Customer cohorts
Customer cohorts answers one question: of the people who bought from you for the first time, how many came back, how soon, and how much more did they spend?
A cohort is a group of buyers whose first order with you fell in the same month. Following a group forward shows whether your new buyers turn into regulars.
Open My Shop → Analytics → Customers. Customer cohorts is the first tab. On this tab the date picker selects first-order dates: pick last March and you see everyone whose first order was in March, and everything they bought after it.
The three cards at the top
The cards cover the buyers whose first order falls in your dates, over a set number of days after that first order. The window starts at 30 days. To change it, scroll down to the table and click any percentage under 60 days, 90 days, 180 days or 365 days. The cards and charts switch to that window.
| Card | What it tells you | Example |
|---|---|---|
| Bought again | How many of these buyers came back for another order within the window. | 12%, with 180 / 1,500 buyers underneath: 180 of 1,500 buyers came back within 30 days. |
| GMV per buyer | How much a buyer spent in the window, first order included. | ฿420: the average buyer spent ฿420 in their first 30 days. |
| Value vs first purchase | How much later orders added on top of the first one. | 1.30×: for every ฿100 of first orders, buyers spent another ฿30 later. 1.00× means nobody spent anything after their first order. |
The small badge compares with the period just before your dates, of the same length. For example, if you picked 1–31 March, it compares with buyers who first ordered in the 31 days before that. On Bought again, the badge is in percentage points: +2 pts means the rate went from, say, 10% to 12%.
Hover or click the small line on a card to open a bigger chart of the same number for each month.
The three charts
All three show the same buyers and the same window as the cards.
Share of buyers (left)
Buyers grouped by how many purchases they made in the window. For example, bars of 1 = 85%, 2 = 10% and 3 = 3% mean 85 in 100 buyers bought only once. A tall first bar is normal. If the bars after it shrink slowly, you have more regulars. If they drop away fast, most people buy once and don't come back.
Time until next purchase (middle)
This answers when buyers come back for their second purchase. The higher the line on a given day, the more buyers make their second purchase around that day.
The line always starts high and slopes down. A buyer is most likely to come back soon after their first order, and less likely as weeks go by.
- A steep drop means buyers who come back do so quickly.
- A gentle slope means returns are spread out over months.
- The shaded band around the line shows how sure the estimate is. A wide band means few buyers have come back yet, so treat the shape as rough.
- The grey area is where there is real data. Past the grey, the line is a best guess.
Chance of one more purchase (right)
Each bar answers: of the buyers who got this far, how many made one more purchase?
For example, 1→2 = 12%, 2→3 = 35% and 3→4 = 50% mean:
- only 12 in 100 one-time buyers came back for a second purchase,
- but 35 in 100 who bought twice came back for a third,
- and half of those who bought three times came back again.
Bars often rise like this, because the second purchase is the hardest one to win.
The table: Repeat purchases by cohort
One row for each month of first orders in your dates.
- New buyers: how many people made their first order that month.
- 30 / 60 / 90 / 180 / 365 days: how many of them came back within that many days. Darker cells are higher.
For example, the row Jan 2026 and column 90 days showing 18% means that of the people whose first order was in January, 18 in 100 ordered again within 90 days.
Some cells are still filling in:
- 1,200 / 1,500 observed under a rate means only 1,200 of the 1,500 buyers have had the full window so far. The rate can still change.
- Still observing means nobody has had the full window yet. It does not mean 0%.
To compare months fairly, read down a column, so every month has had the same amount of time. Reading across a row shows how one month's buyers kept coming back over time. Hover a cell to see the counts.
- A purchase is one order day for one buyer. Unpaid, cancelled, failed and fully returned orders are left out.
- A rate only counts buyers who have had the full window ("fully observed"). Recent months use only some of their buyers until the rest catch up.
- Buyers first seen in the shop's first 90 days of DataGlass history are left out, because they may have bought before that history begins.
- Time until next purchase is a smooth estimate. It assumes a returning buyer has the same chance of coming back on any day, so it shows the overall pace but not one-off spikes such as payday or a double-date sale. The shaded band is a 95% range.
- Chance of one more purchase: each bar counts only buyers who reached the step before it.
Use cases
- Judge a sale by the buyers it brought in. Pick the sale's dates and note Bought again and Value vs first purchase at 60 or 90 days. Then pick ordinary days of the same length and compare. Much lower numbers mean the sale mostly brought one-time deal hunters.
- Time a win-back message. Look at Bought again at 30, 60 and 90 days, and at where Time until next purchase levels off. If most buyers who come back do so in the first month, send a reminder or voucher at around three to four weeks, before the rest drift away.
- Find where to push. If 1→2 is low but 2→3 is high, getting the second order is the hard part. Put your follow-up effort into first-time buyers.
- Know what a new buyer is worth. GMV per buyer at 180 or 365 days is what a new buyer typically spends in that time. It's sales, not profit, so turn it into profit first. For example, ฿420 × 20% margin = ฿84, so don't spend more than about ฿84 to win one new buyer.
- Spot a drop early. If recent months in the 30 days column are lower than older months, fewer new buyers are coming back. You'll see it here before it shows up in total sales.