Choosing a CRM in 2026: a buyer's checklist for SMB and mid-market

CRM selection has become harder, not easier. Every vendor now demos an AI assistant, every platform claims to be low-code, and feature comparison grids have stopped being useful because everyone ticks every box. Here is how to evaluate on the things that actually differ.

Before you talk to any vendor

Write down the three processes the system must run well, the systems it must integrate with, and the number of people who will touch it in each role. One page. Send it to every vendor in advance. This alone changes the quality of the conversations you get.

Also decide who owns the decision and who merely advises. Selection committees without a named decision-maker default to the safest brand rather than the best fit.

Run scripted demos, not showcase demos

A vendor-led demo shows the platform's best path. Instead, send a script and make every vendor run the identical scenario using a sample of your data.

• Create a lead, qualify it, convert it, and produce a quote — while narrating each click.
• Change an approval threshold live, without a developer.
• Add a field to an object and get it onto a page and into a report in the same session.
• Show the record's audit history and explain who can see what.
• Show what the mobile experience looks like for a rep in a car park.

Watch for the phrase 'we'd handle that in an implementation phase'. Sometimes legitimate, often a tell.

Questions that reveal real differences

1. 1What can a business administrator change without a developer, and where is the exact boundary?
2. 2How are customizations preserved through platform upgrades?
3. 3What does the total cost look like in year three, including storage, sandboxes, API volume, and the tiers required for the features shown in the demo?
4. 4How do we get our data out — full export including attachments and history — and how long does it take?
5. 5What is included in support, and what is chargeable?
6. 6For AI features: where does our data go, is it used for training, and can we disable that?

Evaluating the AI claims

Ask for one concrete workflow, end to end, on your data. Not a chat window — a workflow with an input, an output, and an approval step. Ask what happens when the model is wrong, who reviews it, and how you audit it. Vendors with real capability answer these quickly; vendors with a wrapper change the subject.

Red flags

• Pricing that requires a call to understand at any level of detail.
• No sandbox available during evaluation.
• Reference customers who are all much larger or much smaller than you.
• 'Everything is customizable' with no explanation of what that costs to maintain.
• An implementation quote that has no data migration line item.
• Discount pressure tied to a quarter-end deadline.

Scoring without fooling yourself

Weight your criteria before the demos, not after. A simple split works well: 40% process fit, 25% adaptability without developers, 20% total three-year cost, 15% implementation partner quality. Adjust the weights to your situation, but fix them in advance — otherwise the scoring rationalises the preference you already had.

And weigh the partner as heavily as the platform. Two organizations buying the same CRM routinely get opposite outcomes based on who implemented it.

Buy for the process you run every day, not for the feature you saw once in a demo.