The Keep-or-Kill Audit: 100 Teams Told Us Which SaaS They'd Cancel Tomorrow

Renewal season is where software opinions become software decisions. Between September and October 2026, we surveyed 100 US teams, from five-person startups to a 400-person platform org, and asked two questions that cut through every feature matrix:

Which tool would you cancel tomorrow if nobody pushed back? And which one would you fight for in a budget meeting?

The second question is the interesting one. Nobody fights for a tool they tolerated.

How we asked

Respondents were engineering leads, IT managers, ops directors, and a handful of finance partners who sit in on renewals. We kept it open-ended on purpose. No vendor list, no checkboxes. Just names, reasons, and one follow-up: what happens the week after you turn it off?

That follow-up question did most of the analytical work.

The first-to-go list

The cancellation list was remarkably consistent across company sizes, and it clustered around four types of tools.

1. Duplicate note-takers and AI meeting recorders. The single most-named category. Teams accumulated three or four of these during the agentic-tool boom of 2024 and 2025 and never removed one. The complaint was never that the tool was bad. It was that nobody could remember which one had the good summary.

2. Tools bought for one project and never re-scoped. Survey platforms, diagram apps, migration assistants, a load-testing service. These survive on auto-renewal and inertia. Several respondents admitted they only knew the tool existed because the invoice showed up.

3. Dashboards nobody opens. Analytics add-ons and reporting layers that produce weekly screenshots into a Slack channel. The screenshots are the product. The dashboard itself is scenery.

4. Idle seats. Design tools, observability tiers, and premium support plans where the headcount on the invoice no longer matches the headcount on the team. This is the most fixable category and the most common form of quiet overspend.

The defended list

The keep list looked completely different, and the pattern was almost embarrassingly simple: the tools people defended were the ones where removal has a blast radius.

• Payroll and HRIS, because stopping it stops paychecks.
• Identity providers, because everything downstream breaks.
• On-call and paging, because the alternative is silence at 3 a.m.
• Source control and CI, because nobody wants to be the person who explains the outage.
• Support desks, because the tickets are the record.
• Design file systems, because the history is the asset.

Notice what is not on that list: anything adopted for speed last year. AI coding assistants landed in the middle, defended by heavy users and shrugged at by everyone else. That split is worth watching, but it is not the same as a must-keep.

Why stickiness is not the same as value

A tool can be sticky for reasons that have nothing to do with usefulness. We saw this play out in three patterns repeatedly.

Loud usage. Some tools generate the metrics that make them look adopted: dashboards, weekly digests, seat activations. Executives see activity. Individual contributors see a tab they close. Usage charts measure clicks, not outcomes.

Sunk-cost migration. Teams defended two tools purely because moving the data would take a quarter. That is a legitimate cost, but it is not a reason to renew forever. It is a reason to plan the exit properly.

Seat sprawl dressed as adoption. A tool with 80 seats and 12 weekly active users is not widely adopted. It is widely provisioned. Those are different numbers and only one belongs in a budget conversation.

The three questions that predict a keeper

Across 100 teams, the defenders all passed a short test that has nothing to do with features:

1. What breaks in the first hour if we turn it off? If the answer is a workflow, it stays. If the answer is a report someone reads on Mondays, it goes.
2. Who would complain, and how loudly? Tools defended by named individuals with real names attached to real workflows are safer than tools defended by nobody in particular.
3. Could we migrate in a weekend? Cheap exits make for better renewals, because the decision stays honest.

We think of this as a two-by-two: daily utility on one axis, switching cost on the other. High utility and low switching cost is the best place to be, because you keep the tool for the right reason. Low utility and high switching cost is the danger zone, and it is where most of the cancelled tools in our survey had been quietly living for years.

A pre-invoice checklist

The finance partners in our sample were the most ruthless respondents, and their routine is worth copying. Run it two weeks before the renewal date, not two days after:

• Pull the seat count. Compare it to active users from the last 30 days.
• Name the owner. A tool with no named owner has no defender.
• Write one sentence about what breaks without it.
• Check for a same-category duplicate already being paid for.
• Confirm the export path still works, so the migration option is real and not theoretical.

If a tool fails three of those five, it is a cancellation waiting for someone to say it out loud.

Where we land

The teams in our survey were not drowning in bad software. They were drowning in undecided software. The tools people wanted gone were almost never terrible products. They were products that had stopped being a decision and become a habit, an auto-renew, a tab nobody closes.

The keepers all share the same quality: if you removed them, someone would notice within the hour. That is a brutal filter, and it is also the only one that has ever produced a shortlist worth defending in a budget meeting.