Skip to content

Data Foundation: The Common Finance Model

Data Foundation is the finance data foundation for your PlaidCloud projects. Its Common Finance Model brings together ERP data from multiple ERP systems, and multiple instances of each, into a single model: one chart of accounts, one set of entities and one reporting currency. Reports, metrics and AI assistants built on the model all read the same numbers, whichever ERP they came from.

Your ERP connects through its PlaidCloud connection, so the general ledger, trial balance, master data and budgets arrive directly from the source, with nothing to export. Files add data a system doesn’t hold, such as currency rates, allocation drivers or a budget kept outside the ERP, and can replace the ERP rows they cover. Load conforms every source into the model, translates it into the reporting currency and checks that it ties out, all from the project’s Foundation tab. Where sources use the same code for different things, you say whether they mean the same, and a management consolidation builds the group profit and loss with the intercompany eliminated. Templates, such as Enterprise Profitability, build their reports and models on top of it.

ERP systems & instancesERP A, instance 1ERP A, instance 2ERP BFiles: rates, driversCommon Finance ModelOne chart of accountsOne set of entitiesOne reporting currencyLoaded and checked on the Foundation tabReportsMetricsAI assistants
Every ERP system and every instance lands in one model, so every report, metric and AI answer reads the same numbers.

Your ERP connects to Data Foundation through its PlaidCloud connection. Data Foundation reads the general ledger, trial balance, master data, exchange rates and budgets directly from it, and only ever reads. You can pick an existing connection or create one on the tab. Each system has a page for the administrator who grants the access, listed in Which Systems Can Feed the Model, and the Data Foundation Coverage Matrix shows which datasets each system supplies directly.

  1. In Analyze, open Projects and click New Project. Enter a name, and under Start from, choose a template, such as Enterprise Profitability (template), instead of Blank project. Click Create.

    The project opens on its Foundation tab. It is private to you, and you are its only Architect, until you share it.

  2. On the Foundation tab, open Sources and, in Connect a system, choose your kind of system from its tile. Then pick an existing connection for it, or choose Create a new connection. Each existing connection lists the workflows that already use it, or says Not used by any workflow yet, and a connection that is already a source is listed as already a source and can’t be picked again.

  3. For a new connection, enter a Connection name, choose the Environment, and fill in the credentials your ERP connection asks for.

    Don’t have the credentials yet? The form shows a ready-made access request for your ERP administrator. Click Copy request or Email request, send it, and enter the values they send back.

  4. Click Test connection. PlaidCloud signs in and tries to read each dataset, then lists each as Readable, Missing permission or Error. When something is missing, it names the permissions to ask your ERP administrator to add.

  5. Set Months of history: how many months back to load, from 1 to 120. The default is 24.

  6. For an existing connection, tick the entities to load, or leave them all clear to load every entity. Inactive and elimination entities stay hidden until you choose to show them. Then click Connect.

  7. For a new connection, click Connect to create it. Then tick the entities to load and click Connect again to save them. Until you do, every entity is loaded.

  8. To bring in another ERP system or another instance, open Connect another system and repeat these steps. Connecting always adds a source and leaves the existing ones as they are, so several ERP systems and instances can feed the one consolidated model.

Once connected, the card says which connection, environment, entities and months of history the project loads.

Each load reads only what changed: open periods and the last closed period are read again, and closed periods already loaded are kept. Raising Months of history reads just the added months. A changed exchange rate, entity crosswalk or reporting currency reads again the months it affects.

Every ERP system and instance the project loads from is a source. The Load order list on the Sources page of the Foundation tab shows one line for each, in order, with its name and system, its Companies, Last loaded, Rows and Status. Companies says All or how many are selected. A source that hasn’t loaded yet says Not loaded yet. Files you confirm are sources too.

Click Edit on a source to change how it is used:

Setting What it does
Name The label reports, mapping questions and messages use for the source.
Place in the order Where the source ranks. When two sources cover the same month of the same company, the one higher in the order wins.
Reporting books, separated by commas The accounting books to read, for a system that keeps more than one. All reporting books reads every book.
Companies For a system source, which of its companies (subsidiaries) the source loads. The list is read from the system when you first open the panel and shows the companies the source loads now. Tick the ones to load, or tick none to load all of them. If the system doesn’t list its companies, enter their names separated by commas.
Company, From (2026-01) and To (2026-12) The months this source provides for each company. Leave a month empty for no limit. Write months as the year and month, for example 2026-03.

