The three textbook types, briefly
The classification comes from academic work in the 1990s and it describes what a system is mostly for, not what it is called in a price list. Almost every product on the market does some of all three.
Operational CRM runs the work. Contacts, pipeline, quotes, jobs, tickets, tasks and the automations between them. This is what people picture when they say CRM, and it is the only one most businesses under a hundred people ever need to buy deliberately.
Analytical CRM reads the data the operational system produced and answers questions with it: which sources produce customers who stay, what a customer is worth over five years, which stage leaks. In small and mid sized businesses this is a set of reports and a warehouse query, not a separate purchase.
Collaborative CRM is about everybody who touches the customer seeing the same record: sales, service, delivery, finance, and sometimes partners or dealers through a portal. It is less a product than a decision about permissions and scope.
If a vendor tells you their product is an analytical CRM, ask what it does when a salesperson needs to send a quote. The three types describe emphasis. They do not describe a shopping list.
So treat those three as vocabulary, not as options. The four splits below are the ones that genuinely narrow a shortlist.
The vocabulary inside every type: records, activities and notes
Whatever the type, every CRM is built from the same three pieces, and knowing them makes vendor demos easier to follow.
Records are the things you track: contacts, companies, deals, jobs, matters. Each has fields and an owner. Activities are dated events attached to a record: a call, a meeting, an email, a task with a due date. They drive reminders and reports. Notes are free text attached to a record, the place for what the fields cannot hold.
A CRM note, by definition, is unstructured. That is its strength and its weakness. Notes capture context nobody thought to make a field for, such as a client's preference about how to be contacted or the history behind a complaint. But notes cannot be filtered or counted reliably. When the same kind of note appears on more than a handful of records, it should become a field. That rule alone keeps most systems usable.
Split one: what shape is the thing you track
This is the first question we ask in discovery, and it predicts more about the right system than industry or headcount do. Four shapes cover nearly everything.
| Shape | The central record | What makes it hard | Typical products that fit |
|---|---|---|---|
| Deal | One opportunity that closes once | Quoting, forecasting, probability | Nearly every mainstream CRM |
| Job | A site or asset visited repeatedly | Scheduling, routing, a usable phone view | Field service tools, rarely a pure CRM |
| Relationship | A person tracked for years | Segmentation, consent, many small events | Donor and member platforms |
| Matter | A case with dates and documents | Deadlines, permissions, audit trail | Practice management software |
The reason this matters commercially: the mainstream CRM market is overwhelmingly built for deal shape, because that is where the software industry's own sales teams live. If your business is job shape or matter shape, most of the products you will be shown are a poor structural fit, and no amount of custom fields will change that. You will be storing a job as a deal that reopens, which works until you need to schedule it.
Businesses that carry two shapes at once are the ones that end up with two systems and a spreadsheet joining them. A landscaping firm runs job shape for maintenance rounds and deal shape for design work. A law firm runs matter shape for cases and deal shape for business development. Naming both shapes out loud, early, is the single most useful hour in a CRM project. The CRM examples guide works through fourteen businesses shape by shape.
Split two: horizontal or vertical
A horizontal CRM serves every industry and calls everything a contact, a company and a deal. A vertical CRM is built for one industry and already knows what a matter, a load, a listing or a beneficiary is.
Vertical products win on day one and lose slowly. You get the right vocabulary, the right fields, the compliance features and often the industry integrations already wired. You are also buying a roadmap set by somebody else's average customer, a smaller supplier with fewer engineers, and a migration path that is harder because the data model is specialised. Prices tend to be higher per seat than horizontal equivalents, because the market is smaller.
Horizontal products win on ecosystem and lose on vocabulary. Everything integrates with them, hiring someone who knows them is easy, and the per seat price is competitive. In exchange, somebody at your company spends two months turning a generic tool into an approximation of your industry, and that configuration is a liability that leaves when they do.
Our honest read after a few dozen of these projects: for a common industry with a settled process, a good vertical product is hard to beat and we say so. For a business whose process is genuinely its own, the choice is between configuring a horizontal product forever and building the thing once. That is the case for a custom build, and it is also why we publish forty five industry pages rather than one generic pitch.
What CRM means in banking, hotels, retail and franchises
The same word covers very different systems by industry, which is why examples from one sector mislead buyers in another.
| Industry | What CRM usually means there | The central record |
|---|---|---|
| Banking | Relationship management for business and private clients, alongside the core banking system | The client relationship and its products |
| Hotels | Guest profiles, stay history and preferences, fed by the property management system | The guest and each stay |
| Retail | Loyalty, purchase history and segments for marketing, often closer to a CDP | The shopper and their orders |
| Franchises | Two layers: franchisee recruitment for head office and local customer pipelines for each site | The franchisee and the site |
In banking and hotels the CRM sits next to a core system that owns the transactions, so the CRM is mostly analytical and collaborative. In retail it is largely marketing. Franchises are the interesting case, because head office needs a view across every site while each franchisee needs a simple operational CRM of their own. The franchise CRM page and the hotel CRM page show both structures.
Split three: how it is deployed and who holds the keys
Deployment used to be a technical question. It is now mostly a commercial one, about who can raise your price and who can lose your data.
| Model | Where it runs | What you control | The real risk |
|---|---|---|---|
| Public SaaS | Vendor's cloud, shared | Configuration only | Price rises, feature removal, export on their terms |
| Private or single tenant SaaS | Vendor's cloud, isolated | Configuration, some data residency | Same commercial risk, higher bill |
| Self hosted open source | Your server | Everything, including the upgrades | You now run software, badly, unless someone owns it |
| Custom, hosted for you | Your hosting account | Code, data, database, backups | You need a maintainer, which is a contract, not a hire |
| On premise | Your building | Everything, including the hardware | Almost never worth it in 2026 |
One question separates these more clearly than any feature comparison: can you get every record, with relationships and attachments, out of the system this afternoon, without asking permission. If the answer is no, you are renting, and every renewal is a negotiation you enter from behind.
