Most agencies lose clients without seeing it coming. One month everything looks fine — the work is going out the door, invoices are being paid, nobody’s raised a formal complaint. Then a brief email arrives: “We’ve decided to move in a different direction.” Six months of retainer income evaporates, and in hindsight the warning signs were all there.
A client health score is a structured way to surface those warning signs before they become an exit letter. It translates relationship signals — response times, ticket volume, invoice payment behaviour, engagement with deliverables — into a single number you can act on. Done well, it turns client retention from an intuition game into a managed process.
This guide walks through how to build a health score system that actually fits the way a digital agency operates. No SaaS vendor assumptions, no enterprise complexity — just a practical framework you can implement this month.
Why Client Health Scores Matter More Than NPS
Net Promoter Score surveys have become the default retention metric for a lot of agencies. They’re easy to send, they produce a tidy number, and they give you something to report to the management team. The problem is that they’re retrospective and infrequent. By the time a client gives you a 4/10 on a quarterly NPS survey, they’ve been unhappy for weeks.
A health score is live. It updates as behaviour changes, not as survey responses arrive. A client who normally pays in 14 days and suddenly stretches to 45 days has changed their behaviour — that’s a signal. A client who used to reply to your account manager within a few hours and now takes three days to answer a straightforward question has also changed their behaviour. Neither of these events would register on a quarterly NPS survey, but both of them are telling you something important.
The other advantage of a health score is that it’s comparative. You can look across your entire client base and rank who needs attention, rather than relying on whoever shouted last. For a 10-person agency managing 30 clients, that’s the difference between firefighting and running a proper client success programme.
Choosing What to Measure: The Five Signal Categories
A health score is only as good as its inputs. Use the wrong signals and you’ll build a metric that looks busy but doesn’t predict churn. Most agency-client relationships can be described by five categories of signal, each of which captures a different dimension of how well the relationship is actually functioning.
1. Financial signals
How clients handle money tells you a lot about how they feel about the relationship. Days to pay (how long invoices sit before settlement) is one of the clearest leading indicators: a client who was consistently paying in 10 days and now sits at 35 days is either cash-strained or emotionally disengaged. Outstanding invoice value relative to retainer size, disputed invoice frequency, and retainer renewal behaviour all belong here too.
2. Communication signals
Average response time to account manager emails or messages is measurable, and changes in that average are meaningful. So is inbound communication frequency — a client who used to send you 10 messages a week dropping to two is often a client who’s quietly disengaging. Track both volume and responsiveness. If your CRM or shared inbox logs message timestamps, you already have the raw data.
3. Support and issue signals
Ticket volume per month, average ticket severity, and the ratio of complaints to general requests all matter. A spike in support tickets isn’t necessarily a bad sign — it can mean active, engaged usage of a product you’ve built for them. But a spike in high-severity tickets, or repeat tickets about the same issue, indicates friction that isn’t being resolved. Also watch for clients who go completely silent on support: sometimes that means everything’s perfect, but sometimes it means they’ve given up raising issues because they don’t expect action.
4. Project engagement signals
Does the client review and approve deliverables promptly? Do they attend scheduled calls, or do they repeatedly reschedule? Do they provide brief and assets on time, or does every project stall waiting for client inputs? Delayed approvals and poor brief quality aren’t just operational problems — they’re signals about how much the client values the work.
5. Relationship depth signals
This is the qualitative layer that pure data misses. Has the client introduced you to anyone else in their organisation? Have they expanded scope in the past six months, or has it contracted? Did they renew without negotiation, or did they push for a discount? These signals require a human to record them, but they’re often the most predictive of all.
Building the Scoring Framework
The simplest workable health score uses a 0–100 scale. Assign each signal category a weight that reflects its importance to your agency’s relationships, then score each category from 0–100 and take a weighted average. Here’s a starting framework you can adapt:
Starting weight allocation
Invoice payment speed, outstanding debt, retainer renewal
Response time, message volume, engagement consistency
Ticket volume, severity, repeat issues
Approval turnaround, call attendance, brief quality
Scope growth, referrals, executive access
A few important notes on weighting. These percentages work as a starting point, but your agency’s most predictive signals will differ. If you’ve lost clients before, go back and look at the data for the three months before each cancellation. Which category deteriorated first? That’s your most predictive signal, and it should carry more weight.