Click Save to keep your changes. Adding companies, or ticking none, saves straight away. Leaving a company out asks you to confirm Load fewer companies? first, because saving clears everything loaded from that source right away, and the next Load pulls the ticked companies again. If the source’s connection now reads a different account than the one the source was set up for, the save is refused with a message saying so, so a change can never replace another source. Reconnect that account as its own source instead. To drop a source, click Remove this source and confirm Remove. Everything loaded from it is deleted, and you can connect it again later.

Changing sources does not read your ERP again. The bar above every page says You have changed your sources. Press Apply changes to use them. until you click Apply changes, which runs a load from the data already pulled: it conforms the sources again with your new order, mappings, names and files, and rebuilds the model. Click Load when you also want fresh data from the ERPs.

Files supplement your ERP connection with data a system doesn’t hold, such as currency rates, allocation drivers, an Entity crosswalk or a budget kept outside the ERP. Each file belongs to one of the datasets listed under Common Finance Model Tables.

  1. Under Files, choose the Document account to store the file in.

  2. Drop a CSV or Excel file on Drop a CSV or Excel file here, or click Choose a file. The file is uploaded to the /foundation/ folder of the document account.

    For a file that is already in the document account, enter its File path, for example /foundation/budget.csv, and click Add file.

    To read a new file every month, add every file in a folder whose name matches: enter the folder and what File names that contain, start with, end with, match exactly or match as a pattern, and click Add files. Files in folders inside it are not read.

  3. PlaidCloud reads the file and proposes a mapping:

    • Detected delimiter, decimal separator and dates. The delimiter (comma, semicolon, tab or pipe), the decimal separator and the date format are detected for you.
    • Dataset: the Common Finance Model table the file belongs to, matched from the file name and column headings. When the match isn’t clear, choose it from the list.
    • Decimal separator: Period (1,234.56) or Comma (1.234,56).
    • Date format: shown when the file has dates. Change it if the detected format is wrong: YYYY-MM-DD, DD.MM.YYYY, DD/MM/YYYY or MM/DD/YYYY.
    • A grid of Column in file, Type, Examples and Goes to. Change Goes to to send a column to a different field, or to Don’t load.
  4. Answer the questions the file raises, such as Which company is this file for?, Which currency are the amounts in?, Are these amounts year-to-date or for the month?, Are the amounts shown in thousands or millions? or Are revenues shown as negative numbers? Each one is asked only when the file doesn’t already say.

  5. Check where each column goes, then click Confirm mapping. While it works, the note under the button says why it is waiting. In a project that still has the sample data, the first file you confirm also removes the sample, which can take a few minutes.

These options appear on a file when they apply:

Option Use it to
Which source do these rows apply to? Limit a code map file to one source. Every source applies it to all of them.
Where does this file come from? Say whether the file holds This project’s own codes or comes from Another system, which you name, for example SAP Germany export. A file from another system is checked for shared codes like an ERP source.
Split … into … parts at each … (under Advanced) Split one column that holds several values, such as 100-200-300, and send each part to its own field.
Which currency do you report in? Set the reporting currency when the project hasn’t one yet. Write it as three letters, for example USD.
These are consolidated results (OneStream, Anaplan, SAP Group Reporting, FCCS) Load the results of your consolidation system so the group profit and loss can be reconciled.

To take a file out of the project, click Remove next to it under Files, or Remove this file on its mapping, and confirm. Its figures leave your reports the next time you click Apply changes.

When PlaidCloud can’t decide the dataset, the decimal separator or the date order on its own, for example a file whose dates could read as day/month or month/day, it asks you to choose before you confirm. Adding a second file for the same dataset replaces the first. If a later file’s columns don’t match the dataset, the load names the file and the columns and tells you to correct the file or its column names, or set the step’s Schema Mismatch to union.

Accepted files are CSV, TSV and TXT text files, Excel .xlsx files (read from the first sheet) and Parquet files.

Click Load on Status. The load reads your connected ERP and your confirmed files, conforms them into the Common Finance Model, translates them into the reporting currency, runs the readiness checks and rebuilds the template’s models.

