How to choose a CRM by what it costs to run

How to choose a CRM: the short answer#

Most guides on how to choose a CRM answer a different question. They rank products. Meanwhile the person signing the invoice is asking something else entirely. What will this cost to run, and will the team still be using it in year two? In short, this page answers the running question rather than the ranking one.

The three tests, each with its pass mark#

1. Price it over three years, at the seat count you expect to reach.

   Pass mark: you can state a thirty six month total for each shortlisted product, built from your own seats and the tier that carries the feature you need.

2. Record a real deal in what the product already ships.

   Pass mark: your last five closed deals record without more than a couple of custom fields.

3. Export everything yourself, during the trial, before you sign.

   Pass mark: contacts, deal history, notes and the activity log come out as files you can open.

Why the price you were quoted is not what the CRM costs#

The per-seat figure is what a CRM costs to buy. However, it is not what a CRM costs to run. Every quote you have collected covers one line of a bill that has at least six.

One vendor's own ladder spans 4.9 times#

Those lines are easy to name once somebody names them. First, the licence, charged per seat per month. Second, getting your existing records in, which is either your weekend or somebody's invoice. Next comes connecting the CRM to the tools that already hold customer data. Then training, which costs working hours rather than money. Finally there is administration, and administration never stops.

The six lines, and who publishes each number#

Two of those lines are printed on vendor pricing pages, so you can check them today. For example, as of September 2026 HubSpot charges a mandatory one-time onboarding fee on its Professional and Enterprise tiers, published on its own Sales Hub pricing page (opens in new tab). That fee is not optional. It is not inside the monthly figure you were quoted either.

The remaining lines are yours to estimate. Nobody will estimate them for you, and that is the part no pricing page covers. Because the vendor's page exists to sell the licence, the rest of the bill is simply not its subject.

In practice this is where most of the difference between two shortlisted products hides. Two products can quote within a few dollars of each other. Yet one loads your data in an afternoon and the other needs a consultant for a fortnight.

How to choose a CRM by the three-year number, not the monthly one#

The only cost figure that separates two products for your business is your own three-year total. A per-seat price separates nothing. The headline per-seat figure will not separate your shortlist, because it covers one line of a bill that has at least six.

Build the number from five inputs, and all five are yours. First, seats today. Second, seats you expect in year three. Then the tier you actually need, which is often not the tier you were quoted. The published price of that tier. Finally the one-time lines: migration, integration work and training hours.

The arithmetic, stated once#

Two mechanics move that total far more than the headline price does. First, seat growth compounds, so a team of four planning to be twelve is buying three times the licence it priced. Second, tier boundaries force upgrades. A feature listed on a pricing page as available often sits above the plan in your quote. As a result you price the cheap tier and buy the expensive one.

The tier that carries the reports is 12.9 times the tier in your quote#

Here is an illustrative example, not a client result. A four-person sales team prices a mid-tier seat. Then it expects nine seats by year three. Also it needs one reporting feature that lives a tier up. Consequently the three-year licence roughly triples against the number in the first quote.

Your own thirty six month total#

Run this arithmetic before the demo rather than after. Because you arrive holding a number, the conversation changes. Instead of asking what the product does, you ask which tier carries the three things you need.

That is the whole of the cost question. One number, built from your own seats, checked against the vendor's published tier.

Who administers this, and what does that time cost?#

A CRM needs a named person with standing time, or it decays. This is the running cost that appears on no invoice. Also, it is the one most likely to decide whether the system survives.

The work itself is small and constant. Somebody approves a new required field. Somebody decides what a stage means when two people disagree. Later somebody notices that half the deals have no close date, and chases them. Nobody does that work in the gaps between their real job.

Therefore answer three questions today, before the shortlist exists. Who approves a change to the record layout? How many hours a week do they have for it? What happens to the system during the month that person is on leave?

If the honest answer to the second question is zero, the purchase is already in trouble. Still, that is not a reason to pick a different product. Instead it is a reason to fix the staffing question first, or to wait.

