’s embarrassing two months later when something else turns out to be the real driver. Most account managers, especially junior ones, learn quickly that the safe move is to describe what the numbers did and let the client draw their own conclusions. The deck looks confident. The author is shielded. Nobody can be wrong about a number that’s simply restated.
There’s also a labor problem on top of the safety problem. Real interpretation requires understanding cause. To say why CTR moved, the analyst has to know what changed in the account, what changed in the market, what your competitors are doing, what the seasonal pattern looks like in your category, and what your own product cycle is. That’s real thinking, every month, on every account. A template that just renders the numbers takes ten minutes. A template that interprets them takes two hours. Across a book of fifty clients, that math doesn’t work for an agency optimising for margin.
Pick one specific metric movement in the most recent deck and ask: “Walk me through what specifically caused that, what changed in the account or the market, and what we’re doing in response.”
Don’t accept the description. Push for the cause and the response.
It means the report is doing its safe job and nothing else. The agency is describing the weather, not telling you what to wear. There is no commitment to any specific action, because no specific cause has been named, because naming a cause is exposing. The cumulative effect over six months is that you have no idea what your agency is doing, what they’ve learned, or what they’ll do differently — you just have a stack of decks that all sound vaguely positive.
The fix is reframing the report itself. Tell your agency the next deck needs a “what / why / so what” section for every meaningful change, with specific cause attribution and a specific response. Two sentences per movement. If that gets pushback, you’ve learned something more important than any individual number: you’ve learned that the agency would rather report safely than report usefully. That’s the conversation the renewal meeting should be about.
]]>A landing page is a single-purpose page built for the keyword cluster that brought the visitor there. It has one headline, one promise, one form, and no navigation menu inviting the visitor to wander off. It can be tested, iterated, and scored. A homepage is a navigation hub built for fifteen different audiences at once, optimised by committee, and almost impossible to attribute conversion changes to. Sending paid traffic there is the digital equivalent of paying for a sales lead and then handing them a stack of brochures.
Three reasons, in order of frequency.
Landing pages aren’t the agency’s job, technically. A standard PPC retainer covers in-platform work: campaigns, keywords, bidding, reporting. Building landing pages requires a CMS, design, copywriting, and developer time. The agency’s scope ends at the click. The client’s scope, in theory, picks up at the click. In practice nobody owns the click-to-form-fill conversion because both sides have decided it’s the other side’s problem.
The agency doesn’t want to be measured on what happens after the click. As long as the destination is the homepage and the conversion rate is whatever the homepage produces, the agency can hide behind “that’s a website issue, not a campaign issue.” The moment you build a dedicated landing page, the agency is on the hook for landing-page conversion rate — and the gap between “could be 6%” and “currently 1.4%” suddenly has their fingerprints on it.
Nothing has been tested. Even on accounts that do have landing pages, those pages are usually built once and never touched. Same headline, same form, same hero image for fourteen months. No A/B testing (running two versions of a page against each other to see which converts better). No iterative changes. The page might as well be a homepage for all the dynamic optimisation it’s getting.
Two questions.
First: “What percentage of paid spend is currently routed to dedicated landing pages versus the homepage or other site pages?”
Second: “What landing-page tests have you run in the last 90 days, and what did each one teach us?”
It means nobody owns the most important conversion lever in the entire account, and the agency is comfortable with that. The bad answer is technically accurate — a PPC retainer often doesn’t include landing-page production. It is also a tell, because the right response from a real partner is “here’s the gap, here’s how we close it, here’s who needs to do what.”
The fix is one of three things. Either the agency expands scope to include landing-page production and testing (and the retainer adjusts to match), or you bring in a landing-page specialist who works alongside the PPC agency, or you assign internal resources to it. All three are real answers. “Not our scope” without a follow-up suggestion is not a real answer; it’s the agency telling you they’d rather optimise the part of the funnel they get credit for than the part of the funnel that actually limits performance.
If you want a quick sanity check on whether your landing pages are the cap on your account, that’s a good free question.
]]>Most agencies do roughly an hour of customer discovery during onboarding, fill in a one-page ICP (ideal customer profile) document, file it in a Google Drive folder, and never look at it again. Twelve months later they’re still running campaigns based on assumptions made in a forty-five-minute call with someone who isn’t at your company anymore. The campaigns work to the extent they happen to be correct by accident.
