Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Frameworks

Phase-Gate Reviews

Phase-gate reviews explained: splitting a launch or market entry into phases with criteria agreed in advance, the decision at each gate, an example and limits.
Frameworkintermediate9 min readUpdated

What phase-gate reviews are

The biggest marketing commitments rarely have to be made all at once, though they are often treated as if they did. A new market, product or channel can be entered in steps, each small enough to learn from and to stop without great loss. Phase-gate reviews formalise those steps: each phase has an objective and a budget, and between phases sits a gate, a meeting or written decision at which the results so far are judged against criteria agreed before the phase began.

The gate is a decision, not a status update, with a short list of outcomes: go on as planned, go with changes, hold while a specific problem is fixed, or stop. A gate that can only ever say go is a calendar milestone with a meeting attached.

Gates belong to work where the cost of being wrong rises sharply as the work scales. The go-to-market planning guide sequences a launch into pre-launch, launch and sustain, with a gate between each, and the guide to choosing and prioritising markets ends with a pilot whose gate decides whether to expand, hold or close. This page covers the gate method itself, which applies to both and to other large commitments, such as moving serious budget into an untried channel. It does not cover the checks that decide whether a single campaign is ready to go live; the campaign workflow from intake to launch owns those review and approval gates, which catch errors rather than judge how the market responded.

Phases and gates

Most marketing initiatives follow a similar arc: explore, test small, launch under control, then scale. Each gate asks a different question and needs different evidence.

GateQuestion it answersEvidence it needsPossible decisions
Idea to explorationIs this worth research time?A problem statement, a rough sizing, a fit with the strategyExplore, park, reject
Exploration to pilotIs there reason enough to spend real money on a test?Customer research, a view of competitors, a costed pilot plan with its criteriaRun the pilot, research further, stop
Pilot to launchDid the pilot show the response the plan depends on?Pilot results against the criteria, cost per result, early repeat or retention behaviour, operational problemsLaunch, rerun with changes, hold, stop
Launch to scaleDoes the response hold at volume and at an acceptable cost?Steady-state cost and conversion, retention by cohort, capacity and service dataScale, scale more slowly, hold, wind down
Review after scaleIs this still earning its budget?Results against the plan's measures, changes in the marketContinue, reshape, retire

Not every initiative needs every gate. A small test in an existing channel might have one gate at its end; a new country with its own site and media budget might use all five, plus a go-live check such as the new market launch checklist. That checklist is a readiness gate: it confirms the market is ready to receive customers, not that customers want what is offered.

The difference between readiness and performance matters at every gate. Readiness criteria ask whether the work is built: tracking, pages, stock, support, legal review. Performance criteria ask whether the market responded. A gate that checks only readiness approves a launch that is well built and unwanted.

Writing gate criteria in advance

Criteria are the whole method. Written before the phase starts, they turn the gate into a test; written afterwards, they turn it into a story. Useful criteria share four features.

They are specific. "Strong interest" is not a criterion; a stated level of sign-ups or a ceiling on acquisition cost is. Each criterion names the KPI, the threshold, the period it covers and the data source, so nobody argues at the gate about which report to trust.

They include guardrails as well as goals. A pilot can hit its sign-up target by discounting heavily, buying low-quality traffic or overloading the service team. A guardrail metric such as refund rate, complaint volume or margin per order sets the limit within which the goal must be reached. Without one, gates reward the cheapest way to move the headline number.

They say what happens if they are not met: which shortfall leads to a hold, which to a rerun with changes and which to a stop. Writing the consequence in advance is uncomfortable, and that is the point; it is far harder to write after months of effort.

They fit the phase. A pilot cannot prove steady-state economics, so its criteria cover what it can show: response from the chosen segment, early retention, the cost of reaching people in the channel. Asking a pilot to prove profit at scale guarantees a failed gate or a fudged one.

Where uncertainty is high, use ranges: a level above which the work goes on, a level below which it stops, and a band between in which it holds for a defined period. The go-to-market plan template has a gates field for recording criteria and consequences phase by phase, so they travel with the rest of the go-to-market plan rather than living in someone's notes.

Running the review

A gate review needs four things: a decision-maker, an evidence pack, a recommendation and a record.

The decision-maker is one person who controls the next phase's budget. The team that ran the phase prepares the evidence and recommends an outcome but does not make the call, because the people who have spent months on the work are the worst placed to stop it.

The evidence pack sets each criterion beside its result, in the order the criteria were written, before any commentary. Context follows: what changed during the phase, what the criteria did not anticipate, and what the next phase would cost. Circulate it ahead of the meeting so the time is spent on the decision.

The recommendation names one of the agreed outcomes and gives the reason. If the team wants to change a criterion because the phase showed it measured the wrong thing, it says so openly, and the decision-maker rules on that change separately from the verdict. Quietly lowering the bar is the most common way gates fail.

