The Observatory’s method

The HelloMarkus Observatory of local business visibility is a public measure, city by city, of how local businesses are found on Google and Google Maps: their profiles, their reviews, and who ranks at the top when a customer searches.

The October 2026 edition measures one trade, takeaways (quick-service food), in 20 French cities: Angers, Annecy, Bordeaux, Caen, Clermont-Ferrand, Dijon, Grenoble, Le Mans, Lille, Lyon, Marseille, Montpellier, Nantes, Nice, Paris, Perpignan, Rennes, Rouen, Strasbourg, Toulouse. Barbers will follow the next quarter.

The Observatory is signed HelloMarkus. The figures name no business and rank nobody: they say where the businesses of a city stand, in totals, shares and medians.

• The public Google profiles of the businesses and the Google Maps results, collected by DataForSEO, a data provider (its Business Listings, Google Maps and Google Reviews services). For the French barometer, HelloMarkus does not use the Google Places API and stores no profile.
• Google Ads search volumes (monthly average over twelve months, France, in French), also collected by DataForSEO: what customers type together with the name of the city.
• INSEE municipal populations (the French statistics office), read from the Découpage administratif API (geo.api.gouv.fr), to set the number of profiles against the population.

The measurement script is versioned with the website: the same command redoes the measure, with the same random draw.

• Every measure carries the day it was collected: the October 2026 edition was collected on 7 October 2026, and every published figure is signed “HelloMarkus Observatory, October 2026”.
• Search volumes carry their period: the monthly average from September 2025 to August 2026.
• A full measure per city and per trade every quarter (January, April, July, October). Next measure: January 2027. Past editions remain available.

A profile in the trade

A Google profile whose primary category is on the trade’s list, and whose address is in the municipality. For takeaways, the Google categories: fast food restaurant, snack bar, kebab shop, doner kebab restaurant, sandwich shop, hamburger restaurant, taco restaurant, fried chicken takeaway. A secondary category isn’t enough: a bakery that adds “snack bar” to its categories is not counted.

The average rating

The average of the ratings of profiles with at least one review. A profile with no reviews has no rating and is left out of the average; it counts in the share “no reviews at all”.

An incomplete profile

A profile missing at least two things out of five: opening hours, a phone number, a website, at least 10 photos, a description.

An unclaimed profile

A profile where Google still shows “Own this business?”: nobody has taken it in hand.

Unanswered reviews

The 20 most recent reviews of 60 profiles drawn at random among those with at least 5 reviews (all of them, if there are fewer than 60). The draw uses a fixed seed: it can be redone identically. A review is unanswered when the owner hasn’t replied to it. A random draw, because big chains reply and independents don’t: taking the most-followed profiles would give a false figure.

The businesses at the top

The first three Google Maps results, ads removed, for the word customers type (“fast food” for takeaways), searched on a phone at the centre of the municipality and from four points 1.5 km away (1 km under 200,000 residents). We take them as Google shows them, even when their category is not the trade’s. For them: the median number of reviews, photos and rating, the share with a website, and the share of their recent reviews left unanswered.

The median

The middle value: half of the profiles above, half below. It holds up against the few profiles with thousands of reviews.

Districts and postcodes

Profiles grouped by the postcode of their address (the arrondissements in Paris, Lyon and Marseille).

The other cities

A city is compared with the median of the 20 cities measured and with their range (from the lowest to the highest), never by a ranking.

Merged municipalities

In Annecy, profiles whose address still carries the name of a former municipality merged into it (Annecy-le-Vieux, Seynod, Cran-Gevrier, Meythet, Pringy) are counted in the municipality.

• 50 profiles in the municipality to publish a city: below that, no page.
• 20 profiles to publish a district or a postcode.
• 300 reviews read to publish a share of unanswered reviews, including for the businesses at the top and for negative reviews.

Below a threshold, the page says “not enough profiles” or leaves the figure out. It never estimates it.

• The category is chosen by the ownerA takeaway filed as a “restaurant” is not counted; a bakery’s sandwich counter can be.
• Google can be wrong or out of dateA closed business still online, a misfiled address, a wrong postcode: we count what Google shows.
• The Google Maps ranking movesWith the time of day, the phone, the person searching. “At the top” is a snapshot taken on one day, from five points.
• A sample gives an order of magnitude60 profiles drawn at random: the share of unanswered reviews is accurate to about ten points, not to the unit.
• Search volumes are roundedGoogle Ads rounds in steps and groups close variants together; “near me” searches are not included.

• Name a business, for better or worse. A business owner is only quoted with their written consent, and never negatively.
• Copy a review or its author’s name, publish an individual rating, an address or a phone number.
• Keep profile and review details for more than 30 days: they serve the calculation, away from the website, then they are deleted. Only the aggregates remain.
• Draw up a league table, a “worst city” or a ranking of businesses.
• Estimate a figure that has not been measured.

The Observatory’s figures are published under the Creative Commons Attribution 4.0 (CC BY 4.0) licence: you can reuse, quote and adapt them, including in an article or for commercial use, citing “HelloMarkus Observatory, October 2026” with a link to this page.

The aggregates file has one row per city and one per publishable district or postcode: 20 cities, 9,227 profiles collected, 22,635 reviews read. Aggregates only: no names, no addresses, no reviews. The column names are in French: “ville” is the city, “fiches” the number of profiles, “note_moyenne” the average rating, “part_avis_sans_reponse” the share of unanswered reviews.

A figure looks wrong, a profile miscounted? Write to us, in English if you like: we reply within 7 days, and the correction is dated on the page concerned.

• October 2026: first measure. Takeaways in 20 cities, collected on 7 October 2026.

AI assistants, later: measuring what assistants such as ChatGPT, Perplexity or Gemini answer to “meilleur snack à Lyon” (best takeaway in Lyon) will be added once its method is written and published here, not before.

The same measure, outside France: London first for the United Kingdom. Same indicators, same thresholds, same rules, one city before any other.

• The planned source: the Google Places API, Google’s official interface for business profiles. Each figure will carry its source and its date.
• What we keep: the place_id of each profile (the identifier Google allows to be stored) and the aggregates computed from the profiles: totals, shares, medians.
• What we never keep: names, addresses, phone numbers, reviews or individual ratings. The detail serves the calculation, then it is discarded.
• Nothing is published before written confirmation that publishing figures computed from Google Places data fits the Google Maps Platform terms for this use.

Replies to reviews and claimed profiles cannot be read through the Places API: the pilot will say how it measures them, or that they are missing. See the London pilot.