Agency CRM: Managing Clients, Retainers and New Business
Monday pipeline review at a thirty person agency. The board looks good: eleven live opportunities, two in contract, a healthy weighted number at the bottom of the column. Nobody mentions the retainer that has quietly run at forty hours a month against a twenty five hour scope since April, or the client whose twelve month agreement auto renews in nineteen days with no conversation booked, or the account where the only two people who ever replied to email have both left. An agency CRM that models new business and nothing else will show you a green board on the morning you lose a third of your recurring revenue.
This is not a discipline problem. It is a data model problem. Agencies run two revenue motions simultaneously, and almost every CRM on the market was built to describe one of them well.
Why an agency CRM has to run two pipelines at once
A software company has one motion: acquire, close, expand. The CRM object model matches it. An opportunity is created, moves through stages, and closes. After that the record is historical.
An agency has a second motion that never closes. Retained revenue is continuous, it renews on dates nobody put in the CRM, it erodes through scope creep long before it churns, and it lives in the delivery system rather than the sales system. When you force it into an opportunity record, you get a "closed won" row from fourteen months ago that has not changed since, attached to a client relationship that is currently on fire.
| Dimension | New business pipeline | Retained client pipeline |
|---|---|---|
| Unit | Opportunity, finite, ends at signature | Agreement, continuous, ends at notice |
| Time shape | Weeks to months, forward moving | Rolling term with a renewal and a notice deadline |
| Owner | New business lead or principal | Account lead or delivery lead |
| Primary failure | Deal stalls or goes to a competitor | Margin erodes, sponsor leaves, renewal lapses |
| Leading indicators | Meeting held, scope agreed, procurement engaged | Hours delivered vs contracted, response latency, sponsor changes, invoice behaviour |
| Where the evidence lives | CRM, inbox, calendar | Project tracker, time tracker, inbox, Slack, invoicing |
| What most CRMs model | Well | Barely |
Read that last row carefully, because it explains most of the frustration agency owners have with CRM for agencies as a category. The tool is not broken. It is answering the question it was designed to answer while the money is leaking somewhere it cannot see.
The practical consequence: you need stage design for one pipeline, an agreement object with dates for the other, and a way to join them to the systems that hold the delivery truth. Take those in order.
Stage design for the new business pipeline
Agency pipelines rot for a specific reason: stages get defined by what the agency did rather than what the client did. "Proposal sent" is a seller activity. It tells you nothing, because proposals go into voids constantly. Every stage should have an exit criterion that a client had to satisfy.
A workable set for most marketing agency CRM and creative agency CRM setups:
- Qualified conversation. Exit criterion: you know the problem, the approximate budget range, and who signs. Not who briefed you, who signs.
- Chemistry or credentials meeting. Exit: a second meeting is booked, or the client has named the other agencies in the process.
- Scoping. Exit: the client has agreed to a written scope outline, including what is excluded. If nobody on their side will confirm exclusions, you are being used for free thinking.
- Proposal or pitch. Exit: proposal delivered and a decision date confirmed by the client.
- Contract and procurement. Exit: legal or procurement has the paper. This stage is separate on purpose, because it has its own failure mode and its own timeline that has nothing to do with whether they liked the work.
- Won, lost, or no decision.
Three fields earn their place beyond the defaults.
Pitch investment. Hours spent on the opportunity, logged even roughly. Without it you cannot calculate cost per win, and cost per win is the number that tells you whether to keep entering four way pitches. Agencies that track this often discover a whole category of RFP they should decline on sight.
Origin. Referral, inbound, outbound, existing client expansion, network. Keep the list short and mutually exclusive. Referral and network get blurred constantly and they behave completely differently.
Loss reason, from a fixed list. Price, timing, incumbent retained, lost to in house team, no decision, disqualified by us. Free text loss reasons are unanalysable within a quarter. "No decision" deserves its own value because it is usually the largest bucket and it points at qualification rather than pricing.
One structural note: pitches and RFPs need either a separate stage set or a flag, because the shape is different. There is no discovery, the decision date is fixed, and half the qualification questions are unanswerable. Mixing them into the same funnel makes every conversion rate meaningless.
None of this works if the underlying contact records are a mess, and agency contact data is usually worse than average because clients move between brands constantly. The mechanics of getting that foundation right are covered in CRM Contact Management: From Spreadsheet to Real Database, and it is worth doing before you build anything on top.