The record holds the decision, the reason, any conditions and the criteria for the next gate. Kept together, the records show whether the organisation ever holds or stops anything, which is the real test of whether its reviews work.

Worked example (illustrative)

Illustrative example — not a real result

A fitness-studio chain decides whether to scale an online class subscription

Fernquill Fitness, an invented chain of 14 fitness studios, pilots an online class subscription at £12 a month in one city before deciding whether to offer it across every location. Before the 12-week pilot starts, the leadership team agrees three criteria for the pilot-to-launch gate: at least 600 paid sign-ups, at least 70% of week-one subscribers still paying in week eight, and an acquisition cost below £30 per member. A guardrail sits beside them: studio membership cancellations in the pilot city must not rise more than 2 points above the other cities. If sign-ups and cost pass but retention lands between 55% and 70%, the agreed outcome is an 8-week hold to work on retention, not a launch.

At the gate the pilot shows 740 sign-ups and an acquisition cost of £24, but only 58% of week-one subscribers are still paying in week eight. Studio cancellations are flat. Exit surveys point to a thin class library after the first month. The marketing director, who owns the budget, takes the pre-agreed hold: the team adds 40 classes, keeps the 70% retention criterion for the next cohort and books the next gate for week 20.

Limits and criticisms

Gates add overhead. For small, cheap or easily reversed work, a formal review costs more than a wrong decision would: a new subject line needs a measure and a date to look at the result, not a gate meeting. Keeping gates for commitments that are expensive or hard to undo keeps them meaningful.

They turn into rubber stamps. When criteria are vague, or the sponsor has already announced the launch internally, the review approves whatever arrives. The warning sign is a history of gates that never produced a hold or a stop. Specific criteria, written consequences and a decision-maker who is not the sponsor reduce the risk, but cannot remove it, because a gate is still a meeting among people whose reputations ride on the outcome.

Sequential phases slow learning. A strict structure assumes each phase must finish before the next begins, yet creative testing, pricing research and channel tests rarely need to wait for one another. Teams that run everything in series learn more slowly than teams that run independent questions in parallel and keep the gate for the next commitment of money.

Criteria set early can turn out to be wrong, because the phase may show that a different measure matters more. Treating them as untouchable stops good work or continues bad work for the wrong reason. The answer is an open, recorded change, decided apart from the verdict, not abandoning criteria.

Finally, a launch can pass every gate and still be overtaken by a competitor or a shift in demand. Gates reduce the cost of being wrong about the business's own work; they do not predict the market.

Gates decide when to commit more; other frameworks shape each phase. The experimentation loop runs within a pilot, producing the changes the next gate judges. One-way and two-way door decisions help decide which commitments need a gate at all, since easily reversed changes rarely justify one. Scenario planning prepares responses to changes outside the plan, which a gate review can draw on when the market moves between phases. For launches written in SOSTAC form, SOSTAC for a product launch shows how a readiness gate and written triggers sit in the plan's action and control sections.

Common mistakes

  • Gates without consequences. Criteria that describe success but not what happens when it is missed lead to a go by default.
  • Criteria written after the results. Thresholds chosen once the numbers are in measure the team's skill at framing, not the market's response.
  • Only readiness checks. A gate that confirms the work is built, but not that anyone wants it, approves well-made failures.
  • The sponsor as the judge. The person who proposed the work should present it, not decide whether it continues.
  • Open-ended holds. A hold without its own deadline and criteria becomes a phase that never ends.
  • Gating everything. Formal reviews on small, reversible work waste time and teach people to treat gates as paperwork.
  • No guardrails. Goals reached through deep discounts or cheap traffic pass the gate and fail at scale.

Frequently asked questions

How is a phase gate different from an approval step in a campaign workflow?

An approval step checks whether a piece of work is ready to go out: the copy is accurate, the legal review is done and the tracking fires. A phase gate decides whether the whole initiative has earned the next commitment of money and effort, judged by what the last phase showed about the market. A campaign can pass every approval step and still fail a gate because customers did not respond.

Who should make the decision at a gate?

One named person who controls the budget for the next phase, advised by the team that ran the last one. The team presents the evidence against the criteria and recommends an outcome; the decision-maker chooses and records the reason. When a committee decides by consensus, a hold or a stop becomes very hard to reach, because someone in the room is always attached to the work.

What should happen when a gate is failed?

The failure should lead to one of the outcomes written down before the phase began: hold while a named problem is fixed, change the plan and rerun the phase, or stop. Moving the criteria after the results are in, so that the work scrapes through, is the outcome to avoid. A hold needs its own deadline and its own criteria, or it turns into an open-ended extension of the phase it was meant to judge.

Terms used on this page

Where this page is used

Also related

Related