Docs › Your data
Snapshots and versions
Snapshots let one dataset hold several dated copies of the same sheet, such as a backlog exported every Monday, so reports can show today's numbers and how they changed over time. This guide covers snapshot datasets, adding and managing snapshots, thinning old ones, loading only recent ones, and the ways a report's earlier states are kept (report history, saved versions and export packages).
What a snapshot is#
Many systems export a list as it stands on a given day: open orders, a project backlog, account balances. Each export is a snapshot. When you keep every export in one dataset, you can compare them.
- A snapshot column is a date column that says which snapshot each row belongs to, such as "Report Date", "As Of" or "Snapshot Date". Every row of one export carries the same date.
- A snapshot date is one value of that column. All the rows with that date make up one snapshot.
- A dataset with a snapshot column is a snapshot dataset. It holds one row per item per snapshot date.
Two snapshot dates are only the same if they are written the same way: "5 Oct 2026" and "5 Oct 2026 14:30" are two different snapshots.
Plan availability: snapshots need Ultimate. They also need stored rows, so a dataset past Ultimate's limit of 50 stored datasets can't take snapshots. On other plans the snapshot controls are hidden and every sheet is read as one snapshot.
How reports read snapshot datasets#
This is what makes comparing over time work, and it happens without any setup once a dataset has a snapshot column.
- Visuals show the latest snapshot by default. A bar chart of "open orders by region" shows the newest export only, so the same backlog isn't counted once per week.
- A visual whose axis is the snapshot column shows the trend. A line or column chart with "Report Date" on its axis shows one point per snapshot.
- Snapshot dates are never grouped into weeks or months. Snapshots are levels (a backlog, a balance), not flows, so adding four weekly snapshots together would count the same items four times. A snapshot axis always keeps one point per snapshot, whatever date grouping is set.
- A page filter on the snapshot column picks the snapshot shown. Visuals that show one snapshot then show the newest one the filter allows. To look at last month's backlog, filter the snapshot column to that date.
- Cross-filtering: clicking an older snapshot in a trend chart makes the other visuals show that snapshot.
- KPIs and scorecards with a sparkline along the snapshot column read every snapshot for the sparkline and label their change "vs previous snapshot".
The per-visual History setting#
When a visual's dataset has snapshots, its settings show a History section with Snapshots:
| Option | What it does |
|---|---|
| Auto (default) | Every snapshot when the snapshot column is on the visual's axis (a trend); the latest otherwise. |
| Latest only | Always the newest snapshot (or the one a page filter picks). |
| Every snapshot | All snapshots. Use this only with the snapshot column on the axis, so each snapshot is its own point. |
The hint under it names the snapshot column: "The sheet stacks snapshots by Report Date. Auto shows every snapshot when that date is the axis (a trend), and the latest otherwise, so totals aren't counted once per snapshot."
See Visual options for the other per-visual settings.
Making a dataset a snapshot dataset#
There are four ways:
- Your file already stacks several exports with a date column. Set that column as the snapshot column, either on the dataset (Snapshots dialog) or in a single report (Snapshots by).
- The rows you have are one export. In the Snapshots dialog, choose Add a SnapshotDate column and say when the rows are as of. Later uploads are then added as snapshots, each dated when you upload it.
- You upload the next export to an existing dataset with Add or replace then Add as snapshot, or with Add snapshots from files… in the Snapshots dialog. Pick several files to add them all at once, each dated by its file (Adding several files at once). The first time, you pick the column that dates the rows, or have one added. See Adding a snapshot.
- Automatic detection in reports. If a report's dataset has a date column named like a snapshot date and nothing else is set, the report uses it. See below.
