Data collection — primary vs secondary, tally charts, frequency tables: Visual walkthrough

This walkthrough uses the fruit survey from the parent topic. Prefer Hindi? That link has it too.

Step 1 — The raw stream: data before any organizing

WHAT. We start with the honest, unprocessed thing: a sequence of answers in the order people gave them. No counting yet — just a line of symbols.

WHY. You cannot organize what you have not first seen as it truly is. The raw stream is called primary data — information you collected firsthand by asking directly. It arrives one item at a time and in no particular order. Recognising this "one at a time, unordered" shape is the whole reason the tally trick exists.

PICTURE. Below, each box is one person's shout, left to right, in the order it happened. There is no pattern yet — Apple, Banana, Apple, Orange… just chaos in a row.

Figure — Data collection — primary vs secondary, tally charts, frequency tables

Step 2 — Sorting into buckets: what "category" means

WHAT. We decide the bins we will drop each answer into. Here the bins are the three distinct fruits: Apple, Banana, Orange.

WHY. Counting only makes sense once you know what you are counting into. A category is a bin — a group that every observation belongs to exactly one of. If a box could go in two bins, our count would double. So categories must be distinct (no overlap) and complete (every box has a home).

PICTURE. Watch the raw row from Step 1 get sorted: each box slides down into its matching bucket. Same boxes, now grouped by kind. Nothing was added or removed — only rearranged.

Notice the total number of boxes is unchanged. This "nothing lost, nothing gained" idea is our safety check later.

Step 3 — The tally mark: one stroke = one thing

WHAT. Instead of dragging heavy boxes around, we replace each box with a single vertical stroke |. One stroke means one observation.

WHY. Boxes are slow to draw while someone is still shouting answers at you. A stroke is instant — you can mark it the moment you hear the word. This is the point of a tally chart: real-time recording. You never have to remember; you just make a mark.

PICTURE. Each bucket's boxes become a row of strokes. Apple's bucket → eleven strokes. But look how hard eleven loose strokes are to count by eye — that difficulty is exactly the problem Step 4 solves.

Step 4 — The gate of five: crossing every fifth mark

WHAT. Every fifth stroke is drawn across the previous four, making a little "gate": four uprights with one diagonal bar through them.

WHY. Humans instantly recognise a group of five as one chunk (this is called subitising — seeing a small amount without counting). By forcing marks into gates of five, a long row of strokes turns into a short row of fives, which we can count in leaps: 5, 10, 15… This is the single most important convention on the page.

PICTURE. Apple's eleven strokes reorganise into two full gates of five plus one leftover stroke. Below, the two red gates are boxed and the lone leftover stands apart.

For Apple: $g = 2, r = 1 \Rightarrow 2 \times 5 + 1 = 11$ .

Step 5 — The frequency: turning gates into a single number

WHAT. We apply the formula to every row and write down one number per category. That number is the frequency — the total count for that category.

WHY. Gates are great for collecting, but bad for calculating. You cannot easily do arithmetic on drawings of gates. A number like $11$ can be added, compared, and turned into a percentage. So we translate each tally row into its frequency once, at the end.

PICTURE. Each fruit's gates collapse into a number. The arrows show the collapse: Apple's two-gates-plus-one becomes the clean numeral $11$ .

Fruit	Tally	$g \times 5 + r$	Frequency
Apple	2 gates + 1	$2 \times 5 + 1$	$11$
Banana	1 gate + 3	$1 \times 5 + 3$	$8$
Orange	0 gates + 4	$0 \times 5 + 4$	$4$

Step 6 — The safety check: the total must return

WHAT. Add every frequency. The sum is written $\sum f$ (read "the sum of all the f's").

WHY. Remember Step 2: sorting adds and removes nothing. So the frequencies must add back up to the number of people we asked. If they don't, a box got lost or double-counted, and we recount. This is your built-in error alarm.

$\sum f = Apple 11 + Banana 8 + Orange 4 = 23$

Here $\sum$ (Greek capital sigma) just means "add all of these up", and $f$ stands for each frequency. Since $23$ equals the 23 classmates we surveyed — the check passes.

PICTURE. The three number-blocks stack back into one bar of length 23, exactly matching the original raw row from Step 1. The picture literally closes the loop.

Step 7 — Edge and degenerate cases

Real surveys throw curveballs. Each one still fits our machine — here is how.

Case A — a category with frequency $0$ . Suppose nobody chose Mango. Its row is empty: $g = 0, r = 0$ , so $0 \times 5 + 0 = 0$ . We still list it with frequency $0$ , because "zero Mangoes" is information, not absence of information.

Case B — exactly a multiple of five. If Banana had been $10$ , that is $2$ full gates and $r = 0$ leftovers: $2 \times 5 + 0 = 10$ . A leftover of zero is perfectly normal — it just means the last gate closed exactly.

Case C — a single lonely observation. One answer only: $g = 0, r = 1 \Rightarrow 1$ . No gate ever forms. The gate trick only kicks in from the fifth mark onward; below five it is just plain strokes, and that is fine.

PICTURE. Three tiny tally rows side by side: the empty row ( $= 0$ ), the exact double-gate ( $= 10$ ), and the single stroke ( $= 1$ ).

Recall Quick self-test on the edge cases

A category has 3 full gates and 0 leftovers — what is its frequency? ::: $3 \times 5 + 0 = 15$ Nobody chose "Grape". Do you delete that row? ::: No — you list Grape with frequency $0$ ; a zero is real information. Can the leftover count $r$ ever be $5$ ? ::: No — the fifth stroke becomes a gate, so $r$ is always $0, 1, 2, 3,$ or $4$ .

The one-picture summary

This single figure compresses all seven steps: raw stream → sorted buckets → strokes → gates of five → frequencies → total check, with the edge cases hanging off the side.

Recall Feynman retelling — say it like you'd explain it to a friend

"People shout fruits at me one at a time — that's my raw primary data, a messy row (Step 1). I decide three buckets, Apple/Banana/Orange, and drop each shout into the matching bucket without losing any (Step 2). To record fast I don't drag boxes — I just scratch a line for each shout (Step 3). But eleven scratches are impossible to eyeball, so every fifth scratch I draw a bar across the last four, making little gates of five that I can count in leaps (Step 4). At the end each pile of gates becomes one number — its frequency — because numbers do arithmetic and drawings don't (Step 5). Finally I add all the frequencies: they have to come back to 23 people, because sorting never adds or loses anyone; if they don't, I recount (Step 6). Weird cases still fit: an empty bucket is a real zero, exactly ten is two clean gates, and one lonely answer is just one scratch (Step 7). That's the whole machine — from chaos to a table that checks itself."

Where this goes next: once you have this frequency table you can draw it as a bar chart or pictogram, find its mode, bundle scores into class intervals, or turn counts into experimental probabilities.