  • Only one load runs for a project at a time. A second Load or Apply changes says a load is already running, and you can’t change sources while one runs.
  • If other workflows use the same ERP connection, the first load lists them and asks you to confirm, because a load reads from the ERP and can slow those workflows down. Click Load anyway to go ahead.
  • After PlaidCloud is upgraded, an Architect’s Load also updates the project’s generated workflows, creating any working tables they need. Until then the status page lists an amber item, “The workflows were generated by an earlier release. An Architect’s Load updates them.”, and a Manager’s Load warns that only an Architect can update them.
  • Applying an update keeps your own edits to the generated steps. If you changed a step and the update changed it too, your edit is kept and the status page notes the conflict by step name. A step you edited that the update no longer generates is kept but turned off; delete it, or turn it back on if you still need it. A step the update replaces is named, and the earlier one stays in its step versions.
  • If an update stopped part way, or is still running, the status page says so, and its steps may run out of order until it finishes. Press Load again as an Architect to finish it. If it stops again, contact support.
  • A failed load shows red and never hides behind earlier results. If the consolidation fails, the status names the step it stopped at instead of showing old group results. Rows from an earlier load show From an earlier load. Load again. until a load succeeds.
  • If one entity and period fails to read, the rest still load. Run Load again to fetch only what failed.
  • If one dataset of a source fails to read, the rest of the source still loads, and that dataset keeps the rows it had. What to do next on Status lists an amber Data loaded item for each step of a source’s read that failed in the last load, naming the source, the step and its error.

When sources overlap, a load keeps one copy of each figure and holds the others back instead of counting them twice. Held-back rows are recorded with their source, entity and the reason, and the load finishes with a warning that names the counts, for example 120 rows of NetSuite EU also provided by a source ranked ahead. The reasons are:

Reason Meaning
In another accounting book The row belongs to a reporting book the source isn’t set to read.
Outside the periods the source provides The row falls outside the From and To months set for the company.
Also provided by a source ranked ahead A source higher in the order covers the same company and month.
Also in an earlier file of the source An earlier file of the same source already covers the month.
Replaced by a file A file replaces the ERP rows for that entity and month.
Carrying a code another source also uses The code needs your answer. See Map Codes Across Sources.
Carrying a code mapped to two codes The code map gives the code two answers.

Statistical accounts move to the allocation drivers instead of the ledger, and show as moved, not as a problem.

A file replaces only what it covers, and the other rows from your ERP stay.

  • Master files (accounts, departments, customers, products and Segment values) replace the ERP rows with the same code.
  • Transaction files replace the ERP rows for the same entity and month. General ledger and trial balance files also match by ledger, budget files by scenario, and statistical driver files by scenario and driver.

If a file later stops covering a month, the ERP’s data for that month comes back on the next load.

Each source names its entities with its own codes. The Entity crosswalk maps those codes to the entities in your entity master, so lines from every source land on the same entity.

To use one, add a CSV or Excel file and choose the Entity crosswalk dataset:

Column Holds
source_entity The code as the source carries it, such as a company code.
entity The entity it belongs to, as the entity master names it.
  • A code that isn’t in the crosswalk is used as it is.
  • Several codes can map to one entity. Their lines are combined, and a month counts as closed once it is closed for all of them.
  • Codes in the crosswalk apply to every source, so each source’s entity codes must be unique across sources.
  • An Entities file row replaces the ERP’s entity with the same entity code.

The Unmapped entities check flags any line whose entity is neither in the entity master nor mapped to one by the crosswalk. Until an entity master is loaded, it shows – Not applicable.

Two sources can use the same code for different things, or different codes for the same thing. When a code appears in more than one source, Data Foundation holds its rows back until you say whether it means the same everywhere, so each amount is counted once. The Codes waiting for your answer check lists them, the first 10 and then “and K more”, and Answer codes › in the What to do next list on Status opens them.

Answer codes › opens Answers needed, which lists Codes waiting for your answer. It groups the codes by what they are: Account types, Companies, Accounts, Departments, Profit centers, Customers, Products and each extra category. For each code it shows what each source calls it and a suggestion:

  • Looks like the same thing when the sources give the code the same name.
  • Looks like different things when the names differ.
  • We can’t tell. Please answer. when there is nothing to go on.

Answer one code at a time with Yes, the same or No, different, or answer a whole group at once with Accept all (take every suggestion), All the same or All different. For accounts, you can also answer a whole parent group of the chart at once. Where a source carries no account types, choose Use this type for a group of accounts: Revenue, Cost of goods sold, Operating expenses, Non-operating income and expenses, Assets, Liabilities or Equity.

Mappings are kept for each source and dimension, so an answer for one source’s departments never changes another source’s accounts. Answers are saved as you go, with Saved. Press Apply changes to use your answers. Click Apply changes to use them.

You can also supply the mappings in a file. Add a file with the Code map dataset, and choose Which source do these rows apply to? to limit it to one source.

