Let me paint a picture you might recognise. You're three months into a six-month engagement. The agency that was electric in the pitch - sharp, proactive, full of ideas - has gone quiet. Deliverables are arriving late, or arriving on time but not quite right. You've had a couple of conversations about it internally, probably using phrases like "it's not where we need it to be" and "let's give them another few weeks." Nobody's panicking yet. But the doubt has settled in.
That doubt is doing more damage than you think. Right now, you're in the most commercially consequential phase of the entire engagement - and most managing partners don't realise it until it's too late. The decisions you make in the next two to four weeks will determine whether this relationship recovers, or whether you're back at square one, running another procurement process, losing everything the agency has learned about your business, and explaining to your board why the digital programme is six months behind.
The instinct I hear most often? We just need to have a hard conversation with the agency. They know things aren't working. And look - I get it. You're a senior professional. You've managed difficult relationships before. A frank conversation feels like the right move.
But a hard conversation without a diagnosis is just venting. And venting doesn't fix anything. Before you pick up the phone, you need to understand what's actually broken. Because different failures require fundamentally different responses, and treating them all the same way is how recoverable relationships die and unrecoverable ones get prolonged.
In 25 years of working alongside - and occasionally cleaning up after - agency relationships, I've seen four distinct failure categories. They overlap, sometimes messily, but getting clear on which one is dominant changes everything about what you do next.
Communication failure is the most common and the most fixable. The agency isn't proactively telling you what's happening. There are long silences between formal reviews. Problems surface after the fact. You find yourself chasing for updates that should have come to you unprompted. The work itself might be decent - but you can't tell, because you're always three steps behind. I sat in a steering committee last year where the client sponsor said, "I genuinely have no idea whether this project is on track or not." The agency lead looked blindsided. They thought the work was going well. It probably was. But the client's experience of the engagement was anxiety, not progress.
Capability failure is harder to fix and harder to admit. The agency doesn't have the specific technical or creative skills the scope demands. The symptoms: repeated deliverables that miss the standard, rework cycles that feel endless, and - this is the tell - the agency's founders or directors suddenly appearing in the day-to-day work. When the principals start doing the job their team was supposed to handle, that's not a sign of commitment. It's a sign the bench isn't deep enough.
I saw this play out badly at a mid-sized financial services firm a few years back. They'd hired a well-regarded digital agency to rebuild their client portal - a proper job, Salesforce integration, the works. About six weeks in, the agency's co-founder started joining the weekly calls. Then the technical director. By week ten, the original delivery team had essentially been replaced by the people who'd won the pitch. The client was flattered at first. Then they realised the original team had been out of their depth from day one, and the founders were firefighting rather than leading. The project finished four months late and about £80k over budget.
Commitment failure is the one that makes people properly angry. The energy that characterised the pitch has evaporated. The senior strategist who led the chemistry meeting hasn't been seen since week two. Response times have stretched. You're dealing with people you've never met who don't seem to know the brief. This is the classic bait-and-switch, and honestly, it's endemic in the agency world.
Cultural fit failure is the subtlest and often the most intractable. The agency's working style is fundamentally incompatible with yours. They want to move fast and iterate; you need documentation and sign-off. They're informal and spontaneous; your governance structure requires structure. Every interaction feels like friction, and both sides are spending more energy managing the relationship than delivering the work. Neither party is wrong, exactly. But the mismatch is real.
Most deteriorating relationships involve more than one of these. A capability gap often creates a communication failure as the agency starts hiding the problem. A commitment failure can look like a cultural fit issue if you're not looking closely. The diagnosis doesn't have to be perfectly clean - but you need to identify the primary driver, because that determines whether recovery is realistic.
And one more thing, because fairness matters here. Clients contribute to agency relationship failures. Constantly. Unclear briefs, shifting priorities, slow feedback, the internal stakeholder who keeps changing direction without telling the project lead - I've seen all of it. Before you diagnose the agency, take an honest look at your side of the table. Not to excuse poor performance, but because a recovery plan that only addresses one side of the problem isn't a recovery plan. It's a blame exercise.
Right. You've done the diagnosis. You know what's broken, roughly. Now comes the bit everyone dreads.
I won't pretend this is easy. In professional services, the agency might be someone you've known for years. The account director might be a mate. The founder might have been introduced by a board member. The personal dynamics make it genuinely uncomfortable, and anyone who tells you "just be direct" has probably never had to do it while protecting a relationship they might need to preserve.
But the conversation has to happen. And the quality of it determines whether recovery is even possible.
Three principles that I've seen work:
Name the specific problem, not the general feeling. "Things aren't working" gives the agency nothing to act on. "In the last three deliverables, the UX wireframes have been well below the standard we agreed - here are three specific examples, and here's what we expected versus what we received" gives them something concrete. It also stops the conversation drifting into defensive generalities.
Ask for their diagnosis before proposing your solution. This is counterintuitive - you've done the work, you think you know what's wrong. But saying "I want to understand why these issues are occurring before we decide how to address them - what's your view?" does two things. It gives the agency a chance to be honest, which tells you a lot about whether recovery is feasible. And it surfaces information you might not have. I remember one conversation where the agency's diagnosis was "your internal stakeholder has changed the brief four times in six weeks and we've been absorbing it without flagging it." They were right. The problem was real, but the cause was shared.
