Automated competitor analysis around every site, every week
Automated competitor analysis counts rivals, their ratings and new branches around each of your sites on a timetable, then writes the brief. What to measure and how.
Part of the guide AI employee for site selection
Automated competitor analysis replaces the one-off study (a consultant's deck that was right the month it was written) with a flow that measures the same few things around each of your sites every week: how many rivals are within reach, how they are rated, how many belong to chains, which branches are new and which ratings moved. It compares each week with the last and writes a short brief on what changed. You stop asking "what is the competition like here?" once a year and start hearing "two new rivals opened near your Camden site" on Monday.
This is for operators with several sites (gyms, cafes, clinics, salons, pharmacies, hotels) and for the analysts and agencies who report to them. It covers what is worth measuring, how a weekly run differs from a one-off analysis, a made-up example of a gym operator in London, the traps in reading the numbers, and the blocks that do the work.
Why a running analysis beats a one-off study
| One-off study | Weekly flow | |
|---|---|---|
| Answers | What the market looked like then | What changed since last week |
| New rivals | Missed until the next study | Reported the week they appear |
| Ratings | A snapshot | A trend, and the week it moved |
| Effort | Days of searching and copying | Set up once, read in a few minutes |
| Comparable | Rarely, if the method changed | Always: same radius, same types, same sites |
The value is in the comparison: the same measure, taken the same way, over time.
A one-off study still has its place before an entry decision. After you open, the questions are about change, and only a repeated measurement answers them.
What to measure around each site
Nearby places, run around each of your sites with counts, adds these to every site each week:
- Rivals within reach (nearby_count): the places of your types within the distance you chose.
- Nearest rival and its distance (nearby_nearest, nearby_nearest_m).
- Their average rating (nearby_avg_rating): weaker rivals leave room; strong ones set the bar.
- How many are chains (nearby_chains): a chain arriving usually means a bigger fight than an independent.
- Each type on its own, when you search more than one (nearby_type_count).
Two more blocks add the change: Only new, after a second Nearby places search that passes on the rivals themselves, lets through the ones not seen before; What changed lets through rivals whose rating moved or that closed.
Those London figures show why a citywide count says little to a single club: what matters is the handful of gyms within a short walk of each site, and how that handful changes.
A made-up example: six gyms in London
This example is made up. A boutique gym operator has six clubs in London. Its marketing lead wants a Monday brief per club and one for the board.
- The six clubs sit in a table in My data, with each club manager's email.
- Nearby places counts gyms, yoga studios and pilates studios within 800 m of each club.
- A second Nearby places search lists those rivals; Only new keeps the ones that appeared since last week; What changed keeps those whose rating moved by 0.2 or that closed.
- An xMap LLM block, working on the whole list, writes the brief from those fields alone, told never to add a number.
- Make a report turns it into a one-page PDF for the board; each manager gets their own club's lines by email.
Made-up brief: "Clapham: a new pilates studio opened 300 m away; rivals within 800 m rose from 9 to 10; their average rating held at 4.5. Shoreditch: the nearest rival, 150 m away, now shows as closed."
Traps in reading the numbers
- Changing the method. Widening the radius from 800 m to 1 km will add rivals that were always there. Keep the distance and types fixed, or note the change.
- Mistaking a new listing for a new opening. A place new to the data may have opened years ago and only now been listed. Where the database has an opening date, check it before you call it new.
- Reading small moves in averages. An average rating over five rivals moves when one of them gets a bad week. Look at who moved, not only the average.
- Watching only one kind of rival. A gym's customers also leave for studios, climbing walls and swimming pools. Count the types your customers actually switch to.
- Alerting on everything. A weekly brief that lists every unchanged number trains people to skip it. Report change; keep the steady figures for the monthly review.
The AI employee that does it
Set up once, this becomes a standing job, the kind location intelligence automation is about:
- Schedule: every Monday at 07:00.
- My data: your sites.
- Nearby places: rivals within your distance, counted.
- A second Nearby places listing the rivals, then Only new and What changed.
- xMap LLM: the brief, from the figures only.
- Make a report and Email.
On Monday the inbox holds a short brief per site and a PDF for the leadership team: new rivals with their distance, ratings that moved, closures. If you already run several AI employees, Team findings can read what they all found and feed one weekly brief. The templates weekly competitive intelligence for my cafes, competitor density around each site and a weekly brief from your AI employees are versions of this flow.
Questions
How often should competitor analysis run?
Weekly suits most multi-site businesses: openings and closures are rare enough that a daily run is mostly empty, and frequent enough that a quarterly one misses the moment to respond.
What distance should I use around each site?
The distance your customers travel: a short walk for a cafe or a city gym, a few kilometres for a suburban store. Keep it fixed so weeks compare.
Can it compare my rating with my rivals'?
Yes. Look up your own sites with Map business scan, Each place itself, and put their rating beside the rivals' average in the brief.
Will the brief invent figures?
Tell the xMap LLM block to use only the figures given and never add a number; the counts come from the blocks before it, not from the AI.
Templates that do this
Read next
More on site selection and expansion · Every article · Templates