The most expensive thing in marketing isn't a campaign that fails. A failure has an owner, a post-mortem, a fix. The expensive thing is a site that underperforms quietly, a few points below what it should do, with no owner, no alarm, and no line in any report that names the cause. It just leaks, month after month, and everyone signs off the numbers because the numbers look fine.
You know this pattern from the other direction. When a campaign tanks for an obvious reason, wrong audience, weak offer, bad creative, someone owns the miss and someone fixes it by Friday. But when a landing page converts at four percent instead of the six it's capable of, nothing lights up. Four percent isn't a failure. It's just a slightly disappointing normal that nobody has a reason to question. So it survives every review, forever.
Here are the six leaks I find most often when a team asks me to look at a site that's "converting fine". None of them show up as a red number. All of them cost real money.
A leak with no owner is invisible by design. Nobody's job title contains the words "the gap between what this converts and what it could".
What's costing you, and why nobody caught it
The message-match gap
Your ad promises a specific thing to a specific person. They click, and the page they land on opens with "Welcome" and a value proposition written for everyone. The scent is broken. In the half-second it takes to notice the page isn't obviously about the thing they clicked for, a chunk of them are gone, and your media budget paid for every one of those bounces.
This is the single most common leak I see, and it hides because the two halves are owned by two different people. The paid team owns the ad. The web team owns the page. Nobody owns the join, which is precisely where the money falls through. The fix costs nothing but attention: the page's first screen has to answer the promise the click made, in the words the click used.
Why nobody flagged it: both halves look fine in isolation. The leak only exists in the seam between two teams' dashboards.
Form friction that "tested fine years ago"
The form has nine fields because, at some point, sales asked for company size, and someone in ops wanted a phone number, and a field for "how did you hear about us" got bolted on for a report nobody reads any more. Each field felt reasonable the day it was added. Nobody has ever added up what the whole set costs at the point of submission.
Every non-essential field is a small tax on completion, and the taxes compound. I've watched a required phone number alone drop form completion by double digits, because people who'll happily give an email get twitchy about a call they didn't ask for. The question is never "is this field useful". Everything is useful to someone. The question is "is this field worth the conversions it costs", and almost nobody runs the sum.
Why nobody flagged it: each field was added by a different person for a defensible reason. The cost is only visible when you total them, and totalling them isn't anyone's job.
The AI-search blind spot
Your funnel measures the visit onward. But a growing share of high-intent research now finishes inside an AI answer before your site is ever opened. Someone asks ChatGPT or Perplexity to compare the options in your category, and either you're in that answer, cited and quoted, or you're not in the running at all, and you never see the visit that didn't happen.
This is demand lost upstream of everything you measure, which makes it the hardest leak to notice and one of the fastest-growing. It's also winnable right now, because most of your competitors are still optimising for blue links while their buyers have quietly moved into the answer box. At Joblogic I got us to a 45% share of voice in AI answers for our category. That share is a moat while it lasts, and it won't stay open for long.
Why nobody flagged it: it doesn't appear in your analytics at all. You can't see a visit that an AI answer prevented, so the loss is perfectly silent.
Testing theatre
The team runs A/B tests, so the box marked "optimising" is ticked. Look closer and the tests are underpowered, called the moment they wander into significance, and aimed at the safe stuff: button colours, headline tweaks, the fiddly edges. Meanwhile the offer, the page structure and the proof, the things that actually move a rate, go untouched because they feel too big to test.
This is worse than not testing, because it manufactures confidence. Everyone believes the site is being improved, so nobody asks the harder question. A real test is powered to a decision, left alone until it's earned an answer, and pointed at something that would matter if it won. Fewer, braver tests beat a calendar full of timid ones every time.
Why nobody flagged it: activity got mistaken for progress. A full test backlog looks like rigour, so no one interrogates whether any single test could have changed a decision.
The page-speed tax with no line item
The site got slow the way everyone's site gets slow: a tag here, a plugin there, a marketing script the agency added and nobody removed, a hero image nobody compressed. Each addition was signed off on its own merits. The cumulative weight, and the conversions it quietly sheds on mobile, was never anyone's decision, because it was never a single decision.
Slow pages convert worse. That's settled. But "page speed" almost never appears in a conversion report as a cause, so a page can bleed for years without the leak ever being named. I've pulled this one apart in detail in the piece on what your slowest page is really costing you. The short version: performance is a conversion lever wearing an engineering costume, which is exactly why it goes unowned.
Why nobody flagged it: it accreted one approved script at a time, and it lives in the gap between marketing and engineering where neither side counts it as theirs.
The fix that needs a dev ticket
Here's the leak behind the other five. Your team can often see the problem. What they can't do is ship the fix, because the CMS is rigid, the change needs a developer, the developer is on the roadmap until Q3, and the ticket dies in a backlog. So the known problem persists, not because nobody spotted it, but because the cost of shipping the fix is higher than the pain of leaving it.
When shipping a change is hard, changes stop happening, and a site that can't change is a site that can't improve. That's the real tax of platform lock-in, and I've written about escaping it here. The point for this piece: half of "we know but we can't get to it" is a delivery problem dressed as a priority problem.
Why nobody flagged it: it's not framed as a conversion issue at all. It's filed under "the CMS is a pain", which sounds like grumbling rather than money on the floor.
The fix starts with a name and a number
If you take one thing into your next planning meeting, take this: an unowned leak stays unowned until somebody attaches a number to it. "The landing page could convert better" gets nodded at and shelved. "We're spending forty grand a quarter driving traffic to a page whose first screen contradicts the ad" gets a meeting of its own. Same problem, different fate. The number is what turns a shrug into a priority, and the number is almost always sitting in data you already pay for.
So pick the leak you suspect is worst, size it roughly, and give it a single owner, even a temporary one, whose only job is to close that gap. Not the paid team, not the web team, one person accountable for the seam between them. You'll be amazed how fast a problem gets solved once it stops being everybody's and becomes somebody's. Most of these leaks aren't hard to fix. They're hard to assign, and assigning them is the part that's genuinely within your gift this quarter.
What actually changes the numberWhat a senior pair of hands does differently
None of these leaks is exotic. There's no clever growth hack in the list, and that's the point. They persist not because they're hard to understand but because each one lives in a seam, between two teams, between two dashboards, between spotting a fix and being able to ship it. Seams are where accountability evaporates, and evaporated accountability is expensive.
What experience buys you is speed of diagnosis. Somebody who's fixed these patterns across enough sites walks into the same data your team already has and reads it faster, because they know which finding is the cause and which is noise. I spent years on this at Joblogic, where a sign-up flow went from a 3% conversion rate to 17.5%, not through one big idea, but by finding and closing exactly these kinds of quiet leaks, one seam at a time. Twenty years across Dropbox, Oracle and ServiceNow before that was the same job in different clothes: the rate is never bad for one dramatic reason; it's a few points short for six boring ones.
Underperformance without an owner is the most expensive line in the budget, precisely because it never appears in the budget.
The other thing seniority buys is the willingness to say the offer is the problem, not the button. A junior tests what's safe to test. Someone who's been on the hook for a number will tell you, on day one, that no amount of headline tweaking fixes a page whose proof is thin and whose promise doesn't match its traffic. That conversation is uncomfortable, which is why the people who can have it are worth having in the room. If you want the constructive flip side of this piece, the shape of a page that gets these calls right, that's the anatomy of a landing page that converts.
The leak nobody owns is the one that's been costing you longest. Naming it is most of the fix.
If you'd like a straight read on which of the six is quietly costing you, that's what a Zenlio conversion review is. No retainer, no theatre, just the prioritised list I'd work from if the site were mine.
