Before we built Avara, we spent about six months talking to agency and consultancy operators about their client churn. Not about what they thought was causing it, but about what they knew was in the data that they had never analyzed.
The consistent answer was meeting notes and survey verbatims. People were confident that those sources contained signals. They were equally confident that nobody was reading them systematically. We asked: if you went back and read the notes from the three months before your last major non-renewal, what would you expect to find? The answers were remarkably consistent regardless of firm type or size.
Here is what that analysis revealed about the pre-churn signal pattern in agency and consultancy client relationships. This is based on qualitative research across early conversations with agency operators, not a formal quantitative study, and we are presenting it as directional insight rather than precise statistics. The patterns are real; the exact numbers are approximate.
The CRM tells almost nothing about relationship quality
Start with what the CRM shows in the 90 days before a non-renewal. In most cases: nothing that looks different from a healthy account. Meeting frequency is maintained. Email response times are within normal range. The satisfaction score has not dropped dramatically. The contract date is visible but no different from any other account with an upcoming renewal.
This is what makes the agency churn problem different from SaaS churn. In SaaS, declining usage genuinely predicts non-renewal, and that signal shows up in product telemetry. In professional services, clients can fully maintain the operational motions of the engagement while internally having already decided not to renew. They are not disengaging from the meetings. They are disengaging from the relationship. Those are different.
The CRM data structure is built to record the former (meeting happened, email sent, survey submitted). It has no field for the latter (what the client said in the meeting, whether the survey response contained genuine forward-looking investment, whether the quality of interaction has been declining for two months).
The notes contain the signal: three patterns that show up before non-renewals
When we reviewed notes from pre-churn periods in the accounts we had access to through early conversations, three patterns showed up consistently across different firm types.
Specificity collapse in verbatims. In the months before non-renewal, client survey responses tend to become noticeably less specific. Responses that previously named particular deliverables, referenced team members, or connected work product to business outcomes shift to generic positive statements ("team is professional," "deliverables are on track") or brief acknowledgments. The satisfaction scores may not change, but the informational density of the response changes significantly. Clients who are genuinely invested tend to write more and more specifically over time. Clients who are moving toward disengagement write less, more generically, even when they still score positively.
Strategic topic withdrawal. In meeting notes, a related pattern: the client stops raising strategic topics. In healthy engagements, clients often bring adjacent problems to their agency or consultancy, topics that are not strictly in scope but reflect their trust that you can think about their business with them. In the six to eight weeks before non-renewal, this tends to stop. The meeting notes become narrower. Topics are operational, within-scope, completion-focused. The client is not bringing you into their strategic thinking anymore. That is a meaningful shift, and it shows up consistently in the notes before it shows up anywhere else.
Timeline and ownership language changes. A more subtle pattern in meeting notes: the language around ownership and timelines changes. In healthy accounts, clients use "we" language about joint work and tend to reference future deliverables or next phases naturally. In accounts approaching non-renewal, the language often shifts: deliverables are discussed in more contractual terms, timelines become important in a closure-focused way ("just need to make sure this is wrapped up by end of quarter"), and the planning horizon in conversation shortens. This is hard to notice in any single note. As a pattern across five or six notes, it is fairly clear.
What the satisfaction scores were doing at the same time
In the accounts we looked at, satisfaction scores were almost uniformly not predictive of the non-renewal in the 90-day window. Many accounts that churned had stable or even improving scores in that period. Clients who have already made an internal decision about not renewing often have less conflict about giving a positive score on a survey, because the score no longer matters to them. They have removed themselves from the relationship emotionally while maintaining the operational surface.
This is a core limitation of NPS and satisfaction-score-based monitoring for professional services. The score is measuring the client's willingness to give you a positive public endorsement, which is partially decoupled from their willingness to continue the engagement. A client who likes your team but has decided to bring the work in-house may well score you positively on your way out.
We are not saying satisfaction scores are useless. They catch genuine dissatisfaction clearly. What they miss is the client who is satisfied but leaving for reasons that have nothing to do with satisfaction: internal ownership changes, budget restructuring, a strategic shift that changes the type of support they need. The notes capture the signals of these transitions before the scores do, because the conversational and behavioral patterns change before the client's retrospective satisfaction assessment does.
The 8-12 week signal window
Across the accounts we analyzed, the most consistent finding was temporal: the note and verbatim patterns described above typically appear eight to twelve weeks before the renewal conversation, not in the immediate lead-up to it. In the final four to six weeks, the patterns are often more visible but the intervention window has narrowed significantly. Inside four weeks, the internal decision is usually made.
The eight-to-twelve week window is when the signals are present but the decision has not fully crystallized. A proactive conversation initiated in that window, one that acknowledges the signals and gives the client an opening to surface concerns they have not raised directly, has a genuinely different outcome than a renewal conversation initiated at 30 days out. The relationship can still be redirected if the problem is addressable.
This is what makes reading the notes matter operationally, not just analytically. The notes contain the information that opens that eight-week window. Not reading them means discovering the problem when the window is already closed.
What this does not tell you
This analysis describes patterns, not causation. The fact that verbatim specificity declines before non-renewal does not tell you why the client is leaving. It tells you which accounts to look at closely. The diagnosis and the intervention still require human judgment and a direct conversation.
We are also working from a limited data set. The patterns are consistent enough that we are confident they represent real phenomena, but anyone doing rigorous work in this space should treat the directional picture as a hypothesis to test on their own account data, not as established fact. Your firm's client relationships have their own dynamics, and the signals worth tracking may not perfectly match the patterns described here.
What is not open to much debate is the basic structural finding: the pre-churn signal is in the qualitative layer of the data, not in the CRM activity layer. If you are only reading the activity layer, you are reading the shadow of the relationship, and by the time the shadow tells you something is wrong, the conversation has already moved past the intervention window.