Stage gate software: what it has to do, and when you don't need it

You run the gates, and the gates run on a pack someone assembled last week. Stage gate software exists to close that gap. It puts current delivery evidence in front of the person deciding, records the go, hold or kill decision along with the reasoning, and keeps the gate schedule visible so no piece of work proceeds by default. Everything else a vendor bundles around that core — roadmaps, resourcing, timesheets — is general portfolio management. The gate function is narrower, and harder to do well.

This piece assumes you already know what a stage gate is. The question here is what software has to do to earn a place at the gate, how to tell the contenders apart, and when the honest answer is to buy nothing.

What the software has to make true

A gate works when three things are true: someone can say no, evidence sits in front of the decision-maker, and the decision gets recorded with its reasoning. A gate without a stop option is theatre, and software cannot fix it; the organisation grants the stop option. What software can do is make the other two conditions cheap enough that they actually hold, gate after gate, and make drift visible when they don't.

Judge any candidate against three tests.

• Live evidence, not a hand-assembled summary. The decision-maker should see the current state of delivery — spend, progress, risk — read live from the system where the work happens. A summary compiled the week before is an argument. Evidence is a record.
• A decision record with reasoning. Who decided, what was decided, what conditions were attached, what the evidence showed at the time. When someone reviews the investment two years later, that record is the difference between an answer and an argument about memory.
• A visible schedule and status. Every piece of work shows which gate it last passed and when the next one falls due. Work that has drifted past its gate is visible as drifted, and nothing proceeds just because nobody noticed.

Spreadsheets and slide decks fail all three tests, and they fail quietly. Someone assembles the steering pack by hand in the days before the meeting, so the numbers are stale by the time anyone reads them. The decision, if anyone captures it at all, lives in an email thread or a minutes document nobody can find at the next gate. And because no schedule shows which projects are overdue for review, work proceeds by silence. None of this looks like failure on the day; it looks like a normal meeting.

How to evaluate stage gate software

Where the delivery data lives

Start here, because it decides most of the rest. A gate tool that sits apart from the delivery system re-imports the staleness problem: someone still exports, reformats and pastes, and the gate still reviews last week. If work runs in Jira, the gate evidence should read from Jira. If it runs elsewhere, the tool needs a live connection to wherever it runs. An integration on a feature list proves little; ask to see the exact screen a decision-maker would open on gate day.

Evidence over summaries

Some stage gate process software is really presentation software: templates for gate packs, workflows for circulating them. That automates the theatre and leaves the substance untouched. Look for the raw material behind every number on the screen — the epics behind the progress figure, the register entry behind the risk rating, the actuals behind the spend line. If the decision-maker cannot click through to the source, the tool is asking to be trusted, and trust is what gates exist to replace.

The decision and audit record

Make the record structural. Decision, decider, date, reasoning, conditions, and the evidence as it stood — captured at the gate and immutable afterwards. Change requests belong in the same log, because scope drift between gates is where governance leaks. If an auditor asks why stage three was approved, one screen should hold the answer.

Who can say no

Permissions are governance. The tool should distinguish the people who prepare evidence from the people who decide, and stop one person doing both on the same investment. Separation of duties sounds bureaucratic until a project manager approves their own gate. The tool also needs a working reject path. Ask to see what happens when a decision is hold or kill; if nobody can show you, nobody has tested it.

Total cost, counted honestly

Price the licence, then price what it removes. A program running monthly gates across a dozen investments burns real days each cycle on pack assembly — exporting, reconciling, formatting, chasing. That labour is the true baseline. A cheap tool that leaves the assembly in place costs more than a dearer one that deletes it. Ask each vendor which of those hours disappear, and how.

When you don't need software at all

If the portfolio is three projects, one sponsor and a horizon of a couple of quarters, a platform is overhead. A recurring calendar entry, a one-page gate template and a sponsor willing to say no will outperform any tool, because the constraint at that scale is discipline, and software does not supply discipline. Software earns its keep when the portfolio grows past what hand assembly can serve, when turnover means nobody remembers why stage two passed, or when auditors will ask for the record. Buying a platform to compensate for a sponsor who never says no fixes nothing; the theatre just gets better production values.

Where Tollgate sits

Tollgate is a Forge app that runs inside Jira Cloud, built for teams whose delivery already lives there. Each investment carries a business case with NPV, IRR and payback. Work packages link to the Jira epics doing the delivery, so the gate reads progress from the work itself. A risk register, a decision log and a change log sit on the same record, with an audit trail and separation of duties between the people who prepare and the people who approve. Investment Health dashboards roll it up into RAG cards that serve as the steering pack; the meeting reads the dashboard, and nobody assembles a deck. Because Forge apps run on Atlassian's own infrastructure, data stays inside the Jira instance; nothing leaves for a third-party server. Pricing is per user through the Atlassian Marketplace, month to month.

Tollgate is on the Atlassian Marketplace. Install it on one live investment and run the next gate from the dashboard.