Seven panels. One of them tells you what to do.

Do this first

One action, at the top, on the dark band, with the minutes it will take and the reason it won. It is the only thing on the screen competing for your attention, because an owner with twenty minutes does not need a shortlist of nine.

• Ranked by what it is costing you, not by how new it is
• Carries its own evidence, so tap and read the sentences
• Mark it done and it moves to Implemented, with a date
• Fired by rules, never by a model's mood

The week's numbers, each one a door

Your rating against the street, your reply rate, the pace of new reviews, and the dish you are known for. Four figures rather than forty, and every one of them opens into the sentences it was computed from.

• Rating shown on the same rail as your four rivals
• A number the sweep could not reach is absent, never zero
• Every bar carries its own figure, for reading outdoors
• A date range you set, not a window we chose for you

The four doors nearest you

You name the rivals. We read them with exactly the same sweep, in the same week, and put you on one shared axis with them: rating, pace of new reviews, the dish they are known for, and what changed on their listing.

• Add or drop a rival whenever the street changes
• Their Instagram read in your own timezone
• A city average is never offered as a substitute
• Orange is you, on every chart, and nothing else is

Dish intelligence, held to your real menu

What guests name, how often, and whether they name it warmly. A dish only appears here if your actual menu backs it. The fastest way to make an owner distrust a tool is to report enthusiasm for something they do not sell.

• Mentions counted across Google, Zomato and Swiggy
• Matched against the menu we can actually see on your listings
• The rival's signature dish shown beside yours
• Named dishes open into the sentences, like everything else

Drafted replies you edit and send

Each unanswered review arrives with a draft in your own register, naming the thing the guest actually named. Nothing posts by itself and nothing is sent on your behalf. You edit it, you copy it, you decide.

• Never auto-posted, on any platform, under any setting
• Answers the complaint, not a template's idea of it
• Connect Google and you can send from here; otherwise copy out
• The AI writes the sentence; it never produces the figure

What changed since you did the last thing

Mark an action done and it moves to Implemented with the date. Next Monday the console reports what moved since, never because. Nothing here will claim credit for your rating, because nothing here can prove it.

• A dated log of what you did, kept for you
• The figures either side of that date, plainly
• No causal arrow, no attribution model, no confetti

What we could not read

The last panel is a list of things the sweep could not reach this week, and why. It is the panel that makes the other six worth believing, and the one no tool in this category ships.

• Named source by source, with the reason
• A missing figure is missing, not estimated and not blurred
• If a whole source went dark, the week says so at the top

The full list of what a public sweep cannot see →