How to buy enterprise software without regretting it in a year

The software usually works. The regret is about everything around it: a department that already loved one vendor, a demo that never touched the actual workflow, a committee that never sat in the same room, a contract signed because the current licence ran out in three weeks.

Analysts have been saying some version of this for twenty years. Gartner and Forrester have put CRM “failure” somewhere between roughly 50% and 47% depending on the year and the yardstick. Independent 2025 research from Johnny Grow, measuring failure as did not achieve planned objectives, landed at 55%. Only 25% of those projects hit the objective, the timeline, and the budget. The median overrun, when cost slipped, was 30–49%.

That is not a product ranking. It is a process ranking.

• Missed the planned objective55%

  The system went live. The job it was bought to do did not.

• Hit objective, time, and budget25%

  The only definition of success a finance director will accept.

• Cancelled before go-live10%

  You paid for a project, not a system.

Johnny Grow, CRM Failure Report (2025): failure = did not achieve planned objectives. 25% hit objectives, time, and budget together. Median cost overrun among overruns: 30–49%.

The decision is the product

A vendor demo is a sales asset. It is designed to look like your company. It is not evidence that the system will survive your month-end, your Arabic-language invoices, your WhatsApp-to-CRM handoff, or the three people who actually have to live in it.

If you start with “who should we buy,” you have already skipped the only question that changes the outcome: what does this system have to do, for whom, every week, or we should not buy it.

How it usually goes

1. 01A director names a vendor they used at the last company
2. 02Three demo calls, all on the happy path
3. 03IT is asked to ‘make it work’ after the quote is approved
4. 04Contract signed in the last fortnight of the current licence

How it should go

1. 01Write the jobs the system must do — with the people who do them
2. 02Score vendors against those jobs, not against a feature matrix
3. 03Run one ugly demo: your data, your exception, your worst week
4. 04Negotiate on a calendar you control, with usage and a second bid on the table

Who has to be in the room

Buying software as a 40-person company in Dubai is not the same as buying it as a 4,000-person company in London, but the failure mode is related: the people who will use it daily are not the people who will sign.

You need four voices, even if some of them are the same person:

1. The operator — the person whose Monday morning breaks if the system is wrong.
2. The owner of the number — revenue, collections, inventory, whatever the system is supposed to move.
3. The person who will administer it — otherwise you are buying a tool nobody can change in month four.
4. The person who can say no to the contract — or the vendor will keep selling to whoever is friendliest.

If any of those four is “we’ll bring them in later,” later is when the regret starts.

Score the job, not the brochure

A useful shortlist is three vendors, not eight. Eight is how a committee hides from a decision. Three is enough to create competition and few enough that you can actually run the same scenario on each.

Write ten jobs. Not features — jobs. “Raise a quote from a WhatsApp lead without retyping it into two systems.” “Close the month without exporting to Excel.” “See which salespeople are discounting, this week, without a custom report from the partner.”

Then run one demo per vendor against those jobs, on your data, with your messy exception in the room. The vendor who refuses that demo has told you how implementation will feel.

Price is not the number on the quote

Licence fees are the visible line. They are often the smaller one. Implementation, integration, data cleanup, training, and the admin who has to exist after the partner leaves are the rest of year one. Treat the quote as a down payment until those are in the same conversation.

(There is a separate piece on this: the licence is not the cost of the system.)

Negotiate on your clock

The fastest way to buy the wrong system is to buy it because the current one expires on the 30th. Vendors know this. Auto-renewal windows of 30–90 days exist to make your calendar their leverage.

Start the process early enough that walking away is a real option. If walking away is not a real option, you are not negotiating. You are confirming.

What to do this week

• Write the ten jobs. If you cannot, you are not ready for a demo.
• Name the four voices. Put a date on a 90-minute meeting.
• Find the renewal / expiry date on every system this might replace. Put a reminder 180 days out, not 14.
• Cut any shortlist longer than three. If you cannot cut it, the jobs are not written clearly enough.

The tech check on this site is built to start that process without a sales call attached to it. Use it, or run the same sequence yourself. The sequence is the point — not the brand of software at the end of it.

Sources

• Johnny Grow, The CRM Software Failure Rate is 55% (2025)
• Gartner and Forrester CRM implementation studies (commonly cited ~50% / ~47%; definitions of “failure” differ)
• VendorBenchmark / enterprise SaaS contract reviews on auto-renewal notice windows (typically 30–90 days)