Skip to main content

Weekly Plan

In plain English​

The Actions list shows you everything. That's honest, but a list of 30 things is a list most people avoid.

The Weekly Plan shows you just five. They're the top five from that same list — the five most valuable things you could do this week.

Five is small enough to hand out in a Monday meeting and actually finish by Friday. That's the whole idea.

Some of the five have a "one-click" tag. That means the app can do the job for you — you just press the button to approve it. The rest need a person.

How to use it​

  1. Click "Daily Ops" in the left-hand menu, then the "Weekly Plan" tab.
  2. Read the summary line at the top — it tells you how many jobs there are, how much money is involved, and how many the app can do for you.
  3. Go through the five cards. Give each one to a person on your team.
  4. Do the "one-click" ones right there in the meeting — they take seconds.
  5. Click "Mark done" as each gets finished.
  6. Come back next Monday. The list refreshes with the next five most valuable things.

⏱ ~5 min to plan a week · 💳 Pro+ · 🎯 A plan people finish, not a backlog they avoid

Why this matters for your business​

A 34-row worklist is honest and complete, and most teams will not work it. The psychology is well understood: a list long enough to feel infinite gets triaged into "later," and later never arrives. Teams don't need the exhaustive queue on Monday morning. They need to know the few things that, if done this week, will matter more than everything else they could have done instead.

That's the entire job of the Weekly Plan. It reads the same ranked stream as the Actions worklist and returns the top five — with the money at stake summed, and the ones the platform can execute for you flagged. Five is small enough to assign in a standup, finish by Thursday, and review on Friday.

The other half of the design is restraint. The Operator ships in recommend mode. It proposes the plan and sequences it; it does not execute it. Approving each action stays a human decision, and that is deliberate — a system that autonomously acts on a projected recoverable amount is a system that can autonomously be wrong at scale. When you do want hands-off execution, that lives in Automations, where it comes with dry-run mode and approval queues.

What this typically unlocks​

OutcomeTypical result
Weekly planning meeting length~5 min — read the plan, assign it
Plan completion rate~80% at five actions vs. ~25% against a full queue
Money addressed per weekThe highest-scoring slice of everything open
Arguments about priorityEliminated — the ranking is the tiebreak
One-click actions actually usedUp sharply — they're flagged, so nobody hand-does them
Team confidence the list is completeHigh — the full queue is one tab away

What you actually get​

The plan renders as a one-line summary plus five action cards.

The summary is composed deterministically from the plan itself:

5 actions to focus on this week · ≈USD 9,340 at stake · 2 one-click.

When there's nothing to do, it says so plainly rather than inventing filler work:

Nothing needs action right now — your operator will surface work
here as it appears.

Each of the five cards carries:

FieldWhat it's for
Title and detailThe finding, in plain language
DomainWhich part of the business it belongs to
SeverityCritical / Warning / Info
Amount at stakeRecoverable money, when the finding is economic
One-click badgeShown when the platform has a wired fix for it
Mark done / DismissThe same audited status actions as the worklist

Plan-level totals​

TotalMeaning
Action countAlways up to five
Projected impactSum of the recoverable amounts across the plan
One-click countHow many of the five the platform can execute
CurrencyFlagged when the plan spans multiple currencies — amounts are never converted or blended

How it works (without the technical bits)​

The plan uses exactly the same scoring as the worklist — (value at stake × confidence) ÷ effort — so the Weekly Plan is always the true top of the queue, never a separately-curated list that could disagree with it. A finding ranked first in Actions is ranked first in the plan.

Two details worth knowing:

  • The plan is computed live, not stored. Open it Wednesday after clearing three actions and you get the next three highest findings, not three empty slots. It is "the five most important things right now," not "the five things chosen on Monday."
  • Currencies never blend. If your findings span USD and EUR, the plan flags mixed currency rather than converting at a rate that would make the total quietly wrong.

One-click vs. manual​

An action is flagged one-click when the finding carries a wired fix — the platform knows how to execute it. Abandoned-cart recovery is the most common: the recovery send is already built, so the action is "approve the send," not "go build a campaign."

Flagging matters more than it sounds. Teams routinely hand-execute work the platform could do in a click, simply because nobody knew the button existed. Surfacing it on the plan card closes that gap.

Findings without a wired fix — restocking a SKU, renegotiating a supplier price, rewriting decayed content — are correctly marked manual. They still rank; they just cost more, and the scoring already accounts for that.

Real merchant scenarios​

Scenario A — Standup in five minutes​

Setup. Apparel brand, three-person growth team, previously spent 45 minutes each Monday arguing about priorities.