Snapshots by (in the Data pane)#
In the report builder's Data pane, each dataset with date columns shows Snapshots by (Ultimate):
| Option | Meaning |
|---|---|
| Dataset setting: column | Follow the snapshot column set on the dataset. Shown when the dataset has one. |
| Pick by name: column | Shown when the dataset has no setting. The report picks a column by its name (see below). |
| None: one snapshot | Treat the sheet as one snapshot. |
| a date column | Use this column, for this report only. |
When you pick a column on a stored dataset you are allowed to change, the pane asks: "Set column as the snapshot column on the dataset too, so every report and future uploads use it?" Choose Yes to save it on the dataset (the report then follows the dataset setting), or Just this report.
Once a snapshot column is in use, the pane shows "n snapshots, latest date. Visuals show the latest unless their axis is column." and the Load setting.
How "Pick by name" chooses a column. It looks at date columns whose header means a snapshot date, ignoring case, spaces and underscores: for example Report Date, Reporting Date, As Of, As Of Date, Snapshot, Snapshot Date, Snapshot Month, Insert Date, Effective Date, Extract Date, Export Date, Run Date, Load Date, Data Date, Refresh Date, Valuation Date, Statement Date, Period End Date, Period Ending, Week Ending, Month Ending, Report Month and Report Period. A header that contains one of these counts too. A column is only picked if the sheet has at least two rows and the column has no more distinct dates than half the number of rows, so an ordinary date column such as "Order Date" with a different date on every row isn't mistaken for one.
Adding a snapshot#
On the Datasets page, open the ⋯ menu on a dataset's row (or Actions on the dataset's own page), choose Add or replace… and pick a file. Or open Snapshots and click Add snapshots from files…: the same dialog follows, without Replace or Reload part of the data. Picking more than one file in either place opens a different dialog: see Adding several files at once. The tab named like the dataset is used, or the file's first tab. The file is read in your browser first, so the dialog can ask what it needs to before anything is uploaded.
The dialog is titled "file → dataset". Under What should this file do?:
- Add as snapshot: "Keeps every row already here and adds this file's, dated by the snapshot column."
- Replace: "Every row is replaced." On a snapshot dataset: "Snapshots are dropped." Click Replace rows.
For Add as snapshot, what you see next depends on the dataset and the file.
Case 1: the dataset has a snapshot column and every row of the file has a date in it#
Nothing to ask. A note says where the dates come from, for example "Snapshot dates come from Report Date: 1 snapshot in this file, 5 Oct 2026." Click Continue.
Case 2: some rows of the file have no date in the snapshot column#
Or the file has no such column at all. The dialog says, for example, "12 rows have no snapshot date in Report Date." Rows are never stored without a snapshot date, so choose:
- Set a date for them, then pick the date (see Choosing a date).
- Don't save these rows: they are left out of the upload completely.
Case 3: the dataset has no snapshot column yet (the first snapshot)#
Which column is the snapshot date? lists the file's date columns, best-named first, plus Add a SnapshotDate column.
- Pick a file column to make it the dataset's snapshot column. Rows with no date in it are handled as in case 2.
- Or choose Add a SnapshotDate column: a new column called SnapshotDate is added, and every row of the file gets the date you pick ("The file's n rows are as of:").
The rows already in the dataset need a date too, so the dataset reads as dated snapshots with no blanks: "The n rows already here are as of:". Last upload is offered first, because that is usually when those rows came in.
The same question appears on any later upload if some stored rows still have no snapshot date.
Choosing a date#
Wherever a date is asked for, you choose from:
| Choice | Meaning |
|---|---|
| Now | The current date and time, shown next to it. |
| Last upload: date | When the dataset's rows were last uploaded. Shown only if there was one. |
| Latest snapshot: date | The newest snapshot date. Offered when saving edits in a grid. |
| As of date | A day you type or pick. |
The stamp takes the shape of the dataset's existing snapshot dates (a date only, or a date and time).
What happens when you continue#
- Every row already stored stays as it was, and the file's rows are added.
- Each new row's snapshot date is its value in the snapshot column (or the date you chose).