Also consider whether your client base is homogeneous. An agency that runs pure retainers will weight financial signals differently than one that works on project-by-project billing. Adjust the framework to reflect your actual business model, not some theoretical one.
Turning Raw Signals Into Sub-Scores
Each category score needs to map raw data onto a 0–100 scale. The principle is simple: set a benchmark for what “healthy” looks like, and score relative to that benchmark. Here’s how to handle each category concretely.
Financial sub-score: Start with invoice payment speed. If your standard terms are 30 days and the client pays in 15, that’s 100. Pays on the 30th day, that’s 70. Hits 45 days, score drops to 40. Goes beyond 60 days, score is below 20. Adjust these thresholds to your payment terms. Add a penalty for any invoice currently in dispute, and a bonus for clients who have auto-renewed without renegotiation.
Communication sub-score: Calculate a rolling 30-day average response time for inbound messages from your account manager. Compare it to the client’s own historical baseline — that’s more meaningful than a universal benchmark, since some clients are naturally slower communicators. If the rolling average is within 20% of their historical baseline, score 80+. If it’s deteriorated by 50% or more, score below 40. Also track whether communication volume has halved over the past 60 days.
Support sub-score: Score based on ticket severity-weighted volume. A month with zero tickets or only low-severity tickets scores high (80+). A month with a medium-severity unresolved ticket scores around 60. Multiple high-severity tickets or a repeat of a previously resolved issue should push the score below 40. The key is to weight unresolved issues more heavily than resolved ones — a complaint that was handled well is better than one that was never raised.
Project engagement sub-score: Track approval turnaround times across projects. If the client reviews deliverables within 48 hours consistently, score 90+. Three to five business days is mid-range (60–75). Over a week becomes a drag on the relationship score. Separately, log call attendance: a client who’s attended 100% of scheduled calls in the past quarter gets a bonus, one who’s cancelled two out of the last three is a concern.
Relationship depth sub-score: This is the manual one. Your account managers should update this quarterly. Has the client expanded their scope in the last six months? +20 points. Have they referred someone to the agency? +15 points. Have you lost access to the senior decision-maker (now speaking only to a junior contact)? -20 points. Has scope contracted? -15 points. Start at 70 and adjust from there.
Setting Thresholds and Intervention Triggers
A health score without action triggers is just a number. Define what each score band means and what it requires from your team.
Healthy
Standard account management cadence. Quarterly review meeting. Identify upsell or expansion opportunities. Document what’s working so you can replicate it.
Watch
Amber flag. Increase check-in frequency. Have the account manager make a proactive call — not to discuss a specific issue, but to ask how things are going. Investigate which sub-score has dropped and address the root cause.
At Risk
Escalate to senior account director or agency principal. Schedule a formal relationship review — don’t frame it as a problem meeting, frame it as a strategic planning session. Look for a meaningful gesture (extra deliverable, fee adjustment, executive involvement).
Critical
Treat as an active retention emergency. Principal-level ownership. Identify the two or three specific issues causing the collapse and address them directly. Prepare a recovery plan. If the client is under contract, review terms. If month-to-month, assume cancellation is possible within 60 days and plan accordingly.
The most important discipline here is acting on the Watch band. It’s tempting to treat a 65 as “probably fine” — the client hasn’t said anything, they’re still paying, there’s no fire. But a 65 that’s been trending down from 80 over six weeks is a very different situation from a 65 that’s been stable for three months. Track the direction of movement, not just the current value.
Implementing It Without Drowning in Spreadsheets
For a small agency, the simplest implementation is a shared spreadsheet updated weekly. Build one row per client, score each sub-category against the criteria you’ve defined, and use conditional formatting to colour-code the output. It takes about 90 minutes to set up and 20 minutes a week to maintain once you have the habit. That’s a reasonable starting point — the goal is to start somewhere, not to build a perfect system.
