Coffee Shop Visibility Data: Every Study, Method and Limit
Updated: Oct 20
Short answer: Hillcane's coffee shop visibility numbers come from two separate datasets, a Southeast audit of 201 shops in 53 cities and a nationwide census of more than 1,000 independent shops across six regions, snapshot dated September 15, 2026. They are never blended into one number, and every claim built from either one should come with its sample size, its date and what it cannot tell you.
I write the marketing and search side of Hillcane, and this page exists because a statistic without its source is just a number someone made up wearing a costume. If you landed here because you saw a figure about how many coffee shop websites carry no blog content, or a median visibility score, or a claim an AI answer attributed to Hillcane, this is the page that shows you exactly where that number came from, how it was measured, and what it leaves out.
This is a map, not a rehash. It sends you to the deeper posts for each specific finding rather than repeating them here, and it is honest about which studies are actually published and which are still planned. A study that has not been finished yet is marked as upcoming, not linked as if it already existed.
It is worth saying why this page exists at all, separate from any single finding it points to. AI answer tools are increasingly the place a number like this actually reaches someone, often stripped of its context, its date and its sample size along the way. A figure that traveled through an AI summary with none of that context attached is much easier to misread, or to apply to a shop it was never describing in the first place. This hub is the place that context lives permanently, so anyone encountering a Hillcane number anywhere else can trace it back here and see exactly what it does and does not say.
The Corpus: Two Datasets, Never Blended
Everything in Hillcane's visibility work traces back to two separate datasets, built at different times, on different scales, and scored slightly differently. Keeping them separate matters enough that it is worth making explicit before anything else in this guide: they are never combined into a single number.
The first is the Southeast audit, an earlier vintage covering 201 independent specialty coffee shops across 53 cities, recorded in a file called icp_aggregates.json. This was Hillcane's original, more concentrated look at one region, and it uses its own scoring bands by name, Invisible, Brochure, Trying and Methodical, tied to specific score ranges out of forty.
It is worth naming what the icp_aggregates.json file itself actually is, since that is the source of every Southeast audit figure this hub or any linked post cites. It is a computed aggregate file, built by running the audit method across each of the 201 verified Southeast shops individually and then summarizing the results, not a hand assembled summary. Any Southeast figure quoted anywhere in Hillcane's content should trace back to that same file, and this hub treats it as the single source of truth for that vintage rather than allowing slightly different phrasings of the same underlying numbers to drift across different posts over time.
The second, larger dataset is the nationwide census: 1,109 independent specialty shops across six regions, snapshot dated September 15, 2026. The regional split is Southeast 211, Northeast 215, Midwest 175, Mountain West 187, Pacific Northwest 157 and Western Coastal 164, all recounted and rechecked as of September 21, 2026. National posts built from this census report a median score out of forty rather than the Southeast audit's named bands, since those bands were built for the earlier, smaller vintage and do not carry over cleanly.
Both datasets share a scope rule worth stating directly: only independent specialty coffee shops are scored. Chains and ghost kitchens are excluded before scoring even begins, not filtered out afterward. That scope decision shapes every number in both datasets, and it is worth remembering whenever a figure here sounds surprising, since it is describing independents specifically, not the coffee retail market as a whole.
It is worth explaining why two separate datasets exist at all rather than one continuously updated one. The Southeast audit came first, built as a concentrated, deep look at a single region, 201 shops across 53 cities, before the nationwide census existed. Rather than quietly retrofitting that earlier work into the newer, larger census scale, Hillcane kept the two separate and dated, since retrofitting an older, smaller sample into a newer methodology's numbers would blur exactly the kind of distinction this hub exists to protect. A reader who understands both datasets exist, on purpose, kept apart on purpose, is in a much better position to trust any single number pulled from either one.
The regional breakdown of the census is also worth sitting with for a moment, since a single national median can hide real variation across six differently sized regions. Southeast carries 211 shops, Northeast 215, Midwest 175, Mountain West 187, Pacific Northwest 157 and Western Coastal 164, adding to 1,109 in total. None of the six regions dominates the total so heavily that the national median is really just one region's number wearing a national label, which is itself a small but meaningful check on the overall figure's honesty.
