Phase plan template

A phase plan is how authority over an ad account widens in steps. One row per area per phase: what may change without a separate approval, inside what limits, what still needs a person, who holds the authority, and what has to be true before the next phase begins.

It applies equally to a new team member, an agency, or an automated rule. The mechanism is the same in every case: a narrow permission, a limit, a record of what was done under it, and a reversal rule written before anybody needs it.

Write it at the start of any arrangement where somebody other than the account owner will be making changes. That is nearly every arrangement, and the alternative is the two states most accounts oscillate between: everything needs approval, so nothing happens quickly, or nothing does, so nobody can say what changed.

Use it again at every review, because a phase that is never re-read has already become permanent.

Authority over an ad account usually sits in one of two unsatisfactory states.

In the first, every change needs approval. That sounds safe and produces its own failure: changes that should take a minute take three days, the person closest to the account stops proposing them, and the account drifts because nobody can act at the speed it moves.

In the second, somebody has authority over everything, and nobody can say afterwards what was changed, by whom, or why. This is usually described as trust. It is really an absence of record, and it is discovered at the worst possible moment, which is when something has gone wrong and the history is the only way to understand it.

A phase plan is the third option. Authority is granted narrowly, per area, with a limit on it, and it widens on evidence rather than on familiarity.

Authority is granted per area because the areas do not carry the same risk.

Being comfortable with somebody rotating creative says nothing about being comfortable with them moving budget. Changing an audience definition is reversible; restructuring an account is not, in any practical sense, because the history goes with it. A single blanket permission forces the most cautious answer onto every area, which is why blanket permissions are either too narrow to be useful or too wide to be safe.

So each area has its own row and its own phase, and they advance at different speeds. It is normal for creative authority to reach phase three while structural authority is still at phase one, and the plan should make that visible rather than treating it as inconsistency.

A permission without a limit is a handover. The limit column asks for three things: how much, how often, and on what.

How much bounds the size of a single change — a proportion of a campaign's budget, a number of assets, a range for a bid. How often bounds the frequency, which matters more than people expect: an unlimited number of small changes is a large change made invisibly. And on what bounds the scope: inside one flight, on one channel, on campaigns below a certain size.

Those three make a permission explainable. The test for a good limit is whether the person holding it would be comfortable describing a change they made under it, afterwards, to the person who granted it.

The approval column is written explicitly rather than left as everything not listed above.

That seems redundant and it is not, because at speed people do not reason from an absence. Writing down that any change to the total budget, any new campaign, and anything touching a different flight still needs a person means the boundary can be checked in a second, by somebody who is busy, which is the only time it matters.

It also surfaces disagreement early. The first draft of this column is usually where an account owner and a media buyer discover they had different assumptions.

Two columns decide how phases move, and both are written before the phase starts.

The evidence to enter is what had to be true first: a period of clean reporting, a reconciliation with no unexplained differences, a rule tested against past data, a handover completed. The evidence to advance is what has to be true before the next phase begins.

Writing the second one in advance is the important half. Decided at the time, the question of whether to widen authority is settled by whoever is most optimistic, usually in a meeting where things are going well. Decided in advance, it is settled by whether the named condition happened.

Every row carries a rule for narrowing authority again, and it is written when nothing is wrong.

That timing is the whole point. A reversal decided in the middle of a problem is read as blame, and it becomes a negotiation about whether somebody was at fault. A reversal that follows a rule everybody agreed to in week one is a mechanical consequence, and stepping back a phase becomes a normal move rather than a demotion.

Make the trigger specific: changes made outside the limits, changes made without a record, a result outside an agreed range. And name who decides, so the rule does not depend on somebody volunteering.

The record column asks what is written down each time the authority is used: what changed, from what to what, why, by whom, and when.

Without it, none of the other columns can be checked. The evidence to advance cannot be assessed, the reversal rule cannot be applied, and a review of the phase is a conversation about impressions.

It does not need to be elaborate — a dated line per change is enough — but it does need to exist at the moment of the change rather than being reconstructed later. Reconstructed records are, reliably, records of what somebody remembers intending.

Seven checks. There is one row per area, and no blanket permission. Every permission has a limit on size, frequency and scope. What still needs approval is written explicitly. Every row names who holds the authority. The evidence to enter and to advance were both written before the phase started. Every row has a reversal rule, written while things were going well. And the record that will be kept is named, and somebody has agreed to keep it.