Modeling the retainer: the object your CRM probably does not have
Here is the piece that separates a real client management CRM setup from a sales pipeline with clients in it. You need an agreement object, whether that is a native subscription or contract object, a custom object, or in the simplest instances a well disciplined record type.
Fields that matter:
- Monthly or quarterly fee, and the currency
- Contracted volume: hours, deliverables, or retained days
- Term start and term end
- Notice period in days
- Auto renew flag
- Renewal owner, named, a person not a team
- Rate card version and effective blended rate
- Scope document link, the actual signed one
- Budget holder contact and procurement contact, as separate roles
- Last review date and next review date
- Delivery system project ID
The derived field that changes behaviour more than any other: notice deadline, calculated as term end minus notice period. Agencies diary the renewal date and miss the notice window by two weeks, which means either an unwanted twelve month extension or a repricing conversation held from a position of no leverage. Put the notice deadline on the board, not the renewal date, and set the reminder ninety days before it.
Then give the retainer pipeline its own stages, because "closed won" is not a state a live client relationship can sit in for two years:
Healthy, Watch, At risk, In renewal conversation, Repricing, Notice served, Renewed, Expanded, Churned.
Movement between these should be triggered by evidence rather than by a mood in a Monday meeting. Which brings us to the signals.
Scope creep signals worth wiring in
Scope creep is usually discussed as a delivery problem. It is a commercial signal, and it is the single most reliable leading indicator of an unhappy renewal, because a client consuming forty hours against a twenty five hour scope is either about to receive an awkward invoice or about to be quietly subsidised until someone notices the account is losing money.
| Signal | Where it actually lives | What it usually means |
|---|---|---|
| Delivered hours over contracted hours, trending | Time tracker, project tool | Margin erosion, and a repricing conversation you are already late for |
| Requests arriving outside the agreed channel | Slack, email, WhatsApp | Process has collapsed, work is unlogged, and none of it will appear in the scope review |
| Number of distinct requesters climbing | Project tool, inbox | You have been adopted as an internal team without a change of contract |
| Senior staff on junior rate card work | Time tracker | Effective rate falling even when hours look fine |
| Revisions per deliverable rising | Project tool | Brief quality or approval chain problem, often a sponsor change upstream |
| Sponsor or budget holder replaced | Inbox, CRM contact history, LinkedIn | The single strongest churn predictor in agency work |
| Invoice paid later each month, or disputed | Stripe, QuickBooks, Xero | Budget pressure, or dissatisfaction being expressed through finance |
| Review or QBR overdue | Calendar, CRM | Nobody is having the commercial conversation |
| Response latency from the client increasing | Inbox | Attention has moved elsewhere |
Notice how few of those live in the CRM. That is the whole difficulty. A CRM agency setup can hold the agreement, the dates and the owner, but the signals that tell you whether the agreement is healthy are scattered across the time tracker, the project board, the inbox and the accounting system.
You have three options: manual review, where an account lead assembles this once a quarter and it slips whenever delivery gets busy; heavy integration work, syncing hours and invoice data into CRM fields on a schedule; or a layer that reads across the systems and answers questions without moving the data. Each has real costs, and the third one is where the honest positioning section below comes in.
The handoff from sales to delivery, where the margin actually goes
Most agency margin is lost in the two weeks after signature, not during the pitch.
The pattern is always the same. The proposal contained assumptions: two rounds of revisions, client supplies photography, approvals within three working days, one stakeholder in the approval chain. Those assumptions priced the job. They then stay in the proposal document, which the delivery team never reads, and the team delivers to the client's expectation rather than to the contract. By month three the account is running at a fraction of its intended margin and nobody can point to the moment it went wrong.
A handoff checklist that closes the gap:
- Signed scope with assumptions and exclusions extracted into the project brief, not linked, extracted
- Rate card version and blended rate the job was priced at
- Named delivery lead, and the date they were introduced to the client
- Approval chain: who approves, in what order, with what turnaround
- Escalation path on both sides
- Billing setup: PO number, invoicing schedule, payment terms, who receives invoices
- Kickoff booked before the CRM record moves to won
- Access and credentials requested
And one thing that is not a checklist item but a design decision: a shared client identifier that appears in the CRM, the project tool, the time tracker and the invoicing system. Without it, no analysis is possible in any tool, ever, because there is no join key. Agencies commonly have "Acme Corp", "Acme Corporation" and "ACME" as three separate entities across three systems, and every attempt at profitability reporting dies on that fact. Fix it early. It costs an afternoon at twenty clients and a month at two hundred.
