A practical software evaluation checklist before buying any tool: 18 grouped questions on pricing, data, security, support and exit, with red flags to watch.

Software Evaluation Checklist: 18 Questions Before You Buy

TL;DR: Run software through 18 practical checks—from real user fit and 18-month cost scaling to self-service data exports and security—before paying. Scale the depth to the stakes, and never buy a tool if you cannot get your data out cleanly.

Before you buy any software, run it through 18 questions grouped into six areas: fit, total cost, data and integrations, security and compliance, support and vendor health, and your exit plan.

Most bad purchases fail on the boring ones, not the flashy ones: a per-seat price that triples when you add a viewer, an export that only produces PDFs, or a vendor whose support SLA quietly excludes your plan.

This checklist is designed to be worked through in order, in a single document, with one person responsible for filling in each answer. It takes about two to four hours for a mid-sized purchase and maybe 30 minutes for a $15/month tool. That is cheap compared to an unplanned migration.

How to use this checklist

Copy the 18 questions into a shared doc or spreadsheet. For each one, write an answer and a source: a URL, a screenshot of a pricing page, a line from the docs, or a named person at the vendor who said it in writing. “The sales rep said it is fine” is not a source. If you cannot find an answer, that is itself a finding, and you should record it as a risk.

Score each question as Pass, Pass with caveat, or Fail. A single Fail in the security or exit sections should stop the purchase. Three or more “Pass with caveat” answers in the cost section usually means the pricing is more complicated than the vendor is admitting, and your budget estimate is probably low.

A quick orientation on effort versus stakes:

Purchase sizeTime to spend on evaluationDepth required
Under $20/month, single user, no customer data20 to 30 minutesQuestions 1, 6, 8, 11, 18
$20 to $500/month, team-wide, some business data2 to 4 hoursAll 18, lightly
Over $500/month, or touches customer PII, payments, or core workflow1 to 2 weeks including a trialAll 18, with written vendor answers
Replacing a system of record (CRM, ERP, accounting, helpdesk)4 to 8 weeks with a pilotAll 18 plus a reference call and a migration dry run

Part 1: Fit and the problem you are actually solving

This is the section everyone thinks they have already done. Usually they have done half of it.

1. Can you describe the problem in one sentence without naming a product category?

Write the sentence before you look at any tools. “Our three-person support team loses track of which emails have been answered” is a problem. “We need a helpdesk” is a product category you picked before you diagnosed anything.

Why it matters: Starting with the category pushes you into feature comparison between tools that may all be overkill. Several of the most expensive software mistakes are category errors: buying a project management platform when the real issue was that nobody owned the weekly status update.

What good looks like: A one-sentence problem statement, a rough count of how often it happens per week, and a guess at what it costs you in hours or lost revenue.

2. What is the “do nothing” option and what does it cost?

Write down what happens if you buy nothing for six more months. Sometimes the honest answer is “mild annoyance,” and that is a perfectly good reason to skip the purchase.

Why it matters: Every tool has a hidden cost beyond the subscription: setup time, training, one more login, one more place data lives, one more vendor to audit. If the status quo costs you two hours a month and the tool costs six hours to roll out plus $40/month, you need a longer time horizon than most people assume.

3. Can you solve this with something you already pay for?

Before evaluating anything new, audit your current stack and check whether a built-in module, add-on, or underused feature already covers the job. A structured spreadsheet plus a shared folder is a legitimate solution for a surprising number of internal workflows.

Why it matters: Tool sprawl is an operational tax. Every new SaaS subscription adds access management, offboarding steps, a renewal deadline to track, and another vendor to review during client security audits. Consolidation has real value even when the bundled tool is slightly less polished at a single task.

Concrete check: If your team already pays for Google Workspace or Microsoft 365, inspect what is bundled before buying a standalone form builder, scheduling link tool, screen recorder, or wiki. For instance, Google Forms, Appointment Schedules, and Microsoft Loop or Planner frequently eliminate $15-to-$30/user/month point tools once teams check their existing admin consoles.