Customer understanding is the highest-leverage thing an account team can do, and the lowest-billable.
Spending two hours talking to your sales team about which deals close fast and which ones stall is unbillable on a fixed retainer. Reading the transcripts of three of your closed-won discovery calls is unbillable. Sitting in on a customer success call to hear what your existing customers actually say about why they bought is unbillable. All of it pays off — in better keywords, better ad copy, better audiences, better landing pages — but the payoff is diffuse and the cost is concentrated. So it doesn’t happen.
There is also a competence problem. Asking deep questions about a customer requires curiosity, time, and a willingness to admit you don’t already know. Agency analysts are often optimised for technical platform skill rather than commercial intuition. They can build a tightly structured campaign in six hours; they can’t tell you what your buyer is afraid of. Both skills exist on every senior team. Only one of them tends to be staffed at the level your retainer is paying for.
“Walk me through who you think our customer is. Not from the onboarding doc — from your own understanding now, ten months in. Who buys, why, and what stops them?”
If they don’t have a confident answer, that is the answer.
It means the campaigns are built on the platform’s defaults plus an outdated one-page summary, and the agency hasn’t done meaningful customer thinking since the kickoff call. Performance is whatever the algorithm happens to find. The agency’s value reduces to platform mechanics, which is real value but a fraction of what you’re paying for.
The fix is uncomfortable for both sides. Either the agency commits to a quarterly customer-discovery cadence — sales call shadowing, customer interviews, ICP refresh — and treats it as part of the retainer, or you start treating customer briefing as your own work and feed them the answers each quarter. The first version produces better campaigns. The second produces an agency that is functionally an executor of your strategy rather than a thinking partner. Both can work. The version that doesn’t work is the one where neither side is doing it and the campaigns drift.
]]>The red flag is when “we’re still in the learning period” becomes the standing answer for every month that doesn’t go the right way. It is one of the most reliable signals on this list because it requires no looking at the account to detect. You just have to count how many months in a row you’ve heard it.
It is the most defensible-sounding rhetorical shield in the whole vocabulary.
A learning period is a real concept, documented by Google, with real mechanics. It cannot be dismissed by a non-practitioner without sounding like the non-practitioner doesn’t understand the platform. So an account manager who has nothing useful to say about why a month went badly can reach for “learning period” the way a politician reaches for “it’s complicated” — technically correct, almost always insufficient, and effective at ending the conversation.
The mechanic underneath is usually one of three things. Either the agency is making frequent small changes that keep retriggering the learning state (which is itself a process problem), or the agency made a single change two months ago and is still claiming we’re calibrating (which is a credibility problem), or the agency has made no change at all and is using “learning” as a generic term for “we don’t want to commit to why this happened” (which is the worst version).
Two questions, in this exact order. The order is the trap.
First: “What specific change triggered the current learning period, and on what date?”
Second, regardless of what they answer: “What does the trailing-90-day trend look like? If we’re still calibrating from a change made over three weeks ago, that’s longer than the platform documents. What’s actually happening?”
It means the agency is using a real Google Ads concept as a fog of jargon, in the hope that you don’t know enough to notice. It is the verbal equivalent of slowing down at a yellow light forever. Calling it out by name — “a learning period documents at two to three weeks; we’re past that, what’s the actual issue?” — collapses the rhetoric. Either you get a real answer or you get the next layer of fog, and the next layer of fog tells you everything you need to know about whether to keep going.
In my own auditing work, this is the single most common phrase I see used as cover. When I get into the account and pull change history, the “learning” that’s allegedly happening is one of two things. It’s either nothing — no real changes for weeks — or it’s churn, where the agency keeps tweaking small settings that keep restarting the algorithm. Both explanations are real. Neither is the explanation the client got.
]]>The honest version is closed-loop tracking. The lead leaves Google Ads, lands in your CRM, gets qualified, gets disqualified or progresses, and either becomes a customer or doesn’t. That outcome flows back into Google Ads as an offline conversion (data fed back from your CRM into the ad platform) and the bidding algorithm starts optimising toward the leads that actually close, not just the leads that fill out forms. Setting this up takes a half-day to a day depending on your CRM. The fact that most agencies don’t insist on it tells you what they actually optimise for.
