Every number Roadmantica shows you comes from one of these. Nothing is hidden, and nothing is fudged.
You'll both see each other's answers.
Combined score
C_i = a_i + b_i
How you both rate one dimension, added together — 2 at the lowest, 10 at the highest.
The gap
gap_i = 10 − (a_i + b_i)
- tight
-
gap_i ≤ 2 - some gap
-
2 < gap_i ≤ 4 - wide gap
-
gap_i > 4
The room left on one question. A gap of 1 counts as tight.
Where your weights come from
W_i = (wins_i + 1) / Σ_j (wins_j + 1)
- After a full run
-
Σ_j wins_j = P → W_i = (wins_i + 1) / (P + n) - In the Classic set (n = 8)
-
n = 8, P = 28 → / 36
Every dimension starts with one point, so nothing is ever weightless — even one you never picked still counts.
Your side only
Fit% = round( Σ W_a,i · a_i / 5 × 100 )Equal% = round( Σ a_i / 5n × 100 )
- In the Classic set (n = 8)
-
n = 8 → / 40
What your own answers alone can say, before anyone else replies: your weights on your ratings, and the same ratings counted evenly.
Mutual fit
ω_i = (W_a,i + W_b,i) / 2Fit% = round( Σ ω_i (a_i + b_i) / 10 × 100 )
Both sets of priorities count equally — your weight and theirs are averaged before either rating is used.
The equal-weight comparison
Equal% = round( Σ (a_i + b_i) / 10n × 100 )
- In the Classic set (n = 8)
-
n = 8 → / 80
The same score with every question counted the same. The difference shown on your result is the gap between the two.
The two gap lenses
- As scored
-
length_a = a_i / 10 - As perceived
-
length_a = a_i · W_a,i / [ 5 (W_a,i + W_b,i) ] - The thicker side
-
min( 2.4, W_hi / W_lo ) × base
As scored shows the ratings as given. As perceived shows the same ratings weighted by how much each of you cares about them, and the thicker side belongs to whoever cares more.
Priority scores
base_i = wins_i / (n − 1) × 100
- The band a tie shares
-
[ S_g, S_g + (S_h − S_g) / 2 ] - Where each tied dimension lands
-
S_j = S_g + j / (k − 1) × (S_h − S_g) / 2, j = 0 … k−1j = 0 → S_gj = k−1 → S_g + (S_h − S_g) / 2 - When the tie is at the top
-
S_h = 100
Each win moves a priority up the 0–100 scale. Priorities that tie share the lower half of the space up to the next group, so a tie never takes a place it didn't win.
Rounding
round(x) = floor(x + 0.5)
We round once, at the very end, so small errors never add up.
Why your score can move when your partner re-ranks
Mutual fit averages both people's current priorities, so when they change what matters to them, the number you share changes too. Your answers did not move — the weighting did.
Why there's only one live percentage
An alignment always shows the score from the priorities in force right now, for both of you. Older rankings are kept as history, never as a second live number.
Survey statistics
Every average, share and place on a survey's results follows these rules, from the answers that count. No one person's answers are ever shown on their own.
Which answers count
Only each person's latest answer counts, and only if it took at least 10 seconds. A ranking you reuse from earlier counts too. If too many answers come from one network address, the extra ones are set aside as possible duplicates: they are kept, but left out of every figure. The survey's creator and our admin can see how many.
Average and median
x̄_i = Σ_r s_r,i / N
- The median, with the scores sorted from lowest to highest
-
N = 2m + 1 → s_(m+1)N = 2m → ( s_(m) + s_(m+1) ) / 2
In each answer, every priority has a score from 0 to 100. The average adds a priority's scores across the group's N answers and divides by N; the median is the middle score once they are sorted from lowest to highest, or halfway between the two middle ones.
Spread
σ_i = √( Σ_r ( s_r,i − x̄_i )² / (N − 1) )
How far a priority's scores sit from their average: the sample standard deviation, with N − 1 below the line. A wide spread means people disagree about that priority.
10th and 90th percentiles
h = (N − 1) · q + 1P_q = s_(⌊h⌋) + (h − ⌊h⌋) · ( s_(⌊h⌋+1) − s_(⌊h⌋) )
- The 10th and the 90th
-
q = 0.1 → P_10q = 0.9 → P_90
With the scores sorted from lowest to highest, the 10th percentile sits a tenth of the way along and the 90th nine tenths, read between the two nearest scores. Most scores fall between them — the box under “Scores people gave it”.
Average rank and the share at each place
ρ̄_i = Σ_r ρ_r,i / Nf_i(j) = 100 · #{ r : ρ_r,i = j } / N, j = 1 … n
Each answer puts every priority at one place, 1 for the top. The average rank is the mean of a priority's places; the share at a place is the percentage of answers that put it there — the bar under “Where people ranked it”.
Score steps
t_j = 100 · j / (n − 1), j = 0 … n−1
- The step a score s is counted at
-
j = ⌈ s · (n − 1) / 100 − ½ ⌉
With n priorities, scores fall on n evenly spaced steps from 0 to 100. Each score is counted at its nearest step (the lower one in a tie), and the dots under “Scores people gave it” show how many people are at each step.
The order of the ladder
Priorities are placed by their average, highest first. Equal averages are ordered by the larger share who put the priority first, then the higher median, then the set's own order. The order is decided on the exact figures, before anything is rounded, so two priorities showing the same average can still hold different places.
“1 in k”
k = ⌊ 100 / p + ½ ⌋
- “Almost”, when
-
100 / k − p > 1
“Comes first for 1 in 5” means about a fifth of people put it first. We add “almost” when slightly fewer did. When more than two in three put it first, we say “most people”, or “everyone” when it is within a point of all of them. “Last for 1 in m” works the same way, without “almost”.
Whole numbers
x → ⌊ x + ½ ⌋
Every figure is rounded to a whole number at the very end, after the order is decided. Each share is rounded on its own, so a bar may add up to 99 or 101.
The rule of ten
- A group of N_g answers, in a survey of N, is hidden when
-
N_g < 10N − N_g < 10
A survey shows nothing until 10 answers count. A filtered group is hidden if it has fewer than 10 answers, or if fewer than 10 answers fall outside it, because subtracting it from the total would give those few away. Any total that would reveal a hidden group is hidden too.
How sizes are written
10–19 · 20–49 · 50–99 · 100+
Under 10 answers, a survey shows its exact count (“6 of 10”), but a group shows none. From 10 to 99, counts are shown as a range; from 100, exactly — except in breakdowns, which say “100+” so that no hidden group can be worked out.
Comparing two groups
d_i = ⌊ x̄_A,i + ½ ⌋ − ⌊ x̄_B,i + ½ ⌋
- Two groups A and B can be compared when, for some attribute a
-
∃ a : V_A(a) ≠ ∅, V_B(a) ≠ ∅, V_A(a) ∩ V_B(a) = ∅
Two groups can only be compared if nobody can be in both, such as women and men, or under 35 and 35 and over. People who didn't answer that question are in neither group. The difference is between the two averages shown, and only when both groups are visible.
What stands out
- “What stands out”, on the exact averages
-
| x̄_A,i − x̄_B,i | ≥ 5 - A comparison's headline, on the whole-number averages shown
-
| d_i | ≥ 5
Up to four highlights, each shown only when every group it mentions is visible: the priority most often put first; the one people disagree on most; the biggest difference between women and men, from 5 points; and how people who skipped About you differ from those who answered it. Highlights use exact figures and comparisons use rounded ones, so a difference just under 5 can show as 5.