Almost nobody comes to us asking for a CRM. They come to us saying “we are losing enquiries”, “nobody knows what is in the pipeline”, or “when a salesperson leaves, their customers leave with them”. Those are all the same problem, and a CRM is only the shape of the answer.
Which means the real question is not which CRM is best but what is the cheapest way to make follow-up impossible to forget. Sometimes that is a $30-per-seat subscription. Sometimes it is a build. Here is how to tell them apart.
What a CRM is actually for
Strip away the feature lists and a CRM does four jobs:
- Capture. Every enquiry lands in one place, whatever channel it came from: form, phone, WhatsApp, walk-in, marketplace.
- Ownership. Every deal has one name against it and one next action with a date. Nothing sits in a state called “someone should probably call them”.
- Memory. The history belongs to the company, not the salesperson’s phone.
- Visibility. You can see the pipeline, the conversion rate at each stage, and where deals stall, today rather than at month-end.
Anything beyond those four is a nice-to-have. Plenty of businesses pay for AI scoring and forecasting dashboards while failing at job one.
Seven signs your CRM is losing you deals
- Enquiries arrive in a shared inbox or a WhatsApp group and get picked up by whoever is free.
- Your pipeline lives in a spreadsheet that one person maintains, and that person is the only one who trusts it.
- Two people call the same customer in the same week; nobody calls half the others at all.
- Quotes are built by hand in Word and their status is unknown after they are sent.
- When someone resigns, you cannot reconstruct their open deals.
- You pay for a CRM, and your team keeps a private spreadsheet next to it. That spreadsheet is the specification for the system you should have.
- You cannot answer “how many enquiries did we get last month, and what happened to them?” in under a minute.
Point six is the important one. A parallel spreadsheet is not laziness; it is your team telling you the tool does not match the work.
When off-the-shelf is the right call
Buy, and stop reading, if most of these are true:
- Your sales process is recognisable: enquiry, qualify, quote, negotiate, win or lose.
- Fewer than about 20 people will use it.
- You need it running next week, not next month.
- You are willing to adapt your process slightly to the tool rather than the reverse.
- Your integrations are mainstream: e-mail, calendar, a common accounting package.
In that situation a subscription CRM configured properly (and it must be configured, not just purchased) is the cheapest path to a working pipeline. Spending $8,000 to build what $40 a month solves is bad engineering and worse business.
When a custom CRM pays for itself
Build when one or more of these is true, and the arithmetic will usually agree with you:
Your process is not a sales funnel
Trading houses track shipments, LCs and per-container costing against a deal. Clinics track patients and recurring treatment cycles. Contractors track site visits, variations and retentions. Construction, freight, education and healthcare pipelines rarely survive contact with a generic five-stage funnel, and every workaround you add is a place data goes to die.
Licences punish growth
Per-seat pricing is fine at eight users and painful at eighty. Once the annual subscription line approaches the cost of a build, you are renting something you could own.
The CRM has to live inside operations
If winning a deal must create a job, allocate stock, trigger a purchase order and raise an invoice, then sales and operations are one system. Bolting a generic CRM onto an ERP through a middleware tool is often more expensive and more fragile than building the sales layer where the operational data already lives.
You need automation the vendor charges a premium for
Automatic quote generation from a price matrix, WhatsApp follow-up sequences, an AI calling agent that qualifies inbound leads and books them straight into a rep’s calendar: these sit in the top-tier plan of most products, or are impossible in them.
Data residency or access rules are strict
Some clients cannot put customer data in another company’s cloud, or need audit control that a shared platform will not give.
The five-year cost comparison
Sticker prices mislead. Here is the shape of the real comparison for a 25-person sales team. These are illustrative ranges, not a quote, and deliberately unfavourable to nobody.
| Off-the-shelf (25 seats) | Custom build | |
|---|---|---|
| Year 1 | Licences plus setup and configuration; often $12,000–$25,000 once a mid-tier plan and implementation are counted | Build of $10,000–$30,000 depending on scope, plus modest hosting |
| Years 2–5 | The same licence cost every year, usually rising, and again if you add people | Hosting plus maintenance and changes: typically 15–25% of build cost a year |
| Customisation | Within the platform’s limits; deep changes need a specialist consultant | Anything, at the price of engineering time |
| Integration | Good for popular tools, expensive or impossible for your bespoke systems | Whatever has an API, on your terms |
| Speed to live | Days | Two to four weeks for the first usable version |
| Risk | Vendor pricing changes, feature removal, platform lock-in | You own the code, and you own the maintenance |
Two honest points. Custom is not automatically cheaper. It is cheaper at scale or at complexity, and it moves cost from a subscription line to a maintenance line. And a badly built custom CRM is far worse than a well-configured subscription, because you inherit every bug.
Want the arithmetic run on your numbers?
Send us your team size, your current tools and the part of the process that keeps breaking. You get a straight recommendation, including “stay where you are” when that is the right answer.
What any CRM must do (and what it need not)
Non-negotiable:
- Capture from every channel you actually use, including WhatsApp if that is where your customers are.
- One owner and one dated next action on every open deal, with a view of what is overdue.
- Full activity history against the customer, not the user.
- Stage-by-stage conversion reporting and a source report, so you know which marketing actually pays.
- Mobile use that works for someone standing in a car park.
- Data export you control.
Usually optional, often oversold: AI lead scoring before you have clean data, forecasting dashboards for a five-person team, gamification leaderboards, and any module nobody has asked for twice.
Making the team actually use it
Most CRM failures are adoption failures, and adoption is a design problem:
- Keep the logging cost under 20 seconds. Four fields, not fourteen. Every extra required field costs you compliance.
- Capture automatically wherever possible. Calls, e-mails and messages should log themselves; a voice agent or call summary tool removes the after-call typing entirely.
- Give the salesperson something back on day one: their own follow-up list, their own numbers. A CRM that only feeds management reports gets treated as surveillance.
- Migrate the real history. Starting empty means people keep using the old spreadsheet “just for a while”, which becomes forever.
- Pick one process owner internally who decides what a stage means. Ambiguity in stage definitions produces useless pipelines.
The verdict
Standard process, small team, need it now: buy, and spend the saved money on configuring it properly. Unusual process, growing headcount, or a CRM that has to sit inside operations: build, start with the pipeline slice only, and get it live in weeks rather than quarters.
We build custom CRMs as part of the business software practice, usually connected to the invoicing and operations side so a won deal actually turns into work. If you want a straight answer on which route fits, describe your process in two lines.