- If any of the file's snapshot dates already exist in the dataset, you are asked what to do first. See Collisions.
- When it finishes, a note sums up, for example: "Added 1,310 rows to Backlog: 1 snapshot, 5 Oct 2026." It also mentions rows not saved, columns that changed type, and new columns.
- Reports open on the dataset reload with the new rows.
If something is still missing when the file reaches the server (for example rows without a date), nothing is saved and you see a message such as "12 rows of the file have no snapshot date, so nothing was saved. Try again and give them a date."
What happens to columns#
- New columns in the file are added to the dataset. Earlier snapshots read them as blank: "New columns: Owner (blank on earlier snapshots)."
- Missing columns: a column the file doesn't have reads as blank on its new rows.
- Type clashes: if the file has values that don't fit a column's stored type (text in a number column, for example), the column becomes text for every row: "Amount now holds text, since the file's values didn't fit." The upload is never stopped by a type clash.
Adding several files at once#
When you have a backlog of exports, one per week say, you can add them all in one go. Choose Add or replace… on the dataset (or Add snapshots from files… in its Snapshots dialog) and select several files. Each file becomes a snapshot, dated by the file.
The dialog is titled "n files → dataset" and lists each File with its Rows and Snapshot date. Each file's date starts as:
- The date in the file's name, for example "Backlog 2026-10-05.xlsx", "backlog_20261005.csv", "Stock 5 Oct 2026.xlsx" or "Orders 10-05-2026.csv". A month alone ("Stock October 2026.xlsx", "Stock 2026-10.xlsx") means the 1st of that month. If some names put the day first ("25-10-2026"), every name is read day first.
- Otherwise Modified: when the file was last changed, which is usually when it was exported.
Change any file's date with its list: From the name, Modified, Now or As of… a day you type.
- A file that already has the snapshot column keeps its own dates: it shows "From the file: 2 snapshots in its Report Date column, …". Only its rows with no date there take the file's date.
- Dataset with no snapshot column yet: the dialog first asks Which column is the snapshot date?, as for one file. It defaults to a column named like a snapshot date that every file has, else Add a SnapshotDate column. It also asks when the rows already here are as of.
- Files with the same date make one snapshot together, and the dialog says so.
- All or nothing. Every file is sent as one upload: if anything is refused, no file is saved.
- Dates already in the dataset are asked about once for all the files (below).
- Every date in one upload takes the same shape: the dataset's, or the dates the files already hold. Otherwise, when any file is dated by a day (its name or As of), every file is dated by its day only.
Click Add n snapshots. The note afterwards lists the snapshots added, as for one file.
When a snapshot date already exists#
Each snapshot date can only exist once in a dataset. If the file has rows for a date the dataset already holds, you are asked before anything is saved.
One date: "A snapshot dated 5 Oct 2026 is already here (1,240 rows). The file has 1,310 rows for that date." Choose:
- Add to: keep the rows already in that snapshot and add the file's.
- Overwrite: remove that snapshot's rows first, so the file's rows take their place.
Several dates: a table lists each Snapshot with Rows here and Rows in file, and a Do choice for each (Add to is selected by default). Apply to all sets every row at once. Click Continue.
Behind the scenes:
- The file is uploaded once. Your answer is applied to the rows already sent, so nothing is uploaded twice. Unanswered uploads are discarded after 24 hours.
- Overwritten rows are removed the same way a deleted row is, with their history kept until the dataset is next tidied up (compacted).
- Cancel saves nothing.
The Snapshots dialog#
Choose Snapshots from the ⋯ menu on a dataset's row on the Datasets page, or from Actions (or Open beside Snapshots in the details) on the dataset's own page. Anyone who can open the dataset can view it. Only the uploader or an admin (with access to every row) can change it; others see "Only the dataset's uploader or an admin can change these settings." The subtitle shows "n snapshots, n rows" or "Not a snapshot dataset yet".