When two sources provide the same months, the Which source covers which months check asks you to pick. Click Choose sources › in the What to do next list on Status to open Source coverage, which lists each overlap by company. Choose the source that counts, enter the month it counts from in That month is, for example 2026-04, click Save choice, then click Apply changes in the bar above the page. The other source keeps the earlier months.

The same dialog warns when two sources give different exchange rates for a currency; the rates from the source higher in the order are used.

Each load translates the general ledger, trial balance, sales, inventory and budget amounts from their local currency into one reporting currency. Reports and models read the translated amounts, so entities in different currencies, from any system, total correctly.

To choose the reporting currency, set the project variable reporting_currency to a currency code such as USD. See Managing Project Variables. The template creates it blank. While it is blank, each load uses the first of these that names exactly one currency:

  1. The group currency of the entities.
  2. The local currency of the top-level entities, the ones with no parent.
  3. The local currency of the files.

If none of them names exactly one currency, Currency translation shows a problem until you set reporting_currency.

Amounts already in the reporting currency stay as they are, so a project whose data is all in one currency needs no rates.

Sales, inventory and budget files can carry a currency too. Map a column named Currency or Local Currency and it is picked up automatically, and the file’s local amounts are translated to the reporting currency. A group or reporting amount that is already in the file is kept as it is. Product standard cost is not translated.

A rate converts one unit of a currency into the reporting currency for a month: the translated amount is the local amount times the rate.

  • From your ERP, each load reads the exchange rates the ERP holds.
  • From a file, add a CSV or Excel file and choose the Currency rates dataset. A rate in the file replaces the ERP’s rate for the same currencies and month. Use a file to report in a currency your ERPs don’t publish rates into, or to add planned rates for budget months.
Column Holds
from_currency The currency to translate from, for example EUR.
to_currency The reporting currency, for example USD.
period The month, as YYYY-MM.
average_rate The average rate for the month.
closing_rate The rate at the end of the month.
historical_rate The historical rate, for equity and for accounts set to the historical rate.

When a currency has no rate into the reporting currency, its rate the other way round, from the reporting currency into it, is turned around, keeping 12 decimal places. Failing that, its rate into another currency is multiplied by that currency’s rate into the reporting currency.

Each row is translated from its own currency, so a line in a different currency from its entity is translated from the currency it is in. The first rule that fits the row applies:

Row Rate
Statistical and non-posting accounts Not translated.
Already in the reporting currency Not translated.
Account with a rate type That rate: average, closing or historical. Historical falls back to closing when there is no historical rate. From an ERP, the rate type is the account’s own rate type. In a chart of accounts file, map a rate type column to fx_rate_type with average, closing or historical.
Sales line The average rate for the period.
Inventory balance The closing rate for the period.
Budget row By account, like the general ledger: the average rate for profit and loss accounts and the closing rate for balance sheet accounts.
Profit and loss account The average rate for the period.
Equity account The historical rate, or the closing rate when there is no historical rate.
Any other balance sheet account The closing rate for the period.

Translated amounts aren’t rounded, so totals don’t drift.

Each load rebuilds the dimensions from the master data of every source. There are seven dimensions to start with, and up to ten extra categories:

Dimension Built from
Entity Entities and their parents
Account Chart of accounts and parent accounts
Department Departments and their parents
Customer Customers and their parent customers
Product Products and their parent products
Segment 1 Segment values for segment_1
Segment 2 Segment values for segment_2
Segment 3 to Segment 10 Segment values for segment_3 to segment_10, when your sources carry them
  • A record with no parent, or one that names itself as its parent, loads as a top-level member.
  • Because each load rebuilds the hierarchies, change a parent in the ERP or in your file, not in the dimension.
  • A problem in one hierarchy affects only that dimension. The load and the template’s models still complete.
  • What to do next on Status lists an amber item for each hierarchy step that failed in the last load, naming the step and its error.
  • You can rename each extra category to what it holds, for example Class or Region. Under Names on the Sources page, enter the names your reports use and click Save names. Names use letters, numbers, spaces, dots, dashes and underscores, up to 60 characters.
  • Each source says which of your category names each of its own extra categories goes under. Click Edit on the source, choose the name for each, or choose Leave out, and click Save. Sources that call a category by different names, such as Class and Business area, then share one slot.

To supply segment hierarchies from a file, choose the Segment values dataset:

Column Holds
segment Which segment the value belongs to: segment_1 to segment_10.
segment_value The value, as the general ledger, sales and budget lines carry it.
segment_value_name Its name.
parent_segment_value Its parent within the same segment.