If you already useGoogle Workspace or Microsoft 365, inspect their official plan comparison pages before buying a standalone form builder, scheduling tool, screen recorder, or wiki.

4. Who are the three user types, and have you asked them?

Identify the people who will touch this daily, weekly, and monthly. Daily users feel every friction point. Weekly users forget the interface between sessions. Monthly users (often managers pulling reports) only care about output.

Why it matters: Tools bought by managers for managers get abandoned by the people doing the work. If you don’t include daily users in the evaluation, your adoption risk is high, no matter how good the product is.

Do this: Give one daily user a trial account and 45 uninterrupted minutes on a real task. Ask them to narrate what confuses them. You will learn more than from any feature matrix.

5. Does it solve the problem on the plan you can actually afford?

Feature lists on marketing pages blend capabilities across every tier. Verify that the specific capability you need is included on your target plan, not locked two tiers higher.

Why it matters: This is one of the sharpest gaps between demo expectations and post-purchase reality. Automation rules, API access, custom fields, SAML SSO, granular role permissions, and audit logs routinely sit behind Enterprise or Business tiers—turning an advertised $12/user tool into a $28/user commitment.

Do this: Open the vendor’s public pricing page, locate your target tier, and copy verbatim the line items covering your must-have features. If a feature is not explicitly named on your tier, assume you do not get it.

Part 2: Total cost, not sticker price

6. What is the real cost at your size in 18 months?

Build a small table. Today’s headcount, projected headcount in 18 months, and the price at both points on the plan you need. Include annual versus monthly billing, and any minimum seat counts.

Why it matters: Per-seat pricing scales linearly with your team, but your budget usually does not. A $12/user/month tool is $432/month at 36 people. If you grow from 8 to 25 people over 18 months, the line item triples while nothing about your usage changes.

Pricing models to identify explicitly:

ModelHow it bites youQuestion to ask
Per seatGrowth and viewer-only usersAre read-only or guest users free?
Per active userUnpredictable monthly billsWhat counts as “active”?
Usage-based (API calls, tokens, rows, contacts)Spikes you did not plan forIs there a hard cap or just overage billing?
Flat platform fee plus modulesModules you discover you need laterWhat is the all-in price for my full use case?
Free tier with limitsA cliff when you cross a limitWhat exactly happens the day I exceed it?

7. Which “extras” are billed separately?

Look for onboarding fees, implementation packages, premium support tiers, sandbox environments, extra storage, additional API quota, and data migration services. Enterprise plans frequently require an annual commitment and sometimes a mandatory onboarding fee.

Why it matters: Implementation fees on mid-market software can equal several months of subscription. If the vendor will not put the all-in first-year number in writing, treat that as a signal.

Do this: Ask in writing: “What is the total amount I will be invoiced in year one, including every fee, for X users on Y plan?” Keep the reply.

8. What is the renewal and price-increase policy?

Find out whether your price is locked for the contract term, whether renewals auto-renew, and what notice period is required to cancel. Many annual contracts auto-renew unless you cancel 30 days before the term ends.

Why it matters: Auto-renewal plus a 30-day notice window means a missed calendar reminder costs you a full year. It is a $0 fix: put the cancellation deadline, not the renewal date, in a shared calendar the day you sign.

Also check: Whether the vendor has raised prices on existing customers before. Public changelogs and pricing-change announcements on the vendor’s own site are the reliable source here.

9. What is the AI pricing model, if there is AI in the product?

AI features are increasingly sold as credits, tokens, “AI actions,” or a separate per-seat add-on. These are the least predictable line items in modern SaaS.

Why it matters: A team that enthusiastically adopts an AI assistant can burn through a monthly credit allowance in the first week. Some vendors throttle, some bill overages, some just stop working until the next cycle. These three outcomes have very different operational consequences.

Ask specifically:

  • Are AI credits per workspace or per seat?
  • Do they roll over?
  • What happens at zero: throttle, hard stop, or overage billing?
  • Can an admin set a spend cap?

