How Many Stage Gates Should a Project Actually Have?

## The short answer

Most projects work well with **three to five stage gates**. Fewer than three and you lose meaningful control points; more than six and the process becomes a tax that teams resent and circumvent. The exact number should scale with how risky, valuable and irreversible the decisions are — not simply with how big the project looks on a Gantt chart.

## Why "more gates" is not "more control"

It is tempting to assume that adding gates adds rigour. In practice, each gate carries a cost: preparation time, board attendance, scheduling delay and the temptation to write documents purely to pass. Beyond a handful of gates, you get diminishing returns on assurance and rising friction. The aim is the *minimum number of decision points that genuinely change what happens next*.

A useful test: if failing a gate would not realistically change the decision to continue, that gate is theatre. Remove it.

## A sensible default: five gates

A robust general-purpose model uses five gates that each ask a distinct question:

1. **Concept** — Is this problem worth exploring at all?

2. **Business case** — Do the expected benefits justify the expected cost and risk?

3. **Design / readiness** — Is the chosen approach viable and is the team ready to build?

4. **Pre-launch / go-live** — Is it safe, compliant and operationally ready to release?

5. **Closure / benefits** — Did we deliver the value, and what do we carry forward?

Every one of these changes a real decision. You can collapse some for smaller work.

## Scale the number to the three Rs

Rather than a fixed count, tune gates to risk profile:

- **Risk** — How likely and how damaging is failure? High-risk work earns more gates, particularly around design and go-live.

- **Reach (value)** — Larger investments and wider impact justify more funding checkpoints.

- **Reversibility** — Easily reversed decisions (a feature flag) need less ceremony than irreversible ones (a data migration, a public commitment, a regulatory filing).

Low risk, low value, easily reversed? Two or three gates. High on all three? Five or six, with a hard go-live gate.

## Tier your governance

The cleanest way to manage this across a portfolio is to define **governance tiers** and assign each initiative to one:

- **Tier 1 (light)** — small, low-risk work. One concept check and one closure review.

- **Tier 2 (standard)** — the five-gate default.

- **Tier 3 (heavy)** — major or regulated programmes, with additional gates for compliance, security and operational acceptance.

A tiering rule set means teams know upfront what governance applies, and boards stop applying enterprise ceremony to a two-week task. Enterprise delivery platforms — the kind neart.ai builds — typically encode these tiers so the right gates appear automatically based on an initiative's risk and value attributes.

## Signs you have too many gates

- Teams batch up work to "survive" until the next gate rather than flow continuously.

- Gate packs are written for the board, not used by the team.

- Decisions are rubber-stamped because stopping at that point is no longer realistic.

- The calendar, not the work, drives when value is delivered.

## Signs you have too few

- Spend escalates with no checkpoint to question it.

- Go-live happens without a clear safety and readiness decision.

- Benefits are never reviewed, so nobody learns whether the investment paid off.

- Failing initiatives run on because no one is empowered to stop them.

## Make each gate cheap to run

Once you have the right *number*, the next lever is *cost per gate*. Standardise a one-page decision record, draw evidence from live delivery artefacts, and time-box the board conversation. A cheap gate can be run more readily; an expensive gate tempts teams to avoid it. The number of gates and the cost of each gate are two dials — tune both.

## Takeaway

Aim for three to five gates as a baseline, then adjust by risk, value and reversibility rather than raw project size. Tier your governance so small work gets light-touch control and major work gets the full set, and keep each gate cheap enough to run that nobody is tempted to route around it.