Location scouting workbench for picking your next site

Build me a location scouting workbench I open when we are deciding where to open our next site. It has three screens: Scout, Shortlist comparison, and Settings.

The Scout screen is where I start. I type a candidate address, optionally adjust the search radius and the cannibalization drive time threshold from their configured defaults, and hit Scout. The handler pins the address with the Geolocation Forward Geocode operation, then calls Search Nearby Places twice around those coordinates: once for the business types configured as competitors, and once for the anchor types that generate foot traffic such as grocery stores, gyms, coffee shops and big box retail. Take the most relevant results from each search, capped at about ten competitors and ten anchors, and call Get Place Details on those to fill in rating, review count, opening hours and formatted address.

In the same handler run, read the locations we already operate from a Google Sheets tab with Get Values, where each row has a site name and address. Then call Calculate Distance Matrix with the candidate as the origin and those existing sites as destinations to get driving time and distance to each one. Cap the destination list at roughly twenty five sites, nearest by straight line first, because Distance Matrix bills per element (origins times destinations) and an unbounded list gets expensive fast for a chain with a lot of locations.

The result renders as a scorecard for that candidate: how many competitors sit inside the radius and how that compares to the shortlist average, the average competitor rating and total review volume I would be up against, a list of nearby anchors with their distance and rating, and the closest existing location by drive time. When that closest existing location is inside the cannibalization threshold, show a prominent warning naming the site and the drive time. Show the underlying competitor and anchor lists in expandable tables so I can look at the actual businesses, not just the summary numbers.

Saving a candidate appends it to the shortlist tab with Google Sheets Append Values. Write both the computed scores and the fetched place results into that row, including a column holding the raw competitor and anchor payload, so the app can rebuild the full scorecard from the sheet alone. This matters for cost: Nearby Search runs about four cents per request and Place Details about two cents each, so place data must be fetched exactly once per candidate at scout time and never re-queried on a view render. Add an explicit Rescout button on a saved candidate for when I deliberately want fresh data, and show when the cached data was captured.

The Shortlist comparison screen reads the shortlist tab with Get Values and puts every saved site side by side in one table: address, competitor count, average competitor rating, anchor count, drive time to the nearest existing location, cannibalization flag, and the agent recommendation once it exists. Make the columns sortable, highlight the best and worst value in each numeric column, and let me click a row to open its saved scorecard. Nothing on this screen may trigger a new maps call, it reads entirely from the sheet.

Each shortlisted site gets a "Deep scout this site" button that kicks off a background agent. The agent takes that row, researches the surrounding trade area in depth (it can use Search Places by Text and Get Place Details for additional context such as traffic generators, complementary categories and the quality of the local competition, plus general web research on the neighbourhood), reads the rest of the shortlist from the sheet so it can weigh this site against the alternatives rather than judging it in isolation, and produces a written recommendation with its reasoning and a clear verdict of pursue, hold or pass. It writes the recommendation, reasoning and verdict back into that shortlist row with Google Sheets Update Values, and posts the brief to our expansion channel with Slack Bot Send a Message, leading with the address and verdict so the channel can skim it.

The app should show a pending state on the row while the agent is running, and surface the verdict and recommendation in both the scorecard and the comparison table once it lands. The Settings screen holds the spreadsheet id, the existing locations tab and shortlist tab names, the competitor and anchor business type lists, the default search radius, the cannibalization drive time threshold, and the Slack channel for briefs.