Skip to content
Back to Guides
Strategy7 min read5 chapters

Reviewing Agent Approvals: The Five-Minute Daily Routine

How to read a proposal, a five-minute daily routine for the Approvals queue, when to approve, reject or let a proposal expire, and how to check what ran.

Shubham Raghav

Chapter 1Why the Queue Is the Control

At Ask first, the level every brand starts at, nothing an agent wants to change in your accounts happens until a person approves it. That makes the Approvals queue the control. It only works if somebody opens it.

A proposal nobody answers expires after seven days. Nothing happens, which feels safe, and often is not: a pause that was right on Monday and never made is an ad that kept spending all week. An expired proposal is a decision nobody made.

Ask first protects you only if somebody asks back. A five-minute look once a day is enough for most brands, and the routine below is built to fit in it.

Chapter 2Reading One Proposal

The number on the inbox icon in the header, and on the Approvals row in the Agents menu, is the same count: everything waiting for a decision. The Pending tab lists it. Each row is built to be decided from the row itself.

Part of the rowWhat to check
Who and whenWhich agent or person raised it, how long ago, and how long it has left before it expires.
Where it came fromThe finding or recommendation behind it, when an agent raised it from one.
AccountThe exact account and object the change touches. Check it is the one you think.
ChangeWhat it is now and what it would become.
ExpectedWhat the agent expects to happen, and why.
Forecast and Similar movesWhere there is history, what similar changes did before. No history is said plainly rather than filled in.
MeasuredFilled in once a change has run and its result is in.

Some rows carry a note instead of an Approve button. A row can say that a past rejection still applies to this action, who rejected it and when, and what lifts it. A row for an action you set to Never, or one that cannot succeed for this brand, says why and offers only Reject. That is deliberate: a button that would fail is worse than no button.

Chapter 3The Five-Minute Routine

  1. Open Pending once a day, at the same time

    The morning works for most teams, before the day's spend is committed. Use the From agents and From people chips if more than one person adds to the queue.

  2. Start with what expires first

    A row with a day left is a decision you are about to make by default. Decide those first.

  3. Clear the routine ones

    Pauses and small moves you would approve every time take seconds each. You can select several and approve them together; anything that cannot be undone is left out of a batch and has to be approved on its own.

  4. Open the ones that move money or cannot be undone

    Read Account, Change and Expected properly. For a change that cannot be undone, Review and send asks you to type SEND, which is the moment to be sure.

  5. Reject with a reason

    A reason travels with the rejection, and a rejection can keep the same action from going through again until whatever lifts it happens, which the row then says. A bare no tells the next reader nothing.

Who can approve

Approving needs Admin on the brand. Anyone else sees the queue as view only, which is useful for an analyst who reviews and a person who decides, but it means the queue stalls if the only Admin is away. Make sure two people can approve.

If the workspace is connected to Slack, a proposal posted there carries Approve and Reject buttons. They go through exactly the same checks as the buttons on the page, so a change refused on the page is refused in Slack too.

Chapter 4Approve, Reject, or Leave It

ButtonWhen it appearsWhat it does
ApproveA change that can be undoneSends the change. It runs through every limit on the Autonomy page first, and is refused with a reason if it breaks one.
Review and sendA change that cannot be undoneOpens the detail, where you type SEND to confirm.
Approve this onceA change outside what its goal was authorised to doOpens the detail to confirm this one change. It does not widen the goal.
RejectEvery rowDeclines it, with your reason if you give one.

Leaving a row is also a choice, and the queue is honest about it: the expiry is on the row. If you keep leaving the same kind of proposal, that is a signal about the plan, not about the queue. Either the agent is proposing something you do not want, which a rejection with a reason records, or it is proposing something you always approve, which is a candidate for running on its own. Choosing an autonomy level covers that decision, and setting spend limits for agents covers the limits every approved change is checked against.

Chapter 5Checking What Ran

Approving is half the job. The History tab shows what happened to each change after it was decided: the call that was made to the platform, the status that came back, the state before and after, and until when it can be undone. Where a change can be reversed, the row offers Undo.

Once a week, open the Activity page and its Weekly receipts tab for the wider view of what the agents did, and compare the Measured column on last week's changes with what was Expected. A change that keeps landing below its expectation is worth a rejection with a reason the next time it is proposed.

What to look for when a record says success and nothing happened is in the scans that reported success and stored nothing.

Why an unanswered queue is the weak point of asking first is in the delegation gap.

The daily five minutes

  • Open Pending at the same time every day.

  • Decide anything with a day or less left.

  • Batch-approve the routine, reversible changes.

  • Read Account, Change and Expected on anything that moves money.

  • Type SEND only after reading what cannot be undone.

  • Reject with a reason, every time.

  • Once a week, compare Measured with Expected in History.

Every brand starts at Ask first. The agents propose, you decide from one queue, and every decision and its result stays on the record.

Written by Shubham Raghav, Founder, Cresva. Questions? Email us.