Snapshot column#
Pick one of the dataset's stored date columns, or None. "(added by an upload)" marks a SnapshotDate column created by an upload.
When the dataset has no snapshot column yet (and no column called SnapshotDate), the list also offers Add a SnapshotDate column. Pick it, then choose when the rows already in the dataset are as of: Last upload (offered first), Now or As of a day. Click Add column. Every stored row gets that date, and SnapshotDate becomes the snapshot column. A dataset with no rows just gains the column. Each later Add as snapshot asks for a date for the file's rows, with Now picked, so every upload is dated when it came in. Every report uses this column unless the report picks its own ("Reports can still pick their own."). Choosing None makes it an ordinary dataset again; reports that pick their own snapshot column keep it.
Dates and times#
How date-times in this dataset are kept:
- Wall-clock: shown as written, the same in every time zone. 6 PM stays 6 PM for everyone. This is the default.
- Zoned: kept with the time zone they were entered in, shown in each viewer's. 6 PM in Chicago shows as 7 PM in New York.
To switch, pick the other option, choose a time zone, and click Apply. On a dataset with rows, a warning explains how every stored date-time will be rewritten, for example "Every date and time already in this dataset will be set to your time zone, America/Chicago…". Click Rewrite dates and times to go ahead.
- Each value gets the offset that applied on its own date, so July values get daylight time and January values standard time.
- Dates without a time are never changed.
- Switching from zoned to wall-clock can make two snapshot dates read as the same clock time (during a daylight-saving change). The dialog lists the pairs that would merge into one snapshot, and the button becomes Add to: merge them.
- Saved versions of reports keep the values as they were.
Lock older snapshots#
Tick Only the latest snapshot can be edited so grids can change only the newest snapshot's rows. Earlier snapshots stay as they were recorded. Removing or replacing a whole snapshot still works from this dialog. The lock applies to everyone, including the owner and admins.
Snapshots list#
Each snapshot date is listed with its row count (counting only rows you can see). For people who can change the dataset:
- Add snapshots from files… (above the list, shown on any dataset you can change, snapshot dataset or not): pick a file and it is added as a snapshot, as under Adding a snapshot, or pick several (Adding several files at once). On a dataset with no snapshot column yet, the dialog first asks which column dates the rows, or adds a SnapshotDate column.
- Remove: asks "Remove the date snapshot (n rows)?" Every row with that date is removed from the dataset and every report on it. You can upload it again as a snapshot.
- Replace with file…: pick a file. Every row of the file is given this snapshot's date and replaces the snapshot's rows. Nothing else is asked. The note says "The date snapshot of dataset now holds the file's n rows."
Retention#
Editing snapshot data#
Rows of a snapshot dataset are edited like any other on the dataset view page (Ultimate). See Datasets. Two extra rules apply:
- Locked older snapshots. With the lock on, rows outside the latest snapshot show a lock instead of edit and delete buttons, and the page says "Older snapshots are locked: only rows dated date can be edited here." The server also refuses edits that would date a row earlier than the latest snapshot. A row dated after the latest is allowed, and becomes the new latest.
- Every row needs a snapshot date. If you add a row without one, or clear it, Save changes first asks "n rows have no snapshot date". Choose Set a date (Latest snapshot, Now or As of) or Don't save these rows ("Their changes are dropped; the rest of your changes are saved."), then Save changes.
Thinning old snapshots (retention)#
A dataset that gains a snapshot every week grows forever. A retention rule removes old snapshots automatically while keeping a useful history, for example every snapshot for three months and then one per month.
In the Snapshots dialog, under Retention:
- Choose Keep every snapshot (no rule) or Thin old snapshots.
- Pick a Rule:
- All for 3 weeks, then Sundays
- All for 3 months, then the 1st
- All for 3 months, then the 1st, 15th and last
- Custom
- Adjust the rule if you like:
- Keep every snapshot for n days / weeks / months / years. 0 means there is no such period, and the days below apply to every snapshot.