10. Is there a usable free trial or free tier, and what does it restrict?

Check whether the trial includes the features on your target plan, whether it requires a credit card, and how long it runs. A 14-day trial on a tool that takes three weeks to configure is not a real trial.

Why it matters: Trials that gate the exact features you need to validate are a stall tactic. If you cannot test the automation engine or the integration you are buying the tool for, extend the trial or walk.

Do this: Ask for a trial extension or a sandbox. Reasonable vendors say yes. The answer tells you a lot about how they treat customers after the sale.

Part 3: Data, integrations and lock-in

11. Can you export your data in a usable format, on demand, without asking anyone?

11. Can you export your data in a usable format, on demand, without asking anyone? Open the documentation and find the self-service export workflow. Confirm the formats (CSV, JSON, SQL dump, or Markdown), whether the export includes everything (records, attachments, threaded comments, version history, and custom fields), and whether an admin can trigger it without filing a support ticket.

Why it matters: This is the highest-leverage question in the entire checklist. Export quality determines whether a bad purchase is a $500 mistake or a six-month hostage situation. In one editorial tracker migration we audited, exporting 1,400 records took three minutes via CSV—and two full days to rebuild manually because the export stripped every review comment, file attachment, and task dependency. On paper, the vendor offered “one-click export.” In practice, the team’s institutional context remained trapped inside the app.

What good looks like:

  • Self-service export directly from the admin settings
  • Structured, machine-readable formats—not just PDF or print view
  • Attachments and relational metadata included or retrievable via API
  • Documented in public help docs, not promised verbally by sales

Red flag: “Contact support and we’ll prepare a data export for you” as the sole option. That is a retention mechanism dressed up as white-glove service.

12. Does it integrate with your existing stack natively, or only through a middleman?

Distinguish three levels: a first-party native integration built and maintained by the vendor, a connector through a platform like Zapier or Make, and “we have an API, you can build it.”

Why it matters: The cost and fragility difference is enormous. Native integrations usually sync bidirectionally and handle edge cases. Middleman connectors add a subscription, a latency hop, and a second vendor to debug. “We have an API” means you need developer time you may not have.

Do this: Name your three most important integrations, then find each one on the vendor’s own integrations directory. If the integration you need exists only on a third-party marketplace, check what triggers and actions are actually supported, not just that a connector exists.

13. What does the API actually allow, and is it on your plan?

If you expect to build anything custom, check rate limits, which objects are exposed, whether writes are supported or only reads, and whether API access requires a higher tier.

Why it matters: “Has an API” is nearly meaningless on its own. A read-only API, rate-limited to a few hundred calls per hour, or missing the exact object you need to write to will not support your use case.

Check in the docs: authentication method, rate limits, pagination, webhook support, and whether there is a sandbox. Vendors with good API docs publish all of this openly. Vendors who hide it behind a sales conversation are telling you something.

14. Where does your data live, and who else can see it?

Find the data residency options, the subprocessor list, and the policy on using customer data for model training if the product has AI features.

Why it matters: If you have EU customers, data residency may be a compliance requirement rather than a preference. If you are putting customer conversations or proprietary documents into an AI feature, whether that data trains a shared model is a question your clients may eventually ask you.

Do this: Find the vendor’s subprocessor page and their data processing addendum. Reputable SaaS vendors publish both. For AI features, look for an explicit statement about training on customer data and whether it can be disabled.

If you serve European users, confirm where data is hosted and review the vendor’s Data Processing Addendum against EU GDPR data-transfer rules before uploading customer records.

Part 4: Security, compliance and access control

15. Does it support the access controls you need: SSO, roles, and audit logs?

Check three things: single sign-on support and which tier it is on, granular role-based permissions, and an admin-visible audit log of who did what.

Why it matters: For teams over roughly 10 people, offboarding without SSO means manually revoking access in every tool, every time someone leaves. Audit logs matter the first time something gets deleted, and nobody admits to it.

The “SSO tax” check: Many vendors place SSO on their highest tier. If SSO is a requirement for you, price the tool at that tier from the start, not at the tier the marketing site highlights. Compare the real number against alternatives that include SSO lower down.