The billing side of this handoff has its own machinery, and the reporting cadence that comes out of it is covered in Financial Reporting Automation: From Close to Board Deck. The delivery side, where utilisation and project status reporting live, is covered in AI Agents in Work Management: Analytics and Reporting.
Choosing a CRM for agencies: what to compare
Agencies shop for a CRM agency solution and find three genuinely different categories. Comparing them on feature lists is a waste of an afternoon. Compare them on which of the two pipelines they model natively.
| Category | Examples of shape | Models new business | Models retainers | Realistic fit |
|---|---|---|---|---|
| Lightweight pipeline tools | Simple deal boards, Kanban style | Adequately | No | Under ten people, referral led, one owner doing sales |
| General CRM platforms | HubSpot, Pipedrive, Zoho and similar | Very well | Only with custom objects and work | Ten to a hundred people with someone who will own admin |
| Agency management systems | Bundled CRM plus projects, time and billing | Adequately, sometimes weakly | Natively, including utilisation and profitability | Fifteen people and up, delivery heavy, retainer led |
| Open source and self hosted | Self run stacks | Varies | Only what you build | Technical teams with real reasons to control the data |
| Spreadsheets | The honest starting point | Poorly past twenty deals | No | Genuinely fine at five clients, ruinous at thirty |
The tradeoff nobody states plainly: bundled agency management systems model retainers, utilisation and profitability natively because that is their origin, and their sales pipeline is usually the weakest module in the suite. General CRMs have a far better new business pipeline and no concept of delivered hours. So you either accept a mediocre pipeline inside a strong delivery system, or a strong pipeline sitting beside a delivery system that knows nothing about it. There is no product that is genuinely excellent at both, and any vendor claiming otherwise is asking you to ignore one half of the demo.
For a broader view of the general CRM field sorted by team size and budget, see Best CRM Software: A Shortlist by Team Size and Budget. If controlling the data yourself is a live consideration, Open Source CRM: Self-Hosted Options and Real Tradeoffs covers what self hosting actually costs in practice. And if your agency also runs ongoing support or managed service work, the ticket history side is worth reading in Customer Service CRM: Support Tickets Meet Full History, because support volume per client is another early churn signal.
Questions worth asking any vendor:
- Can I create a first class object for a retainer with its own stages, dates and owner, or am I bending opportunities?
- Can I calculate and alert on a date that is derived from two other fields, such as notice deadline?
- Can I see delivered hours against contracted hours on the client record without exporting anything?
- What happens to reporting when a client has three concurrent agreements at different rates?
- Can I export everything, including custom objects and activity history, in a usable format?
That last one is not paranoia. It is the difference between a two week migration and a six month one.
Where Skopx fits, and where it does not
Skopx is not a CRM. It is not a project tool, not a data warehouse, and not a dashboard builder. It will not store your client records, run your pipeline stages, or replace whatever you pick from the table above. If you are looking for a system of record, keep looking, and buy one.
What it does is the join. Skopx is an AI workspace that connects nearly 1,000 tools a company already uses, including Gmail, Slack, Stripe, HubSpot, QuickBooks and Google Analytics, and answers questions with cited data from those tools. For an agency owner that means the retainer health question stops being a quarterly assembly job. You ask which retainers are at risk, and the answer comes back built from the CRM agreement record, the hours in the project tracker, the invoice status in the accounting system and the response pattern in the inbox, with the underlying records attached so you can check the reasoning rather than trust it.
The parts that matter for the two pipeline problem:
- Chat over connected systems. Cross system questions that no single tool can answer: which clients are over scope this month, which agreements hit their notice deadline in the next ninety days, which accounts have not had a review booked since the last renewal.
- A morning brief. The overnight version of the same thing, so a sponsor change or a failed payment reaches the owner on the day rather than at the next QBR.
- An insights engine that surfaces anomalies without being asked, which is the only way scope creep gets caught, since nobody schedules a check for a problem they do not know exists.
- Workflows built by describing them in chat, for the recurring sweeps: the ninety day notice reminder, the weekly over scope check, the handoff checklist that fires when a deal moves to won.
- BYOK. Bring your own key for any major model, zero markup. Pricing is $5 per month for Solo and $16 per seat per month for Team.