It is the cleanest example on this list of agency incentives diverging from client outcomes.
A lead form fill is fast feedback. It happens within a click of the ad. It is countable, attributable, and easy to put on a slide. The agency’s monthly report can show “142 leads at $87 cost per lead” on the third of every month, the moment the previous month closes. That number renews retainers.
A closed lead is slow feedback. The B2B sales cycle for the kind of business that pays an agency $5,000–$30,000 a month is usually 30 to 120 days. The agency would have to wait two to four months to know whether a given month’s campaign actually produced revenue. By the time the truth surfaces, three more renewal cycles have happened. The structural incentive is to optimise for fast feedback even when fast feedback isn’t the right outcome.
Add the technical work involved — CRM integration, conversion mapping, value assignment, ongoing data quality checks — and you have a feature that is genuinely useful to the client, costly to set up, and quietly de-prioritised by an agency triaging where to spend its hours.
“What’s the closed-loop tracking setup, and which campaigns are producing leads that actually close?”
That question has a direct answer or it doesn’t. There is no middle.
It means the campaigns are being optimised toward the wrong target and have been the entire time. Every dollar of the bidding algorithm’s machine learning has been trained on “produce more form fills” and not “produce more customers.” In a healthy market this looks fine because the conversion-rate-to-customer is roughly stable. In a soft market or a market where lead quality has dropped, it’s catastrophic and invisible — the agency keeps reporting “steady cost per lead” while your sales team complains the leads are unworkable, and nobody on either side has the data to connect the two.
The fix is technical and not particularly hard. A half-day with a competent agency, a conversation with whoever owns your CRM, a working integration via Zapier or a native connector, and a thirty-day calibration period before the algorithm has enough closed-loop signal to optimise on. If your agency tells you it’s “not their lane,” the conversation about whose lane it is, is the actual conversation about whether to keep this agency.
]]>This is the single most diagnostic red flag on this list, because it costs nothing to produce a real report. The numbers exist in the platform. Pulling them takes ten minutes. Choosing not to put them in front of you is, almost always, a deliberate choice.
Three reasons, with the third being the one that should worry you.
Pre-formatted templates. Most agencies use a standard report template across all clients. The template was built once, by someone who optimised for “looks good in a presentation,” and it leads with totals because totals are easy to make look healthy. The agency doesn’t custom-build a template per client unless asked. Asking is enough to fix it on a half-decent agency.
Inertia. The report goes out. The client doesn’t ask for changes. The report keeps going out in the same format. Months go by. Years go by. Nobody is hiding anything, but nobody is showing you anything either. The cure is also asking.
Selection. This is the bad one. The numbers being shown are the numbers that look good. The numbers being omitted are the numbers that don’t. Month-over-month conversion comparison gets left out specifically because it would tell you that conversions have been slowly trending down for four months, while clicks and impressions have been trending up. The aggregate report shows a healthy account. The MoM trend on conversions tells the truth.
The simplest possible request: “Going forward, the monthly report needs a column for last month and a column for the same month last year, on conversions, cost per conversion, conversion rate, and ROAS. Per campaign. With percentage deltas.”
You don’t need their permission. You’re telling them what the report looks like now.
The bad answer is the agency telling you, gently, that they don’t want to put MoM comparisons in writing. There is no other reason to push back on a request this small. The work is ten minutes. The data is in the platform. The only cost is that some months the numbers will tell an inconvenient story and the agency will have to explain it on the call instead of glossing over it.
Some version of this conversation is the test of whether the relationship is worth keeping. If you ask for a real reporting format and the agency says yes and delivers it within one cycle, you have an agency that can be partnered with even if other things on this list are also true. If you ask and you get pushback, soft language, or a quarter of delays, the reporting format isn’t the actual problem. The reporting format is the symptom of an agency-client relationship in which the agency has decided what you’re allowed to see.
The hardest part of fixing this isn’t technical. It’s the moment in the next report meeting when the new MoM column shows red on conversions and the agency has to actually explain why — and you, sitting in your seat, have to be ready to ask the follow-up question. That’s the moment this whole site exists for. If you want help preparing for it, send me a question.