Set a specific timeline with a specific review. Not "let's see how the next few weeks go" - that's just kicking the can. "I'd like to see these three specific changes in the next four weeks, and I'm proposing we have a formal review on [date] to assess whether we're on a better footing." Concrete. Measurable. Impossible to fudge.
So both sides have agreed to try. Good. But "trying harder" isn't a plan. The relationship deteriorated because the original engagement structure didn't prevent it from deteriorating. Doing more of the same, with better intentions, won't fix that.
A structured recovery needs three things the original engagement probably didn't have.
Written expectations that both parties treat as binding. Not the original statement of work - that ship has sailed. New, specific commitments about deliverable quality, communication frequency, and decision-making process. Who reviews what, by when. What "good" looks like for the next deliverable, described in enough detail that both sides can independently assess whether it's been met.
I know this feels bureaucratic. It is. That's the point.
The pushback I always get here is "we don't want to turn this into a legal exercise." Fair enough. But ambiguity is what got you here. The original brief probably felt clear to everyone in the room. It wasn't. The new commitments don't need to be a contract - they need to be specific enough that there's no room for two different interpretations of whether something has happened or not.
Shorter review cycles. Weekly structured reviews for the recovery period, not milestone reviews. I've seen plenty of agency relationships where the monthly review was the only honest conversation in the entire month - and by the time problems surfaced there, they'd been compounding for weeks. A weekly rhythm catches small problems while they're still small. It also creates a paper trail, which matters if you end up needing to exit.
Three specific, measurable improvements. Not "better communication" - that means different things to everyone. Something like: "A written progress update sent by Friday of each week for the next eight weeks, confirmed received by the client lead." Or: "Design deliverables reviewed by the agency's creative director before client submission, with a sign-off noted on the cover sheet." Specific enough that at the four-week review, you can look at the evidence and say definitively whether it's happened or it hasn't.
The reason people resist this level of specificity is that it feels like you're treating a professional relationship like a performance improvement plan. You are. That's what a recovery is.
Sometimes the right answer is to cut your losses. I want to be honest about that, because there's a bias - in professional services especially - toward preserving relationships. Changing agencies feels like failure. It's disruptive, expensive, and politically awkward. So people persist with relationships that are clearly broken, hoping the next conversation, the next sprint, the next deliverable will be the turning point.
Here are the signs that recovery isn't coming.
The agency has offered no credible diagnosis. Just reassurances. "We hear you, we'll do better" without any explanation of why the problems occurred means there's no mechanism for sustained improvement. You're relying on willpower, and willpower fades.
Both sides are spending more time managing the relationship than managing the work. When every meeting starts with twenty minutes of relationship maintenance before anyone discusses deliverables, you're in palliative care, not recovery.
Your trust in the agency's competence has been so thoroughly eroded that you can't objectively evaluate their work anymore. This one's painful to admit, but it matters. If a deliverable arrives and your first instinct is suspicion rather than assessment, the relationship has passed a point of no return. Even if the next piece of work is excellent, you won't experience it that way.
The capability gap you identified requires senior re-staffing that the agency has either refused or promised and not delivered. If the team can't do the work and the agency won't change the team, nothing else matters.
If you've reached the decision to exit, the temptation is to rip the plaster off. Don't. A badly managed exit is almost as damaging as a badly managed relationship.
Protect your IP and documentation first. Before exit conversations begin - not after - make sure all work product, documentation, source code, and access credentials are transferred to you. Once you've communicated the decision, your leverage changes. I've seen firms lose access to staging environments, design files, and analytics accounts because they announced the exit before securing the assets. Get the keys before you hand back the car.
Get the transition plan in writing. What the agency will hand over, in what format, on what timeline. This should be agreed before the formal exit conversation, or at least negotiated as part of it. Don't assume goodwill will carry the handover - it might, but you don't want to depend on it.
Consider a managed handover rather than a clean break. Where the relationship allows it, a two-to-four-week transition period where the outgoing agency briefs the incoming partner is less disruptive than starting from scratch. The outgoing agency knows things about your systems, your stakeholders, and the decisions made along the way that aren't written down anywhere. That institutional knowledge has value even when the relationship doesn't.
The managing partners I respect most - the ones who consistently get good outcomes from agency relationships - share one characteristic. They treat the first sign of trouble as a signal to engage, not a signal to wait. They have the awkward conversation at week three, not week twelve.
That's not because they're confrontational. It's because they understand that the cost of delay in a deteriorating agency relationship compounds daily. Every week you spend hoping things will improve is a week of budget spent on substandard work, a week of internal trust eroding, a week closer to the point where recovery becomes impossible.
So if you're reading this and recognising your own situation - even a bit - the time to act is now. Not next month. Not after the next deliverable.
If you want an independent view on whether your agency relationship is recoverable before committing to either a recovery or an exit, we offer a project recovery assessment. We've also built a relationship diagnosis checklist covering the four failure categories with observable symptoms for each, a go/recovery/exit decision guide, and the recovery plan structure I've described here. It's free and takes about ten minutes. Most people find it gives them a clearer picture than another internal conversation about whether things might get better on their own.