In particular, watch for the pattern where the administrator is assumed rather than named. Somebody technical will pick it up. Somebody in operations already knows the tool. Because nobody agreed to it, the work lands on whoever complains last.

Test one: how to choose a CRM using your own last five deals#

A product fits when a real deal from your business can be recorded in what the product already has. No custom fields, no consultant, no workflow build. In short, that is the first test that separates products where a feature list cannot.

Run it on your last five closed deals. For each one, write down what was actually sold. Note who touched it, what stages it moved through, and what somebody had to know at each stage. Finally, open the trial and try to record all five.

Watch for the moment you reach for a custom field. One custom field is normal. However, five means the product models a different kind of business than yours. Ten means you are about to pay a vendor to become a platform you then have to maintain.

This failure is quiet, which is exactly why it needs a test. Because the demo uses the vendor's example deal rather than yours, a product that needs heavy shaping still demos perfectly. Meanwhile the gap only appears when somebody tries to log a real sale.

Whereas most of this page is about cost, this test is about shape. Also, it is the one that decides adoption, since a salesperson will not fight a form twice.

If your five deals fit nothing on the shortlist, the question changes shape. At that point you are choosing between bending your process and building something. We build custom software, so treat this as an interested view rather than a neutral one: our comparison of custom-built software against off-the-shelf products walks through where that line actually falls.

Test two: what has to be true about your data before day one?#

A CRM inherits whatever you put into it. Therefore the second test is about your records rather than the product. Before you migrate anything, find out whether your current data can be loaded at all.

Take 200 real records from wherever they live now. Load them into the trial. Next, count what broke. Duplicate companies, missing owners, a phone column holding three formats, notes trapped in an email thread nobody exported.

Most of those problems are yours. They would exist in any product you chose. Still, you want to meet them in a trial rather than in week one of a paid contract.

In particular, count the records that need a human decision before they can load. That count is your migration cost, expressed in hours rather than dollars. For instance, 40 records needing a judgement call is an afternoon. Whereas 400 is a fortnight somebody has to find.

Anyone moving off a spreadsheet has an advantage here and rarely knows it. Because your data is already in one place and already flat, most of it maps cleanly. In contrast, a team leaving an older CRM is usually carrying ten years of fields that no living person can explain.

Test three: can you get everything back out?#

Export everything yourself, during the trial, before you sign. That means contacts, companies, deal history, notes and the activity log. If any of those will not come out as a file you can open, you have found the most expensive thing on this page.

• Export contacts and companies as files you can open.
• Export deal history, rather than only the current values.
• Export notes, and check they carry their dates and authors.
• Export the activity log the same way.
• Open every file before you decide anything.
• Find out whether the export is a button or a support ticket.
• Read what the contract says happens to your data at the end of the term.

Almost nobody runs this test, because it feels like planning to fail. In practice it is the opposite. It is the difference between choosing a supplier and acquiring a landlord.

Therefore do it concretely. In the trial account, run the export. Open the files. Check that notes carry their dates and authors. Check that deal history is more than a list of current values. Finally, find out whether the export is a self-service button or a support ticket with a fee attached.

Read the contract for the same thing. Look for what happens to your data at the end of the term. Also look for how long it is retained, and in what format it is returned. Ask for that in writing before signature rather than after.

This is general information, not legal advice. Consult a licensed attorney before you rely on any contract reading, including this one.

Do this on day five of the trial, not on day one. By then the account holds enough of your own records for the export to prove something.

An export that works is also how you keep the rest of your position. Because a vendor whose product you can leave is a vendor you can negotiate with, this test pays for itself at every renewal.

Read the primary source, not the comparison post#

Verify every number that matters against the vendor's own documentation. Comparison content rounds figures. Then the rounded figure travels, and it is not the figure in the contract.

Take the daily API allowance on HubSpot's Professional tier. As of September 2026, HubSpot's own developer platform usage guidelines (opens in new tab) state 625,000 calls a day. A round-up may carry a nearby figure instead, and only the vendor's number is the one in the contract. Any gap matters only near the ceiling, and you learn you are near the ceiling by reading the real number.