After each load, the Readiness grid lists every check for every entity and period, with one of these statuses:

Status Meaning
● Ready The check passed.
▲ Needs attention Worth a look, but the numbers are usable.
■ Problem Fix this before you rely on the results.
– Not applicable There is no data for this check yet.

The checks are:

Check What it confirms
Data loaded Every fiscal period has general ledger lines, and every ERP read succeeded. A closed period with no lines is a problem; an open period with none needs attention.
Trial balance ties to GL The trial balance equals the general ledger for each entity, period and account. Where both the trial balance and the general ledger carry departments for an account, it also ties out each department.
Sales tie to GL Sales lines equal the general ledger revenue and COGS accounts for each period.
Unknown products/customers Every sales line names a product and customer that exist in the product and customer masters.
Accounts missing from chart Every general ledger line names an account in the chart of accounts. Lines with an unknown account are left out of the profit and loss. An account with postings but no account type is also a problem, and a source whose chart types Revenue accounts but no cost of goods sold account needs attention.
Currency translation Every general ledger, trial balance, sales, inventory and budget row has an amount in the reporting currency. A profit and loss or sales row with no rate is a problem, because the profit and loss reads translated amounts. A balance sheet, inventory or budget row with no rate needs attention. A budget month with no rate yet shows amber, so load planned rates for budget months. See Currency Translation.
Unmapped entities Every general ledger, trial balance and sales line names an entity in the entity master, directly or through the entity crosswalk. Not applicable until an entity master is loaded. See Entity Crosswalk.

Loads with more than one source add these checks:

Check What it confirms
Reporting book A source that carries several accounting books for an entity has had its reporting book chosen. Until you choose, only the first book counts.
Codes waiting for your answer Every code used by more than one source has been answered. See Map Codes Across Sources.
Extra categories from your sources Every extra category a source carries has a name or has been left out.
Which source covers which months No two sources provide the same months of a company unless you chose which counts.
Account types and hierarchy Every parent that a dimension names exists, and no member is its own parent or ancestor. A missing parent puts its children at the top of the hierarchy, and a loop stops that hierarchy loading. Totals are unaffected.
Fiscal calendar Each entity’s fiscal periods are calendar months, and every month of rows can be placed in one fiscal period. It needs attention when an entity’s periods are not calendar months (4-4-5, 13 periods), or when a month’s rows could be in more than one fiscal period or the fiscal calendar lacks the month.

Each row that isn’t ready says what to do in the What to do column, such as loading the chart of accounts, loading currency rates or running Load again for a period.

The grid also shows two setup checks, which appear only when they need attention:

Check When it appears
Overhead allocation Only one of the project variables overhead_department and overhead_driver is set, or no allocation driver matches overhead_driver, so the allocation would do nothing. Set both, or clear both to turn overhead allocation off.
Budget comparison Budget rows are loaded, but none are in the version the project variable budget_version names, so the budget comparison would be empty.

Each load builds a management consolidation: the group profit and loss of every entity in the group, in the reporting currency, with intercompany eliminated. It is for management reporting. Your consolidation system stays the record for the statutory group; reconcile to it below.

On the Sources page, open Group and row access to see the companies in the group. A company you clear stays in your reports on its own, but is left out of the group total. Click Save group. The change applies at the next Load or Apply changes. A new company is in the group until you clear it.

The group’s fiscal year starts in the month set by the project variable group_fy_start_month, which is 1 unless you change it.

Table Holds
Group P&L One row per entity, period and line (revenue, cost of goods sold, operating expenses, non-operating, intercompany difference), in the reporting currency. Revenue and costs carry their natural sign.
Group P&L year to date The same, year to date, to the group’s last fully loaded period.
Intercompany pairs One row per pair of entities, group and period, with each side’s amount and the residual.
Intercompany lines The ledger lines behind a group figure or a pair, with the trading partner.
Reconciliation The group P&L year to date against your consolidation system.

Three entity names appear in these tables:

Entity Meaning
Eliminations The intercompany eliminations. They net to zero, so leave them in for the group figures and filter them out for the sum of the entities before elimination.
Group total The consolidation system’s total, in the reconciliation.
Unmapped Lines with no period, and accounts or entities of your consolidation system that this project doesn’t have.

The metrics Consolidated income, Consolidated result and IC difference, and an analysis path for each table, answer questions about the group.