]]>The problem with Performance Max is that it deliberately reduces transparency. You don’t see ad placements the way you would in a Display campaign. You don’t see the full search terms report the way you would in a Search campaign. You see what Google chooses to show you, and Google’s incentives are not perfectly aligned with yours. If your agency hasn’t built guardrails around that fact, you’re paying for a blank check.
Two reasons.
It is the path of least resistance. Performance Max is what Google’s account reps push during their quarterly check-ins with agencies. The pitch is real and it works: less manual labor, broader inventory access, automated creative. For an agency managing a hundred accounts, an automated campaign type that produces decent-looking aggregate numbers with minimal touch is a profitability lifeline. The implicit trade is that some accounts get under-served because the optimisation is delegated to Google.
It produces conversions that look attributable. Performance Max is aggressive about claiming credit for conversions, including ones that would have happened through brand search anyway. If your brand is searchable and PMax is running without a brand exclusion, the campaign will report a strong cost per acquisition number that is partly your existing brand demand cannibalised. The agency reports a healthy CPA. The CPA is, in fact, partially fictional.
Three questions, and the order matters.
First: “What brand-name searches is PMax allowed to bid on, and how do you know?”
Second: “Show me the account-level negative keyword list and the category exclusions inside the PMax campaign settings.”
Third: “What does the most recent search-term insights review tell us, and when is the next one scheduled?”
It means PMax has been turned on, the surface-level numbers look fine, and nobody is doing the work to figure out whether the surface-level numbers are real. The phrase “adding too many exclusions can limit the algorithm” is the rhetorical move that signals an agency has decided not to bound the campaign — which is convenient for them and expensive for you.
The first thing I do on any audit involving Performance Max is check the brand exclusion. If it’s missing on a brandable business with any organic search demand, I can usually find ten to thirty percent of the campaign’s reported “conversions” were going to happen regardless. Strip those out and the actual incremental cost per acquisition is wildly different from what the agency has been reporting. That conversation tends to change the renewal meeting.
The fix is rarely killing PMax. The fix is bounding it: brand exclusion, account-level negatives, category exclusions, monthly search-term insights review with the findings shared in writing. Twenty minutes of setup, an hour a month of review. If your agency won’t agree to that, you’re not really running Performance Max — you’re running “whatever Google felt like buying that month.”
]]>Audience segmentation (splitting your traffic into groups based on what they’ve already done with your business) and using those groups as bid modifiers, exclusion lists, or dedicated campaigns is not advanced. It is table stakes. The fact that it isn’t happening on most accounts I audit is a tell about how the work is actually being done.
Same root cause as the broad match story: it’s labor that doesn’t show up in the report deck.
Setting up proper audience structure means defining segments (recent buyers, abandoned cart users, contact form submitters, customers, lookalike of customers, prospects who’ve hit your pricing page, prospects who haven’t), making sure each one is wired correctly to your tag setup or your CRM, building bid modifiers or separate campaigns for each, and revisiting the structure every quarter as your business evolves. It’s a half-day of work to set up cleanly and an hour a quarter to maintain. On a $5,000-a-month retainer, an agency that takes their margin seriously has to triage what gets done. Audience segmentation often loses the triage.
There’s also a competence issue. Doing this well requires the analyst to ask uncomfortable questions about your funnel: where do prospects drop off, what is the typical sales cycle, what does intent actually look like in your business. Junior analysts without account-management training don’t ask those questions, so the campaigns get built without that input, so audiences get treated as one bucket.
“Show me the audience segmentation strategy currently running on our search campaigns. I want to see the segments, whether each one is set to Observation or Targeting, and what the bid adjustments are.”
A real agency will send you a one-page document within a day. A struggling agency will send you a screenshot of the Audiences tab and call it a strategy.
The bad answer is half-true, which is what makes it a red flag instead of a lie. Smart Bidding (Google’s machine-learning auction bidding) does optimise across audiences if you give it audience signals to optimise on. But it can only optimise for what it can see, and the difference between an existing customer and a cold prospect — or between someone who hit your pricing page yesterday and someone who came in via a content marketing search a month ago — is not visible to the algorithm unless someone has actually wired those segments in.
What the bad answer reveals is that the agency has either delegated thinking to Google’s automation or is using “Smart Bidding” as the rhetorical justification for not having done the segmentation work. Either way, it’s the same outcome for you: cold traffic and warm traffic getting bid the same, your existing customers getting served paid ads they don’t need, and a budget that’s averaging itself into mediocrity.