- Then keep some days of the week (tick Sun to Sat), some days of the month (tick 1 to 31 and Last), or every snapshot.
- Drop anything older than n days / weeks / months / years (optional). This must be longer than the keep-every period.
- Click Preview to see what the rule would do, for example "Keeps 13 of 48 snapshots. Removes 35 snapshots, 1.4 million rows." It lists the snapshots kept, and how many saved versions of reports still hold the rows.
- Click Save rule and confirm in the "Thin old snapshots?" dialog.
How it works:
- A sentence under the rule describes it, for example "Every snapshot is kept for 3 months, then only the 1st of each month. Nothing older than 2 years is kept."
- The rule is applied automatically on the next maintenance pass, and again on every pass after, so snapshots that age past the period are thinned as time goes on. Saving the rule removes nothing immediately.
- Ages are counted back from today's date (in UTC). A snapshot's day is its date as written; on a zoned dataset, the day in the time zone it was entered in. A snapshot whose date can't be read is never removed.
- A rule that keeps everything can't be saved: "This rule keeps every snapshot. Pick days to keep after the first period, or an age to drop them at."
- Removed snapshots can't be brought back once the dataset is tidied up. You can upload their rows again as a snapshot, but their edit history is gone.
- Saved versions of reports made before still read the removed rows, and their storage is freed only when those saved versions are deleted.
- Clearing the snapshot column also clears the rule.
Only the uploader or an admin can change the rule; others see its description.
Loading a window of snapshots#
To open a big snapshot dataset quickly, a report can load just the recent snapshots.
In the Data pane, under the snapshot note, Load offers:
- Every snapshot (default).
- The last… n weeks, months or years.
How the window works:
- It is counted back from midnight today, in your time zone. "The last 12 weeks" keeps every snapshot dated on or after that day.
- If no snapshot is that recent, the newest snapshot alone is loaded, so the report never goes blank.
- Nothing is removed. Choose Every snapshot to load them all again.
- The choice is saved with the report, so everyone who opens it gets the same window.
- For stored datasets, only the rows in the window are fetched from the server. Changing the window fetches the rows again. Any grid edits you made to that dataset in the report are dropped when it is fetched again.
- When a window is in use, the pane shows "Showing 12 of 48 snapshots."
The large dataset prompt offers the same window (Load the last…, default 12 weeks) when a report opens a stored dataset of more than 500,000 rows without one.
Connections with snapshot datasets#
Connections that involve a snapshot dataset follow two rules, in both the workspace's Connections section and the builder's Connections tab. See Connecting datasets for connections in general.
- A snapshot dataset can be the many side of a one-to-many from a dataset that isn't one. Customers to Orders-by-snapshot works: each customer links to its orders in every snapshot.
- Two snapshot datasets must also match on their snapshot dates, so each snapshot links to the same date's rows. The workspace dialog adds that pair for you, shown locked with the note "Snapshot dates are matched automatically so each snapshot links to the same date's rows." In the builder, click Match the snapshot dates.
Anything else is refused with a reason, for example:
- "Backlog holds snapshots, so it can only be the many side: every Customers row would match once per snapshot. Swap the sides."
- "Backlog holds snapshots and Customers doesn't, so they can't be one-to-one…"
If a dataset's snapshot column changes later and a connection no longer meets these rules, it is marked Not in use with the reason, and reports leave it out until it is fixed. Row access filtering keeps flowing along it, so filtered people never see more than before.
Plan availability: snapshots are Ultimate, and saving a workspace connection needs Plus or Ultimate on the server.
Comparing over time: a worked example#
- Upload the first export of your backlog, for example "Backlog.xlsx" with a "Report Date" column.
- Open Snapshots on the dataset and set Snapshot column to Report Date. (If the column is named like a snapshot date, reports already pick it by name.)