We build in the fourth row. The code and the database live in the client's own hosting account, we hold a deploy key rather than the keys, and the support agreement is cancellable. That is not a moral position, it is a commercial one: it means we have to keep being worth paying.
Split four: bought, configured or built
This is the split people mean when they ask about types of CRM, even when they use the textbook words. Three routes, and the right one is usually obvious once the cost is written down honestly.
| Bought as is | Bought and configured | Built | |
|---|---|---|---|
| Time to first use | Days | 4 to 10 weeks | 3 to 8 weeks |
| Up front cost | Nothing to low | 3k to 25k of consulting | 4.9k to 19.9k |
| Cost per year | Per seat, grows with the team | Per seat, plus maintenance of the configuration | Hosting and support, flat |
| Process fit | You adapt | Close, with compromises at the edges | Exact, and you own the edges |
| Who can change it | Anyone, within limits | Whoever knows the platform | Any competent developer, it is your code |
| Exit | Export and start again | Export, and the configuration is lost | You already have everything |
The configured middle row is the one that surprises people. Paying a partner twenty thousand to make a platform behave like your business costs as much as building, produces something you cannot take with you, and still carries the per seat bill afterwards. We win more work against expensive configuration projects than against the platforms themselves.
Buying as is remains right more often than an agency in our position should admit. If your process is ordinary, your team is small and you would be one of ten thousand customers doing the same thing, buy. Our pricing page starts by saying when not to hire us.
Where marketing automation, CDP and ERP sit
Three adjacent categories get called CRM types in articles written by people who have not implemented any of them. They are different systems with different jobs.
- Marketing automation sends the messages and reacts to opens, clicks and forms. It owns deliverability and unsubscribes. It is a poor system of record. The comparison is in CRM versus marketing automation.
- A CDP assembles behaviour from many sources into unified profiles and pushes audiences to other tools. It sends nothing and nobody logs into it daily. See CDP versus CRM.
- An ERP runs stock, purchasing, production and the ledger. It records what happened to money and goods. See ERP versus CRM.
- A CMS publishes your website. It is not a place to keep customers, however many plugins say otherwise. See CRM versus CMS.
The useful mental model is ownership of a question. The CRM answers who should we contact and what is the state of their work. The others answer what did we send, who is in this audience, what does the ledger say, and what does the public see. When one system is asked a question it does not own, the answer is a spreadsheet, and the spreadsheet becomes the real CRM.
A decision path that takes ten minutes
Answer these in order and stop at the first clear yes.
- Do you have fewer than about two hundred customers and one person following up? Use a free tier product or a shared sheet with a review habit, and revisit in a year. Spending money here buys nothing.
- Is your process ordinary deal shape, with a team under fifteen? Buy a mainstream product. The market is competitive and you will get a good tool for a predictable price.
- Is there a credible vertical product for your industry, with references you can call? Trial it properly with your own data. This is the best value in the market when it fits.
- Are you about to pay a partner more than ten thousand to configure a platform? Price a build against it before you sign, because the two numbers are usually closer than expected and only one of them leaves you owning something.
- Does your process carry two shapes, or an unusual compliance requirement, or a per seat bill that has passed what a build costs? Build, and keep the scope narrow on the first release.
Nothing above says buy the most features. Features are the cheapest thing in this market and the least correlated with whether a CRM is still being used in year three.
What actually decides success, once you have picked
We have watched good systems fail and ordinary systems succeed, and the pattern is consistent enough to state plainly.
Adoption beats capability. A system that is opened every morning by everyone beats a better one that half the team avoids. The most reliable predictor of adoption is whether the tool makes someone's day shorter in week one, which usually means the phone view and the quote document, not the reporting.
One owner per record. Every enquiry, job or matter has a named person and a next date. Systems without this rule decay into a list nobody is responsible for, regardless of type.
Fewer fields than you want. Every required field is a tax on every record, paid forever. Start with the fields you would actually filter or report on, and add the rest when someone asks twice.
Somebody owns the system. Not a committee, not the vendor. One person who decides what changes and who reviews the data monthly. This is the difference between a CRM and a graveyard, and no purchasing decision substitutes for it. The habits that hold a system together are in CRM best practices.
If you think you need a custom one
Write four lists before you talk to anybody, including us. The nouns your team talks about as units of work, the stages the main noun moves through in your own words, the sentences that start when X happens somebody has to remember to Y, and the five numbers you would read every Monday.
Those four lists are a specification. They make quotes comparable, they expose the two shape problem early, and they belong to you whoever you hire. Send them to us and a fixed price with a two page scope comes back inside two business days, in writing, with no call required. If the lists say buy something instead, we will say that, and it costs you nothing to find out.
The whole map in one picture
Four splits, in the order they narrow the field. Read it downward and stop when the answer is obvious.
The order matters. Most buying processes run it backwards: a budget appears, a shortlist of three vendors follows, demos are watched, and the shape question never gets asked at all. That is how a job shape business ends up a year later storing site visits as reopened deals, with a scheduling spreadsheet beside the CRM that everyone actually trusts more.
Running the questions in this order costs you an afternoon. It also produces a document that makes every quote you receive comparable, because vendors are answering the same four questions instead of each showing you the part of the problem they are best at.