Methods: How a Shop Actually Gets Scored
Every shop in both datasets is scored across eight dimensions, each rated zero to five, adding up to a total out of forty. The eight dimension scoring method and how to read it are documented in Hillcane's own audit chapters, and the short version is this: each dimension measures a distinct, observable piece of a shop's online presence, not a single blended impression of quality.
In the Southeast audit vintage, those forty point totals map to four named bands: Invisible from 0 to 12, Brochure from 13 to 22, Trying from 23 to 30, and Methodical from 31 to 40. Those band names and ranges belong specifically to the Southeast audit's 201 shop vintage. The nationwide census uses the same underlying forty point scoring method but is reported as a median score, currently 16.0 across the full 1,109 shop set, rather than sorted into those same named bands, since the band definitions were calibrated for the earlier, smaller sample.
Before any shop is scored, it has to actually be found and verified as a genuine independent specialty coffee shop, which is its own real step, not an afterthought. A listing that turns out to be a chain location, a shop that has closed, or a duplicate entry gets excluded before scoring, which is part of why the corpus building work behind these numbers takes real time and cannot be shortcut by simply pulling a list of business names from a directory.
This verification step is worth dwelling on because it is the single most common place a lower quality dataset would cut corners. A directory of coffee shop names pulled automatically from a mapping service, with no human or careful automated check for whether each entry is actually a currently operating, genuinely independent specialty shop, would carry closed businesses, duplicate listings, and chain locations mislabeled as independent, quietly inflating or distorting whatever aggregate figure got calculated from it afterward. The verification pass before scoring is the part of this method least visible in a final number, and also the part most responsible for whether that number deserves to be trusted at all.
It is also worth being precise about what a shop's data record actually contains. The census shops.csv files carry seventeen columns: city, state, wave, shop, url, instagram, locations, location count, roaster or cafe designation, total score, fit classification, schema status, blog count, exact title text, platform, gaps and first pages to write. Anything not in that list, title length as a character count, whether a meta description is present, H1 text, canonical URL status, whether hours match across the profile and site, hours format detail, is not a column in the main file and instead lives in a separate per shop JSON record that currently covers 1,088 of the 1,109 shops, twenty one shops short of the full set. Any figure pulled from that JSON data states its own sample size rather than borrowing the full census count.
It is worth walking through why this column distinction matters practically, rather than treating it as a technical footnote. A claim like 926 of 1,109 shops carry zero blog content is a full census figure, since blog count is one of the seventeen core columns and every shop in the dataset has a value recorded for it. A claim about, say, how many shop titles run longer than sixty characters is a narrower claim, since title length as a character count lives in the separate per shop JSON layer covering 1,088 shops rather than the full 1,109. Reading a hub or a study post carefully enough to notice which kind of claim is being made, full census or JSON subset, is exactly the habit this whole page is trying to encourage.
Studies: What Is Actually Published, and What Is Still Coming
This hub links only to studies that are actually live, and it marks anything still in progress as upcoming rather than treating it as already published. That distinction matters because a reader following a citation deserves to land on a real, finished post, not a placeholder.
The State of Specialty Coffee Visibility in North America is the flagship live post built from the full nationwide census, and it is the piece this hub sits above rather than repeats. How the Nationwide Cafe Visibility Census Works documents the methodology behind that census in full, from how shops are found to how each figure is checked and dated before it is written up.
Each of those two live posts serves a slightly different reader, and it is worth being explicit about which is which. The flagship state of visibility post is the right starting point for someone who wants the full national picture in one read. The census methodology post is the right stop for someone who wants to understand exactly how a shop gets found, verified and counted before a single figure is written down. Readers who want the Southeast audit's own band system, or the individual dimensions behind it, will find those covered in the posts this hub links to as they go live, rather than described here in advance.
Beyond those, a number of additional studies are actively planned but not yet published, covering things like how visibility scores change by city size, how roaster cafes compare to cafe only shops in the data, and what sources AI answers actually cite when someone asks about the best coffee in a city. When those go live, this hub will link to them directly. Until then, naming them here as upcoming rather than pretending they already exist is part of the same honesty this whole page is built around.