Monday's plan:

5 actions to focus on this week · ≈USD 12,180 at stake · 2 one-click.
#ActionAt stakeTypeOwner
1112 abandoned checkouts eligible for recovery$5,400One-clickApproved on the spot
234 products not sellable by AI agents$3,100ManualCatalog lead
3Margin breach on 8 SKUs after cost rise$2,240ManualBuyer
4Replenishment window open for 210 customers$1,440One-clickApproved on the spot
5Three top pages decayed below baseline—ManualContent lead

Result. Standup dropped to six minutes. Two actions executed during the meeting. The remaining three had named owners before anyone left the room.

Scenario B — The plan that refreshed mid-week​

Setup. Merchant cleared four of five actions by Wednesday and reopened the plan expecting one item left.

What they saw. Five again — the next four highest findings had moved up, including a new critical readiness gap detected Tuesday night after a supplier feed dropped image URLs on 60 products.

Why that's the right behavior. A plan frozen on Monday would have hidden a Tuesday critical until the following week. Because the plan is computed live, the most important five are always the current five.

Scenario C — Recommend mode earning its keep​

Setup. Home-goods merchant. Plan surfaced a $4,100 out-of-stock leak on a hero SKU with a recommendation to restock.

What the merchant knew that the system didn't. The SKU was being discontinued — the supplier had exited the category. The "recoverable" $4,100 was not recoverable at any price.

Action. Dismissed, with the reason noted. The recoverable total corrected immediately, and the next-highest finding took the slot.

The point. Under autonomous execution, this would have triggered a reorder of discontinued stock. Recommend mode meant a human with context made the call in four seconds.

Scenario D — Mixed currency, correctly refused​

Setup. Merchant selling through both a US and an EU store with findings in USD and EUR.

What the plan showed. The five actions with their own currencies intact, and a mixed-currency flag on the total rather than a single blended figure.

Why refusing to sum is the right answer. Converting at a snapshot rate produces a number that looks precise and is quietly wrong — and worse, one that changes when the rate moves even though nothing in the business did. Per-store plans, viewed separately, gave the team numbers they could act on.

Best practices​

✅ Run it as a Monday ritual. The value compounds when the plan drives the standup instead of supplementing it.

✅ Assign every action an owner in the meeting. Five is small enough that unowned items are conspicuous.

✅ Execute one-click actions during the meeting. They take seconds and remove the risk of them ageing out.

✅ Review Friday against Monday. Completion rate over a few weeks tells you whether five is the right number for your team's capacity.

✅ Dismiss with context when you know something the system doesn't. Discontinued lines, planned clearances, and known seasonal effects are exactly the cases where your judgment beats the detector.

✅ Keep the full worklist one click away. The plan builds trust precisely because it isn't hiding anything.

❌ Don't treat the five as a fixed weekly contract. It re-curates live — that's a feature, not drift.

❌ Don't hand-execute one-click actions. If it's badged, the platform does it faster and records it properly.

❌ Don't read projected impact as revenue you'll bank. Conservative recovery rates are already applied; it's a prioritization figure.

❌ Don't dismiss to clear the plan. Dismissal corrects the money rollup. Dismissing real work to reach zero makes the headline number lie to you next week.

Plan tiers​

CapabilityFreeStarterProAgencyEnterprise
Weekly Plan (Growth Operator)——✓✓✓
Projected impact total——✓✓✓
One-click badging——✓✓✓
Execute one-click from the plan——✓✓✓
Mixed-currency handling——✓✓✓
Multi-store plan roll-up———✓✓
Scheduled plan delivery to email/Slack———✓✓

Frequently asked​

Why five? Five is the largest number a small team reliably finishes in a week alongside their normal work. The full queue is always available in Actions — the plan is about completion rate, not coverage.

Can I change the number? The curated plan is fixed at five by design. If you want to work more, work the Actions worklist directly — it's the same ranking, unbounded.

Does the plan ever execute anything on its own? No. It proposes and sequences; every action is approved by a person. Autonomous execution lives in Automations, with dry-run mode and approval queues around it.

Why did an action disappear before I did it? It auto-resolved on its domain window, or the underlying fact changed — the cart converted, stock came back, the price returned inside guard.

Is the plan the same as the top five in Actions? Yes — identical scoring, identical source. The plan is that slice with plan-level totals and one-click badging added.

Who can see and approve it? Viewing requires growth:read:all; marking done or dismissing requires growth:manage:all. Solo merchants bypass both automatically.

What if my week's real priority isn't on the plan? Do your priority — the plan is advice, not a mandate. If your priority should have ranked and didn't, that usually means the owning engine is under-estimating its confidence or value; worth checking that engine's configuration.

See also​