- Each week, choose Add or replace… from the dataset's ⋯ menu, pick the new export and choose Add as snapshot. If you re-export a week by mistake, choose Overwrite for that date.
- In a report, a bar chart of open items by status shows this week. A line chart with Report Date on its axis shows the trend. A KPI with a sparkline along Report Date shows the change "vs previous snapshot".
- To see an older week everywhere on the page, add a page filter on Report Date and pick that date.
- Once the dataset grows, add a retention rule (for example All for 3 months, then the 1st) and set reports to Load the last 12 weeks.
Report history (undo and redo)#
The clock button in the builder's toolbar opens Report history in place of the selected visual's settings. It lists every change made in this editing session, newest first, with the time, and Report opened at the bottom.
- Click a row to go back to the report as it was right after that change. Undone changes are dimmed.
- Ctrl+Z undoes and Ctrl+Y redoes.
- The last 100 changes are kept.
- Changes that added a page or a dataset are marked +. Undo and redo skip them, and the page or dataset stays (in its starting state) when you go back before it ("kept as added").
- History covers the report's layout and settings. Edits to data in grids are not part of it; use Revert edits in the Data pane. See Datasets.
- History lasts only while the report is open. It starts again each time the report is opened.
Available on every plan. See Report builder.
Saved versions#
A saved version is meant to keep a report exactly as it stood, together with the exact state of the datasets it read at that moment, so it can be read again later even after the data has changed.
What exists today:
- The server can create, list, open and delete saved versions of a report (Ultimate only, by people with edit access; deleting needs full access to the report).
- A saved version keeps reading the rows it captured, even after edits, removed snapshots or retention rules. Replacing the whole dataset ends that: a saved version of a dataset that has since been replaced can no longer be read.
- Retention previews and the time-zone rewrite already count and mention saved versions ("n saved versions of reports still hold these rows").
Export the report with its data#
An .ifpr is one file holding the whole report and its data: the definition (pages, visuals, filters), every dataset's rows with your edits applied, and the log of those edits. The desktop app saves its reports in the same files.
- In the builder toolbar, open Export and choose Download report (.ifpr) ("The report, its data and edits, in one file the desktop app opens too."). A file named after the report, ending in
.ifpr, is downloaded. - Download definition (.json) ("Pages, visuals and filters. No data.") downloads the layout only, as
.report.json. - To open either file, choose Export > Import… ("Open an .ifpr, a definition or a package file."). An
.ifprbrings its rows and edits with it, as do the older.package.jsonpackages. A definition alone binds to the datasets already loaded. A file saved by a newer version of the app is refused with "…was saved by a newer version. Update and try again." Import is available to people who can edit the report. - The file is built in your browser and contains every row you can see, so treat it as you would the original spreadsheet.
Plan availability: Download report (.ifpr) is Plus and Ultimate on the website (anyone can save .ifpr files in the desktop app). Download definition (.json) and Import… are available on Mini, Plus and Ultimate. Free shows Sign up for Save and Export instead. Guests get Print only. A viewer on Plus whose report reads files sees Import data instead of Export (see Datasets).
Plan limits summary#
| Feature | Free | Mini | Plus | Ultimate |
|---|---|---|---|---|
| Snapshot datasets (snapshot column, Snapshots by, History setting) | No | No | No | Yes |
| Add as snapshot, collisions, Remove / Replace with file | No | No | No | Yes |
| Lock older snapshots, Dates and times mode | No | No | No | Yes |
| Retention rules | No | No | No | Yes |
| Load a window of snapshots | No | No | No | Yes |
| Large dataset prompt | No | No | Yes (Load everything only) | Yes |
| Report history (undo and redo) | Yes | Yes | Yes | Yes |
| Saved versions | No | No | No | Server only, no screen yet |
| Export Definition / Import… | No | Yes | Yes | Yes |
| Export Report with data (.ifpr) | No | No | Yes | Yes |