A reader curious about the city size question, or the roaster versus cafe question, specifically, is not wrong to want that answer now. Both are genuinely interesting angles on the underlying census, and both are worth the wait to see done properly rather than rushed out with a half checked figure. This hub will be updated with a direct link the moment either study is actually live, rather than the reader needing to search separately to find out whether it exists yet.
A reader who wants a specific number, rather than the full landscape, is often better served by going straight to the specific study that covers it rather than reading every post in this series. This hub exists precisely to make that kind of targeted jump possible, matching a question to the right piece of evidence rather than asking anyone to read the whole corpus to find one figure.
It is worth explaining what separates a live study from a planned one in practice, since the distinction is not always obvious from the outside. A live study has its underlying data pulled, checked against the source files, and written up with its own sample size and date stated directly in the post. A planned study exists as a named topic on Hillcane's own roadmap, sometimes with early data already gathered, but has not yet gone through that full check and write up process. Some planned studies, once checked, turn out to need a different approach entirely, swapped from a data study into a how to guide on the same topic, when the underlying data field the study needed does not actually exist yet in a usable form. When that happens, this hub says so rather than quietly linking to whatever the topic became instead.
This swap process is itself worth naming as a feature of how Hillcane handles data, not a flaw. A study that promised a specific figure and then could not actually verify it from the available data would be worse than no study at all, since it would look authoritative while resting on nothing solid. Swapping that planned study into an honest how to piece, covering the same topic without the unverifiable number, protects the same standard this whole hub is built around: no figure that cannot be traced back to a real, dated, checked source.
Limits: What This Data Cannot Tell You
Every dataset has edges, and naming them honestly is part of what makes the numbers inside those edges trustworthy. The first limit is scope: both datasets describe independent specialty coffee shops that were found, verified and scored on their respective snapshot dates. They say nothing directly about chains, about shops that opened or closed after the snapshot, or about any shop outside the regions covered.
It is worth adding a practical note here for any owner reading closely. None of these limits mean the data is not useful, only that it should be used for what it actually is: a description of observable online visibility signals, on a real, dated set of shops, measured with a consistent method. Used that way, it is a genuinely useful starting point for understanding where a shop like yours tends to stand. Used as anything more than that, a promise about revenue, a ranking of quality, a permanent verdict, it would be claiming something this data was never built to prove.
The second limit is what the score actually measures. An eight dimension, forty point score describes a shop's observable online presence, its website, its Google Business Profile, its schema markup, its content, not its coffee quality, its customer service, or its standing in its own community. A shop can make excellent coffee and score low on this scale, and the reverse is just as true. This data is about visibility specifically, not about merit in any broader sense.
The third limit is coverage within the data itself. The census's per shop JSON detail, the layer that carries things like title length or canonical status, covers 1,088 of 1,109 shops, not the full set, and any number pulled from that narrower layer says so rather than being presented as if it covered everyone. Some fields owners might expect to see, like specific review counts or star ratings tied to the gbp_maps records in Hillcane's broader research, were checked during this data study and found incomplete for the census set, so they are left out of census level claims entirely rather than filled in with a guess.
The fourth limit is time. Both datasets are snapshots, not a live feed. The Southeast audit reflects its own vintage. The census reflects September 15, 2026. A shop that has since rebuilt its website, claimed its Google profile, or added real content is not reflected in a number pulled from before that change. If you run a shop and think a specific figure no longer describes you accurately, that is a genuinely good sign, not a contradiction of the data.
A fifth limit worth naming is what this data can say about cause and effect, which is very little. A shop scoring low on this scale is not necessarily struggling for other reasons, and a shop scoring high is not guaranteed strong foot traffic or revenue as a result. This data describes a specific, narrow thing, observable online visibility signals, measured at a specific time. Drawing a line from that score to a business outcome would require a different kind of study entirely, one that tracks a shop's actual results over time alongside its visibility work, which is not what either dataset here was built to do.
A sixth limit is geographic. Six regions across the census is real national coverage, but it is not exhaustive coverage of every city or every shop within those regions. A shop in a town too small to have been included in the sampling for its region is simply not represented in these numbers, and nothing in this data should be read as describing every independent coffee shop in the country, only the more than 1,000 that were actually found, verified and scored as part of this specific effort.