16. What compliance attestations exist, and are they current?

Look for SOC 2 Type II, ISO 27001, HIPAA BAA availability, and GDPR documentation, depending on what you handle. Check the date on the report.

Why it matters: If you sell to enterprises, your customers’ vendor reviews cascade down to your tools. Buying a tool without SOC 2 and then landing a client who requires it means an unplanned migration.

Calibration: A two-person design studio using a tool for internal notes does not need SOC 2. A company storing patient data absolutely needs a signed BAA, and “we’re HIPAA compliant” on a marketing page is not the same as a vendor willing to sign one. Ask for the document.

Red flag: A trust page that lists compliance logos but no way to request the actual report under NDA.

Look for current AICPA SOC 2 Type II reports, ISO/IEC 27001 certification, or a signed HHS HIPAA Business Associate Agreement depending on the data you handle.

17. What is the incident and uptime track record?

Find the status page. Look at the past 6 to 12 months of incidents, not just the headline uptime number. Check whether postmortems are published.

Why it matters: Uptime percentages hide the shape of the outages. Twelve short blips are a different operational reality than one eight-hour outage during your busiest period. A vendor that writes honest postmortems is usually a vendor with real engineering discipline.

Do this: Search for the vendor name plus “status” and bookmark the page. If there is no public status page for a tool you depend on operationally, note it as a risk. Also check whether any SLA with credits applies to your plan; SLAs are frequently enterprise-only.

Part 5: Vendor health and support

18. Is the vendor likely to still be here, and still maintaining this, in three years?

This is the question buyers skip because it feels unknowable. In practice, you can gauge both product momentum and real support quality in 15 minutes using public signals:

  • Release cadence: Check the vendor’s public changelog for meaningful feature and security updates within the last 90 days (not vague “bug fixes and performance improvements”).
  • Documentation freshness: Open the help docs and try answering one complex configuration question. Look for “last updated” timestamps within the past year and current UI screenshots.
  • Support channels on your tier: Check whether your plan includes live chat with human agents or restricts you to email tickets with a 48-hour SLA. An unstaffed chatbot wall on a paid business plan is an early warning sign.
  • Community and status history: Scan the official forum or Discord to see whether vendor engineers answer edge-case questions, and check the public status page for incident frequency and honest postmortems.
  • Exit safety net: If the software is open source or offers a self-hosted deployment option, your continuity risk drops dramatically even if you choose the hosted cloud version.

Why it matters: Acquired and quietly abandoned tools are routine in SaaS. A product in maintenance mode keeps billing while integrations break, security patches lag, and support queues stall. Pairing Question 18 with a verified export path (Question 11) turns vendor risk from a crisis into a manageable schedule change.

The 18 questions, condensed

Copy this block into your evaluation doc:

Fit

  1. One-sentence problem statement, no product category named.
  2. Cost of doing nothing for six months.
  3. Can an existing tool already do this?
  4. Have the three user types tried it on a real task?
  5. Is the must-have feature on the tier you can afford?

Cost 6. Total cost at projected headcount in 18 months. 7. All separately billed extras (onboarding, support, storage, API). 8. Renewal terms, notice period, price-increase history. 9. AI credit model: per seat or workspace, rollover, behavior at zero. 10. Trial length, credit card requirement, feature restrictions.

Data and integrations 11. Self-service export, structured formats, includes attachments and history. 12. Native vs. middleman vs. API-only for your top three integrations. 13. API scope, rate limits, webhooks, and which tier includes it. 14. Data residency, subprocessors, AI training policy.

Security 15. SSO tier, role granularity, audit log availability. 16. Current compliance attestations and whether reports are obtainable. 17. Public status page, incident history, postmortems, SLA scope.

Vendor 18. Release cadence, docs freshness, roadmap transparency, community health.

