Same pattern: what rises and falls with cafés across London
Across London, all restaurants, shopping and beauty & spa places follow the pattern of cafés very closely, and the owner-occupier share goes the other way. Here is the card that says so - Same pattern - and the rank correlation behind its words.
Across London, three measures follow the pattern of cafés within a 10-minute walk very closely: all restaurants, shopping, and beauty & spa places, each within a 10-minute walk too. Where one is high, the others are high; where one is low, so are they. One measure goes the other way: the owner-occupier share, which tends to be low where cafés are many. The card that says so is Same pattern, and it opens from any factor of a Score.
It compares the order of London's hexes - hexagons about 350 m across - on two measures at once: rank every hex on each, and see how well the two rankings line up. The field calls that Spearman's rank correlation. The app never shows the number; it says it in three words.
Same patternthe field's term: Spearman's rank correlation
Same pattern as Café in London. Three measures follow cafés very closely, each read within a 10-minute walk, and one goes the other way. The + adds a measure to your Score.
The idea, in one sentence
Put every hex in order on one measure, then on the other: the more alike the two orders, the more closely the measures follow one pattern. The answer comes in three strengths, plus one line for the opposite:
very closely
The two orders almost match. Three filled bars on the card.
closely
Clearly alike, with more hexes out of step. Two filled bars.
somewhat
A real but loose likeness. One filled bar. Anything weaker is not listed at all.
Goes the other way
One line under the list: the measure that falls most where this one rises - shown only when it does so at least somewhat.
Where to click
Open a Score. In its Selected settings panel, above the factor cards, sits the importance bar: one colored piece per factor, as wide as that factor's share of the Score. Tap a piece, and the card for that factor opens.
Where to click
Map type
Score
Selected settings
importance bar
tap the Café piece
Every row on the card does something:
The + adds that measure to your Score as a new factor, read the way it was compared - here, within a 10-minute walk. A measure already on your Score says on your Score instead.
Tapping a row's name opens Cross two with the factor and that measure, so you can see the two patterns side by side, hex by hex.
On a factor that adds several kinds together, a last section, Inside this factor, says which of its own kinds follow one pattern.
When two factors of one Score follow one pattern very closely, the importance bar says so by itself: a small bracket appears under their two pieces, labeled same pattern ›. Tap it and the card names the pair, with See them together › to open them on Cross two.
Two more places it shows up
While you add factors. Press Add measures on a Score and, above the usual list of measures, a card titled Same pattern as your Score in London suggests up to three measures that rise and fall with a factor you already have and are not on your Score yet. Each row says which factor it follows.
The suggestions for the series' Score. All three follow Total population very closely - including Female, which is no discovery: where more people live, more women live.
That last row is worth a second look. Same pattern states facts, not discoveries, so the obvious ones are listed too. A measure worked out from another - population density from population - can follow it very closely for that reason alone.
On Cross two. Open the second-measure picker of a Cross two map, and the top of the list reads Same pattern as … your first measure, with up to three rows and the one that goes the other way. Picking a row draws it as the second measure.
Where to click
Map type
Cross two
second measure
Same pattern as …
The math, honestly
Rank correlation needs only the order of the hexes, never their values. The five hexes below are illustrative - the shape of the sum is exactly what the app does.
Five hexes: cafés and restaurants, each ranked from fewest to most (illustrative)
Hex A · 0 cafés, 1 restaurantdifference 0
rank 1 · rank 1
Hex B · 2 cafés, 4 restaurantsdifference 1
rank 2 · rank 3
Hex C · 5 cafés, 3 restaurantsdifference 1
rank 3 · rank 2
Hex D · 9 cafés, 12 restaurantsdifference 0
rank 4 · rank 4
Hex E · 30 cafés, 25 restaurantsdifference 0
rank 5 · rank 5
1 − 6 × (0 + 1 + 1 + 0 + 0) ÷ (5 × 24)
0.9 → very closely
The shortcut formula above works when no two hexes tie. The app gives tied hexes their average rank and correlates the two columns of ranks, over every hex that could hold a place and has a value for both - thousands of hexes in London, not five.
The words come from fixed cut-offs, the same for every city and every measure:
0.8 or more - very closely (and the bracket, between two factors)
0.6 to 0.8 - closely
0.4 to 0.6 - somewhat
−0.4 or less - goes the other way
Why ranks rather than the values themselves:
Units stop mattering. A count and a share, a total and a rate are compared on where they are high and where they are low.
One extreme hex cannot decide it. A handful of central hexes with dozens of cafés would sway a correlation of raw values; ranked, they are simply the top few.
Each list takes one measure per topic - one category of place, one census topic - so it is not five kinds of restaurant in a row. A measure with a value in fewer than 30 hexes gets no list, only one sentence: “Only [N] hexes in [city] have a value for [measure], so what follows the same pattern can't be told apart from chance.”
A fact about where things are - not a cause, not advice
Same pattern says which measures are high in the same hexes. It does not say why, and it does not say one causes the other: cafés, restaurants and shops all gather where people walk and work. Nor does it tell you what to do with your Score - it never suggests removing or combining a factor. The places are those Overture Maps lists (release 2026-08-19); the people measures come from ONS Census 2021 and are an estimate for each hex. Treat it as decision support, not a forecast.
Try it on the series' Score
The link opens the Score this series uses throughout: cafés and transit stops within a 10-minute walk, plus the residents of each hex. Tap the Café piece of the importance bar, then the Total population piece, and compare the two cards. For what each factor does to the Score, see how the Score works; for what stands near a kind of place rather than what rises with it, see Usually nearby.
Same pattern ranks every hex on two measures and checks how well the two orders line up - Spearman's rank correlation.
The number is never shown: very closely (0.8 or more), closely (0.6), somewhat (0.4), and Goes the other way (−0.4 or less).
Tap a piece of a Score's importance bar to open the card for that factor; a same pattern › bracket joins two factors that follow one pattern very closely.
+ adds a listed measure to your Score; tapping its name opens the pair on Cross two.
It also suggests measures under Add measures and tops Cross two's second-measure picker.
It states where things are high and low together - never a cause, never advice.
Quick questions
What does Same pattern show?
For a factor of your Score, the measures that are high in the same London hexes and low in the same hexes, said as very closely, closely or somewhat, plus the one measure that goes the other way.
How is it worked out?
Every hex is ranked on both measures and the two rankings are correlated - Spearman's rank correlation. 0.8 or more reads very closely, 0.6 closely, 0.4 somewhat and −0.4 or less goes the other way; the number itself is never shown.
Where do I find it?
Tap a piece of a Score's importance bar. It also suggests measures under Add measures, tops the second-measure picker of Cross two, and draws a same pattern › bracket under two factors that follow one pattern very closely.
Why ranks instead of the values?
Ranks compare a count with a share, or a total with a rate, on where each is high and low, and one extreme hex cannot decide the result on its own.
Does Same pattern mean one thing causes the other?
No. It says which measures are high in the same hexes, never why, and it never advises removing or combining a factor. Treat it as decision support, not a forecast.
See it on a real map
The best way to understand the score is to play with it. The full app is on a Free Tier, with no sales call.
Builds MapBees's location-intelligence pipeline end to end — census and Overture Maps ingestion, H3 hex scoring, and the map interface — and writes the methodology behind every number on the site.