The fix is the half-day of work I described above. If the agency won’t do it as part of the existing retainer, that’s a conversation about scope. If they will but it doesn’t actually get done within three weeks, that’s a conversation about whether the retainer is staffed at all.
]]>The way you tell is by looking at the search terms report (the list of actual queries people typed before clicking your ad) and at your negative keyword list (the list of queries you’ve told Google not to match against). If the search terms report is full of queries that have nothing to do with what you sell, and the negative list hasn’t had a meaningful update in three months, you’re funding the long tail of Google’s imagination.
Three reasons, in roughly this order.
It scales. Broad match plus Google’s automated bidding produces volume on demand. Ask your agency for more clicks or more leads and broad match is the lever they pull, because it’s a setting change rather than the slow work of writing better ads, building tighter campaigns, or improving landing pages. The volume comes fast. The relevance is the cost, and the cost is invisible to you unless you go looking for it.
It hides labor. Maintaining a negative keyword list is unglamorous, ongoing work that has to happen every week or two on a real account. It takes ten to forty minutes per pass depending on volume. Multiply that across an agency’s book of clients and you’re looking at significant unbillable hours. Broad-match-without-pruning is what happens when an account is too small to justify that labor, but the agency doesn’t want to renegotiate the retainer downward.
It looks fine in the report. The numbers Google shows you for broad match campaigns — clicks, conversions, ROAS — are real. They just leave out the half of the spend that bought traffic that will never convert. As long as the agency reports on aggregate performance and not on search-term-by-search-term spend, the bad half is statistically invisible.
Two questions, in order. The second one is the test.
First: “Send me the top fifty search terms by spend over the last 90 days, with conversion rate and cost per lead next to each.”
Then, after they send it: “Walk me through the negative keyword cadence. How often is the list reviewed, who does it, and what was added in the last 60 days?”
It means nobody is doing the work. The phrase “negatives are added when we see issues” is functionally equivalent to “we don’t look unless something breaks badly enough to show up in the aggregate report,” which on a smaller account can be never. “The system is working as designed” is the part where you should pay attention — whose design, exactly?
The fix is rarely as dramatic as switching agencies. It is almost always a 90-minute conversation in which you ask for a 30-day negative-keyword cleanup, a written cadence going forward (“every two weeks, list goes to you in writing”), and a target for broad-match share that comes down by ten or fifteen percentage points over the next quarter. A real agency will say yes and have it done in three weeks. An agency that says yes and then nothing happens is telling you something more important than any report ever will.
]]>It is rarely malicious. It is structural.
Click and impression metrics are abundant, immediate, and almost always trending up if budget is steady. Lead metrics are scarcer, slower, and brutally honest. Account managers are trained to lead with the metrics that pattern-match to growth, because the report meeting is the one moment of the month where renewal risk gets calibrated. If the first slide looks like a win, the rest of the meeting is graded against a forgiving baseline.
There is also a competence ladder underneath this. Senior practitioners are comfortable opening with leads and conversion rate (the percentage of clicks that become leads — the only number that connects ad spend to your business). Junior practitioners aren’t, because if leads are flat the senior practitioner has a thoughtful explanation and the junior one has “results take time.” Leading with clicks lets a junior team look fluent.
Ask the question that puts the scoreboard back on your business: “Walk me through leads, cost per lead, and conversion rate for the last 90 days. What changed and why?”
Listen for what comes after “and why.” That is the entire test.
It means one of two things, and they are about equally common.
The first: the person on the call doesn’t actually know what happened. Either they didn’t look before the meeting, or they did look and don’t understand what they saw. This usually points to a junior analyst running the account with thin oversight from above. Not catastrophic, but you are paying senior rates for junior work.
The second: the person on the call does know, and the news is bad enough that they are managing the moment instead of telling you. This is more serious. If your agency reflexively softens bad news instead of naming it, every problem you have is going to surface a month or two later than it should — which is the difference between a recoverable quarter and an unrecoverable one.
Either way, the move is the same: ask for the leads-first version of the report, in writing, before the next meeting. If they push back on producing it, that’s the answer.
]]>