How to choose a CRM for your agency: the buyer's guide

TL;DR

Choosing an agency CRM goes wrong in the same three ways every time: buying for the agency you wish you were (enterprise feature lists you'll never configure), buying on per-seat pricing math that punishes the growth you're planning, and evaluating tools by demo instead of by your own messiest client. The process that works: write down your five ugliest real workflows first, score candidates only against those, run a two-week trial with one real client's data including the migration, and price the decision at the team size you'll be in 18 months — not today. This guide walks the whole decision: requirements, pricing models, the generalist-vs-agency-specific question, evaluation, and migration. Disclosure: we build Phloz, an agency CRM — the process below is the one we'd want used on us, and it doesn't always pick us.

---

Agencies don't outgrow spreadsheets because spreadsheets stop working. They outgrow them the day two account managers give a client two different answers, and nobody can tell whose answer was current. That's the actual job a CRM is hired for: one version of the truth about every client.

Hold onto that framing, because every step below tests against it.

Step 1 — Write your requirements from incidents, not imagination

Skip the feature wishlist. Instead, list the last five times client information failed you. Real examples from agencies we've talked to:

• A client emailed a scope change to one AM's inbox; the team kept executing the old scope for three weeks.
• Nobody could say which GTM container a departed employee had set up for which client.
• A new hire spent their first month asking "where do I find anything about this client?"
• A client called asking about an invoice and the AM couldn't see what had been promised.
• The owner couldn't answer "which clients are actually profitable" without a weekend of spreadsheet archaeology.

Each incident names a requirement: shared client communication, infrastructure memory, a single client profile, visible commitments, per-client economics. Your five incidents define your requirements — and a tool that nails your five beats a tool with 200 features that misses them.

Step 2 — Decide the generalist-vs-agency-specific question honestly

The market splits in two:

Generalist CRMs (HubSpot, Pipedrive, Zoho) are sales-pipeline tools at heart. They're excellent at deals: stages, close dates, win rates. If your bottleneck is winning clients, a generalist sales CRM is genuinely the right hire — we said as much in our honest HubSpot review.

Agency-specific platforms (Phloz and a handful of others) assume the harder problem is keeping clients: the work, the communication, and — in our case specifically — the tracking infrastructure each client runs on. The deal pipeline is secondary; the client lifecycle is primary — the shape a purpose-built marketing agency CRM is organized around. If you're not sure you even need a CRM versus a generic tool, the what-is-an-SEO-CRM explainer walks the "do you need one yet" question for one discipline.

The honest test: count your records. If you have 200 prospects and 8 clients, you have a sales problem — buy a sales CRM. If you have 25 clients and a dozen prospects, you have a delivery-and-retention problem — buy for the client lifecycle. Most agencies past year two are the second kind and are using a tool built for the first.

Step 3 — Do the pricing-model math at month 18, not today

Three pricing models dominate, and they diverge wildly as you grow:

Per-seat (most generalists): looks cheap at 4 people. At 18 people on a $45–100/seat tier you're at $810–1,800/month, and every hire carries a software tax. Worse, per-seat pricing pushes agencies to under-license — viewers sharing logins, juniors locked out — which silently breaks the "one version of the truth" job you hired the CRM for. We ran the full curve in the HubSpot TCO math.

Per-client (Phloz's model — bias noted): cost tracks your revenue driver instead of your headcount. Hiring is free; growing the client roster is what costs, in proportion to the money it brings in. The fit test: agencies with many small clients should check the tier boundaries; the model favors meaningful retainers over thousand-client volume shops.

Flat-tier (some PM-tools-with-CRM-features): predictable, but the tiers usually gate the exact features agencies need (client portals, permissions), so read the gates, not the headline price.

Whatever the model: build a two-line spreadsheet — cost at today's size, cost at your honest 18-month plan — before any demo. Vendors price for the buyer who doesn't.

Step 4 — Ask the questions vendors hope you won't

From watching agencies get burned, the seven questions that separate marketing from product:

1. "Show me a client's complete picture on one screen." Profile, open work, last communication, who's assigned. If the demo needs four tabs, your AMs will need four tabs forever.
2. "How does client email get into the system?" Forwarding rules that humans must remember will be forgotten. Look for real inbound email handling — unique per-client addresses beat 'log it manually' every time.
3. "What do my clients see?" A portal that shows status without a login wall of its own reduces "quick update?" emails more than any internal feature. Check our take on what a client portal should actually include.
4. "What happens when someone leaves?" Reassignment, access revocation, and whether their knowledge (notes, setups, history) survives them.
5. "How do roles and permissions map to MY org?" Owner/admin/member/viewer with client-level assignment is the minimum for agencies; all-or-nothing access is how interns end up in billing.
6. "How do I get my data OUT?" Full export, documented format, no exit fees. A vendor confident in retention makes leaving easy.
7. "Where does tracking infrastructure live?" If your agency does performance work, ask where the answer to "which pixel, which container, which CAPI endpoint, who has access" lives. For most CRMs the honest answer is "a spreadsheet next to the CRM" — which is the gap we built the tracking map for.

Step 5 — Trial with your messiest client, including the migration

A demo with vendor sample data tests nothing. The evaluation that predicts reality:

1. Pick your messiest real client — most contacts, most channels, most history.
2. Migrate that one client fully during the trial: contacts, open work, recent email history, links to their assets. Time it. Multiply by your client count. That number is your real switching cost, and you want it before you've committed, not after.
3. Run the client through one real week: log actual communication, run actual tasks, answer one actual "what's the status?" from memory of the tool alone.
4. Have your most skeptical AM do the week-two pass. The enthusiast's opinion is already priced in; adoption lives or dies with the skeptic.
5. Only then look at the feature list — to break ties, not to lead.

Two weeks, one real client, the skeptic's verdict. That process has never steered an agency we know wrong.

Step 6 — Plan the migration like a client project

The CRM you choose matters less than the migration you run. The pattern that works is the same one we recommend for client onboarding: a checklist, an owner, and a deadline.

• Freeze a cutover date. Parallel-running two systems for "a transition period" means running two systems forever.
• Migrate clients in tiers: active retainers first (full history), dormant clients second (profile + summary only), dead clients as an archive export — not into the new system at all.
• Kill the old system's write access on cutover day. Read-only for ninety days, then export and close. Every agency that kept the old tool writable still has people in it a year later.
• Measure adoption weekly for the first month: percentage of client communication landing in the system, percentage of work items with a current status. Under 80% by week four is a leadership problem, not a tooling problem — address it as one.

The decision in one paragraph

Write the five incidents. Decide whether you're buying for sales or for the client lifecycle. Price it at month 18. Ask the seven questions. Trial with your messiest client and let the skeptic judge. If the lifecycle is your problem and tracking infrastructure is part of your delivery, that's the shape we built — and if a sales pipeline is your problem, buy the sales CRM with our blessing. The only wrong choice is evaluating by demo gloss and pricing by today's headcount.