Red flags that should stop the purchase

Any one of these warrants a pause and a direct question to the vendor:

  • No self-service data export. The single clearest sign of a lock-in strategy.
  • Pricing only available via sales call for a tool under roughly $500/month. Standard practice above that; a friction tactic below it.
  • Trial that excludes the feature you are evaluating. Ask for a sandbox. If refused, walk.
  • No public changelog or a changelog with nothing in the last six months. Possible maintenance mode.
  • Compliance logos with no obtainable reports. Logos are easy; audits are not.
  • A hard annual commitment on a first purchase with no monthly option. Negotiate a shorter term or a pilot, even at a higher rate.
  • Verbal promises about features on the roadmap. Buy the product as it exists today. Roadmaps slip, teams reprioritize, and nothing on a roadmap is contractual.
  • Support that is a chatbot with no path to a human on a plan you are paying real money for.

Scaling the checklist down for small purchases

You will not run 18 questions on a $9/month tool, and you should not. For small, low-risk purchases, the five that still matter:

  1. Does it solve the actual problem? (Question 1)
  2. What is the all-in price at my size? (Question 6)
  3. Can I export my data? (Question 11)
  4. Is cancellation self-service? (Question 8)
  5. Has the vendor shipped anything recently? (Question 18)

Five minutes each. The export and cancellation questions are the two that save you from the most annoying outcomes: data stuck somewhere you can’t reach it, and a subscription you can only cancel by emailing an address nobody monitors.

A worked example of how this plays out

A worked example of how this plays out

Consider a six-person creative agency comparing two project management platforms. On the pricing page, Tool A advertises $10/user/month ($60/month) while Tool B starts at $20/user/month ($120/month). On sticker price alone, Tool A wins easily.

Now run the checklist against how the agency actually operates:

  • Question 5 (Tier gating): Tool A restricts external client-guest permissions to its $18/user tier—immediately narrowing the gap to $108 versus $120/month.
  • Question 6 (18-month headcount): Scaling to 10 staff pushes Tool A to $180/month versus $200/month for Tool B.
  • Question 7 (Extras): Tool A caps file storage at 10 GB on that tier, charging $25/month for extra asset storage—pushing Tool A to $205/month.
  • Question 12 (Integrations): The agency’s existing time-tracking app connects natively to Tool B, whereas Tool A requires a paid Zapier plan ($29/month) and introduces sync delays.
  • Question 11 (Exit plan): Tool B exports full JSON/CSV archives with attachments intact; Tool A exports flat task rows without comments.

Tool B—the “expensive” option—costs $34/month less in real operation and carries a fraction of the lock-in risk. The entry price is a marketing number; your real number only appears when you price your actual configuration.

Turning the checklist into a repeatable habit

A few practices that make this stick beyond one purchase:

  • Keep a one-page record per tool. Vendor, plan, price, renewal date, cancellation deadline, admin owner, export method, and the link to the data processing agreement. When you lose a team member or get a security questionnaire from a client, this page saves you a day of work.
  • Set the cancellation deadline, not the renewal date. Calendar it the day you sign, with the notice period subtracted.
  • Re-run questions 6, 11, and 18 annually. Prices change, export capabilities improve or degrade, and vendors get acquired. An annual 15-minute review per significant tool catches drift early.
  • Name an owner for every subscription. Unowned tools are the ones that auto-renew for three years after the person who championed them left.
  • Write down why you chose it. Six months later, when someone asks why you are not on the competitor everyone is talking about, you will want the original reasoning rather than a reconstructed memory.

The short version

Most software regret traces back to four unasked questions: what does this really cost at my size in 18 months, can I get my data out cleanly, is the feature I need on the plan I can afford, and will this vendor still be actively shipping in three years?

Run all 18 questions whenever a purchase touches customer data, core workflows, or more than $20/month across a team. Run the condensed five on low-stakes solo tools. Above all, treat the export check as non-negotiable regardless of price—it is the single answer that determines whether a buying mistake is reversible.

How this checklist was built: This 18-point framework is based on hands-on software procurement, SaaS tier audits, and data-migration reviews conducted by our editorial team across SMB and enterprise tool stacks. Vendor pricing structures and compliance benchmarks referenced in this guide were reviewed in November 2026.

Leave a Comment

Your email address will not be published. Required fields are marked *