Weekly retainer risk sweep
Monday 07:00
Weekly schedule before the pipeline meeting
Pull live agreements
Fee, contracted hours, term end, notice period, renewal owner
Pull delivered hours
Last 30 days from the project and time tracker
Check invoice status
Late, disputed or failed payments per client
Check client responsiveness
Reply latency and sponsor address changes
Flag exceptions
Over scope, notice deadline inside 90 days, or two or more signals
Post to account leads
One line per client with the source records linked
Where it does not help, stated plainly:
If the data is not in a system, no connector invents it. Scope agreed verbally on a call, hours estimated from memory at the end of the week, a rate card that exists only in the founder's head: none of that becomes visible because you added a tool. The prerequisite is that hours get logged and agreements get recorded somewhere machine readable.
If you need enforcement, that is CRM work. Required fields, stage gates, quote generation, approval routing, permissions per record: buy that from your CRM vendor.
If you need capacity planning, that is a delivery tool. Resource scheduling, forecast utilisation and who is free in three weeks belong in the system that holds the plan.
If you need a formatted board pack, a reporting tool or a spreadsheet will produce a better artifact. The broader question of what belongs in an analytics platform versus a question answering layer is worked through in How to Get Actionable Insights From Analytics Platforms, and if you are weighing what the no cost end of this market can and cannot do, Free AI Data Analysis Tools: Where They Help and Stop is the honest version.
A thirty day sequence to get this running
Week one: identity. Agree one client name and one client ID across CRM, project tool, time tracker and invoicing. Merge duplicates. This is unglamorous and it is the whole foundation.
Week two: agreements. Create the retainer object or record type. Populate fee, contracted volume, term dates, notice period, renewal owner and scope link for every live client. Twenty clients is an afternoon of work. Derive the notice deadline and put it on a shared calendar.
Week three: stages and hygiene. Rewrite new business stages with client side exit criteria. Add pitch investment, origin and a fixed loss reason list. Archive every opportunity untouched for ninety days into no decision so the board reflects reality.
Week four: signals and the handoff. Wire the over scope check and the ninety day notice reminder. Write the handoff checklist and make kickoff booking a condition of marking a deal won. Run the first retainer review using the signal table above and see how many accounts you were wrong about.
By day thirty you will not have perfect data. You will have something better: a board that shows both pipelines, and a set of dates that arrive before the conversation instead of after it.
Frequently asked questions
Do we need a specialist agency CRM or can we use a general one?
Either can work. The question is which half you are willing to build yourself. A general CRM gives you a strong new business pipeline and you build the retainer object and its dates on top. An agency management system gives you retainers, utilisation and billing natively and you accept a weaker sales pipeline. Under about fifteen people, a general CRM plus a disciplined retainer object is usually cheaper and faster. Above that, the delivery side gets complex enough that a purpose built system starts paying for itself.
How should a small creative agency CRM handle retainers without custom objects?
Use a second pipeline of the same object type, with its own stages, and reuse fields with clear renaming. It is inelegant and it works up to roughly thirty active clients. The non negotiable parts are the term end date, the notice period and a named renewal owner. If you can only add three fields, add those three.
Who should own the renewal, the person who sold it or the account lead?
The account lead, with the seller involved in the conversation. The person doing the work knows the health of the relationship and the seller knows the commercial history. What fails reliably is leaving it unassigned, or assigning it to a team rather than a person, because a renewal with no name against it is a renewal nobody is preparing for.
What is the minimum data hygiene for any of this to work?
Three things. One client record per client across every system, with a shared identifier. Hours logged against a project, even approximately, because the over scope signal is the highest value one and it is impossible without them. And term dates on every agreement. Everything else is an improvement on top of those.
Should time tracking data live inside the CRM?
Only as a summary. The time tracker stays the source of truth, and the CRM holds a periodically refreshed rollup, such as delivered hours this month against contracted hours, on the client record. Copying detailed time entries into a CRM creates a synchronisation problem that will be wrong within a week and that nobody will notice is wrong.
How is a client management CRM different from a sales CRM?
A sales CRM optimises for converting an opportunity, so it measures activity, stage velocity and win rate. A client management CRM optimises for keeping and growing a relationship that already exists, so it measures delivery against contract, sponsor stability, review cadence and renewal readiness. Agencies need both, running side by side, which is exactly why the single pipeline default causes so much trouble.
Skopx Team
The Skopx engineering and product team