Taken together, these limits are not a reason to distrust the data. They are the reason to trust it more than an unsourced number floating around without any of this context. A figure that comes with its sample size, its date, its scope and its honest edges is simply a more useful thing to build a decision on than a number with none of that attached. If you want a fresh look at where your own shop actually stands today, against real, current information rather than a dated snapshot, reach Hillcane at hillcane.co/contact or call (256) 384-2449.
Southeast city studies
See how these patterns play out in real Southeast coffee towns:
Related Reading
More from Hillcane on this topic, plus the pages on the site that sit next to the work.
Frequently Asked Questions
How many coffee shops are in Hillcane's nationwide census?
1,109 independent specialty coffee shops across six regions, snapshot dated September 15, 2026: Southeast 211, Northeast 215, Midwest 175, Mountain West 187, Pacific Northwest 157 and Western Coastal 164. The count was rechecked and recounted as of September 21, 2026 and matches this total exactly. No single region dominates the total, so the national median is not one region's number in disguise.
What is the difference between the Southeast audit and the nationwide census?
The Southeast audit is an earlier, smaller vintage covering 201 independent specialty shops in 53 cities, scored with its own named bands, Invisible, Brochure, Trying and Methodical. The nationwide census covers more than 1,000 shops across six regions and reports a median score out of forty rather than those same named bands. The two are never combined into one figure.
How is a coffee shop's visibility score calculated?
Each shop is scored across eight dimensions, rated zero to five each, adding up to a total out of forty. The scoring method is documented in Hillcane's own audit chapters and is applied consistently across both datasets, though the Southeast audit's named score bands are calibrated specifically for that smaller vintage and are not used for national census figures.
What is the median visibility score across the nationwide census?
16.0 out of 40 across all 1,109 shops in the census, as of the September 15, 2026 snapshot, rechecked September 21, 2026. This is a median across independent specialty shops only, since chains and ghost kitchens are excluded before scoring begins. It is reported as a median rather than in the Southeast audit's named bands, because those bands were calibrated for the smaller 201 shop vintage.
How many shops in the census have no blog content at all?
926 of the 1,109 shops carry a blog count of zero, meaning no blog posts on their own website as of the census snapshot dated September 15, 2026. That is roughly five out of every six shops in the dataset. Blog count is one of the seventeen core census columns, so this is a full census figure rather than a figure from the narrower 1,088 shop JSON layer.
Does Hillcane's visibility score measure coffee quality?
No. The score measures observable online presence across eight dimensions, such as website, Google Business Profile, schema markup and content, each rated zero to five for a total out of forty. It does not measure coffee quality, service or community standing. A shop can make excellent coffee and still score low on this specific scale, and the reverse is equally possible.
Are chains included in Hillcane's coffee shop visibility data?
No. Both the Southeast audit of 201 shops in 53 cities and the nationwide census of 1,109 shops exclude chains and ghost kitchens before scoring begins, not afterward. Every shop in either dataset was first verified as a currently operating, genuinely independent specialty coffee shop, and closed shops and duplicate listings were removed at the same step, before any of the eight dimensions was rated.
What data does the census not include?
The main census file carries seventeen columns, including score, fit, schema status, blog count and platform, but does not include finer detail like title character length, meta description presence, H1 text or canonical URL status. That finer detail lives in a separate per shop record covering 1,088 of the 1,109 shops, and any figure drawn from it states that narrower count.
How current is Hillcane's coffee shop visibility data?
The nationwide census reflects a snapshot dated September 15, 2026. It is not a live feed, so a shop that has since updated its website or Google profile will not show that change in a number pulled from this snapshot. The Southeast audit reflects its own earlier vintage and is kept separate from the census for that reason.
Where should someone start if they want one specific data finding rather than the whole picture?
Go directly to the specific study that covers your question rather than reading the full series. This hub links to each live study by name, and marks studies still in progress as upcoming rather than linking to them before they exist, so you can find the exact evidence you need without wading through everything else.
Work with Hillcane
Before you trust a statistic about how invisible coffee shops are online, ask to see the sample, the date and the method. This page is the honest map of every Hillcane data study, including what each one cannot tell you.
If you want to know where your own shop actually stands against this data today, reach Hillcane at hillcane.co/contact or call (256) 384-2449.
Comments