The daily ceiling moves with the tier too#

The same applies to limits nobody quotes at all. For instance, Salesforce publishes a flat API allocation for Developer Edition in its limits quick reference (opens in new tab). Also, Zoho publishes a credit system with a per-tier ceiling in its API limits documentation (opens in new tab). Those numbers decide whether your future integration works at your volume.

Treat the category's failure statistics the same way. You will see claims that a large share of CRM projects miss their objectives. A figure that travels with no stated population and no publisher does not belong in your business case.

The number in the contract is the vendor's, not the round-up's#

The trial week: what to test on which day#

Every vendor says try it. Almost none says what to try. Therefore here is a five-day protocol you can run inside a standard free trial. It is the part a demo cannot do for you, because a demo is driven by the person selling.

Each day carries a pass mark, and the pass mark is the point.

Day three is the day that predicts adoption. Because a CRM the team will not update from a phone fills with stale deals, this is worth timing rather than assuming.

Run the same week against two products, never one. In contrast to a demo, this produces a comparison you can defend to whoever signs.

When you should not buy a CRM yet#

Sometimes the honest answer is that you should not buy one this quarter. No vendor's blog will say so, because the trial button is why the page exists. We build and customise CRMs, and we do not sell one, so here is the honest version.

Stay on the spreadsheet while two things are true. First, fewer than about a hundred active contacts and a handful of open deals at any time. Second, one person holds the whole customer relationship, and nobody else needs to see it to do their job.

Below that volume a spreadsheet is not a failure of discipline. Instead it is a correct tool, and leaving it early buys administration you cannot yet staff. Therefore wait until the pain is real. A second person needs the history, or a deal is dropped because nobody could see it.

Two other disqualifications are worth naming plainly. First, do not buy a CRM to fix a process nobody has agreed on, because the software will simply encode the disagreement. Second, do not buy while the person who would administer it has no hours to give.

One more case deserves naming. If your deals are won and lost inside a single channel that already keeps its own history, a CRM may only duplicate that record. Check where the history really lives before you buy a second place to keep it.

None of that is an argument against CRMs. Instead it is an argument for buying one at the point where it will actually be used.

How to choose a CRM once a product has passed all three tests#

If a product cleared the run cost, the fit test and the export, buy it and stop shopping. Because the remaining differences between shortlisted products are small, another month of evaluation costs more than it saves.

Two exits follow from here, and both are part of choosing well. First, running the product well, which is a different discipline from choosing it. Most of the value sits there rather than in the choice. Our notes on keeping a CRM accurate after go-live cover the habits that decide whether the records stay usable.

Second, the exit for the reader whose five deals fit nothing. If your business sells something the category does not model, an off-the-shelf CRM keeps asking you to be a different company. Then the question becomes what building costs instead. Our guide to what custom software development actually costs gives the ranges before anybody talks to you.

The one-page checklist for how to choose a CRM#

Here is the whole method on one page, in the order it should be run. Print it, or paste it into the document your shortlist already lives in.

Before you look at any product, answer two questions. Who administers this, and how many hours a week do they have? What does one closed deal in your business actually look like, written down?

Then, for each shortlisted product, do five things. First, build the three-year total from your own seat forecast and the tier that carries the feature you need. Second, record your last five deals in the trial without customising anything. Third, load 200 real records and count what needed a human decision. Also, time somebody untrained logging a call from a phone. Finally, export everything and open the files.

Then check every number you were quoted against the vendor's own documentation rather than a comparison post.

A product that clears all of that is your answer. Any product that fails one line has told you something specific, which is more than a demo ever does.

A note on where this comes from#

Atyantik has run 50+ enterprise engagements across 7 countries. All of them run from India. We also build and customise CRMs for companies whose process does not fit a product off the shelf. We do not sell a CRM. Therefore this page can say when not to buy one.

If you have run the three tests and want a second opinion on what you found, our CRM development and customisation work explains how we approach it. Read it when you are ready. There is nothing to fill in here.