The practical constraint is data collection. The financial signals are easy if your invoicing lives in a system with reporting — you can pull payment lag data directly. Communication signals are harder if you’re not logging message timestamps anywhere. This is where using an integrated platform rather than stitching together separate tools pays dividends: if your CRM, shared inbox, support tickets, and invoicing all live in the same system, health score data is a query away rather than a manual aggregation exercise.
In Marque CRM, client health scores are built into the platform and update automatically based on the signals your team is already generating through normal work. Invoice payment behaviour feeds in from the billing module. Support ticket volume and severity feed in from the helpdesk. You still input the relationship depth signals manually — a human judgement that should stay with your account managers — but the quantitative signals are automated. The result is a health score that costs roughly five minutes per client per week to maintain, rather than 20.
Whether you build it in a spreadsheet or use a purpose-built tool, the discipline of reviewing your lowest-scoring clients in your weekly or fortnightly management meeting is what makes it useful. A health score that no one looks at is just data.
Five Mistakes Agencies Make With Health Scores
Using too many signals. More signals sound more rigorous, but they add noise as often as they add signal. Start with the five categories above and eight to twelve concrete metrics. You can add more sophistication once you’ve validated the model against actual churn events.
Weighting all signals equally. An agency that primarily does retained SEO will have different predictive signals than one doing project-based web development. Take the time to weight signals based on your own churn history, not a generic framework.
Updating scores too infrequently. Monthly updates miss the moment when a score is deteriorating. Weekly updates are the minimum useful cadence for a score you want to act on in time.
Not adjusting for client size. A client paying £15,000 a month taking 45 days to pay an invoice is a different situation from one paying £500 a month doing the same thing. Consider whether your score should incorporate revenue weight so that your highest-value relationships receive proportionately more attention in the Watch band.
Using the score to rank clients rather than understand them. The ranking is useful for prioritising attention, but the score itself doesn’t tell you what action to take. Always drill into which sub-score has moved and why. A drop in the financial score needs a different response from a drop in the project engagement score — conflating them leads to the wrong intervention.
The Retention Maths: Why This Is Worth Your Time
Client acquisition costs for a digital agency are significant. Depending on your lead generation model, winning a new retained client — from first contact to signed contract — typically costs between one and three months of their retainer value in staff time and business development overhead. A client paying £3,000 a month on retainer represents somewhere between £3,000 and £9,000 in acquisition cost.
Saving that client through early intervention — even if it requires two hours of senior account director time and an additional deliverable worth £500 — is still a dramatically better outcome than losing them and starting the acquisition cycle again. The maths of retention almost always beats the maths of acquisition, which is why the most profitable agencies tend to have churn rates below 8% annually.
A well-implemented health score system running across 20 clients will typically surface two or three at-risk relationships at any given time that the team wasn’t actively aware of. If you save even one of those through proactive intervention, and that client has a 12-month average tenure at £2,500 a month, you’ve generated £30,000 in retained revenue that would otherwise have walked out the door. The upfront investment in building the system pays for itself on the first churn event you prevent.
Health scores also improve the quality of your client conversations. When an account manager enters a quarterly review knowing exactly where the score stands and which signals have moved, they can have a direct, productive conversation rather than a vague check-in. Clients consistently describe feeling more confident in agencies that demonstrate they’re paying attention — the score isn’t just an internal tool, it feeds into how professionally you manage the relationship.
Getting Started This Week
Building a client health score system doesn’t require a big infrastructure project. Define your five signal categories. Assign weights based on your knowledge of your own churn history. Score each existing client this week using whatever data you have to hand — even rough scores are better than none. Set a calendar reminder to update them weekly and review the bottom quartile in your next management meeting.
The agencies that retain clients well don’t do it through superior luck or unusually loyal clients. They do it through systematic attention — knowing which relationships need intervention before those clients have mentally checked out. A health score is the mechanism that makes that possible at scale, without requiring every senior person in the business to hold 30 client relationships in their head simultaneously.
If you want to explore how Marque CRM’s built-in health scores and client management tools work in practice, take a look at the full feature set or start a trial below. The score updates automatically as your team works — no additional data entry required.