Intercompany lines are matched by trading partner. The partner on the line is used; when a line has none, its customer’s partner is. Both sides of a pair are eliminated, and the amount the two sides don’t match by is booked to an intercompany difference line. Each pair gets a reason:

Reason Meaning
Matched Both sides booked the same amount.
Only one side booked One entity booked the transaction and its partner didn’t.
Not eliminated A line of the pair has no exchange rate, so neither side is eliminated.

A pair is judged on its year to date. The Intercompany balances row turns amber when a pair’s difference is over the project variable group_amber (default 1) and red over group_red (default 1000). Lines with no partner, or with a partner that isn’t an entity, are not eliminated and are counted on that row. A source that carries partners beside one that doesn’t shows amber.

A source’s own elimination entities stay out of the group P&L, and the Consolidation row says so.

Add the results your consolidation system reports as a file, and tick These are consolidated results (OneStream, Anaplan, SAP Group Reporting, FCCS) when you confirm its mapping. Answer Which version of the numbers is this? with the system’s name. The Reconciliation table then shows ours, reported and diff for each entity and line, year to date. A difference is usually explained on the Eliminations, Group total and Unmapped rows, or by the exchange rates the system uses. The Reconciliation to the source row lists every line over group_amber.

Group results follow the project’s row access. When row access stops the group results, the status shows amber. A person who sees only some entities sees those entities’ sums before elimination. The Eliminations rows and the intercompany tables need access to every entity.

If the project limits who sees which rows by a column the group tables don’t carry, such as department, the group results stay off for everyone and Consolidation says Group views are off. Only an Architect can then tick Everyone on the project may see the group results under Group. The change applies at the next Load or Apply changes.

After a load from your ERP connection succeeds with no problems, Refresh nightly appears on Status. Click it to run Load every night at 2:00 in your time zone. Status then shows Nightly refresh: Every night at 2:00 and who set it up.

The nightly load runs as the person who set it up. If that person can no longer use the ERP connection, the grid says the nightly load is skipped; set it up again. A project that has only sample data can’t be scheduled.

A file that is missing or empty on a nightly load finishes with a warning that names the path, not an error, and the rest of the model still loads.

Business names, under Names on the Sources page of the Foundation tab, give the codes in your model the names your business uses. Reports, metrics and the assistant show these names, and a name you save applies at once. Later loads keep it.

  1. Click Add a name and choose the Kind: Category, P&L line, Metric or Column.
  2. Enter the Code: segment_1 to segment_10 for a category, the line’s key for a P&L line, the metric’s name, or table.column for a column.
  3. Enter the Name, at most 60 characters on one line.
  4. Click Save names.

Saving replaces the whole list. A row with an empty name removes that name, and Remove selected removes the rows you tick. Naming the same code twice is refused. Business names are also written into the analysis path’s description so the assistant uses your words, and that description has a length limit: if a name would make it too long, nothing is saved and you are told.

Explain this number shows what makes up one profit and loss figure. On the Readiness grid, a row for an entity and month has Explain this number, which opens the entity’s operating margin for that month. The window splits the number into layers that add up to the total at the top:

Layer What it holds
From the general ledger What the ledger posted to the line.
Allocated in Cost spread to this entity by your allocation rules.
Allocated out Cost this entity spread to others.
Eliminated in the group Intercompany that was eliminated in the group result.

Each layer has a sentence saying what it is and its largest source documents, with their source system, document and account. Only layers that have rows are shown. The window follows your row access: you explain only the figures you can see. A number with nothing behind it says Nothing to explain.

Allocations spread shared costs, such as overhead, to departments, products and customers. The model has two stages: Stage 1: overhead to departments and Stage 2: departments to products and customers. Open Actions > Allocations… to see or change how each is done. Without changes, they work as they did before the wizard existed. A reader can open the rules but not save them.

For each stage the wizard asks four questions:

Step Question Choices
Pool Which costs do you want to spread? The kind of account, and for stage 1 the department that holds the cost, or All departments.
Driver What do you want to spread them by? Statistic you loaded as allocation drivers, such as headcount or square feet, Net sales amount, Sales quantity or Spread evenly. Use the driver of Same period or Same period last year.
Target Where do they land? Stage 1: departments. Stage 2: products, customers or both.
Method What happens when a driver is negative? Ignore negative values or Stop and tell me.

Under the steps, Preview for a month, written as year and month such as 2026-09, shows what the rules, including ones you haven’t saved, would do: Pool 120,000 · Allocated 120,000 · Unallocated 0. Anything unallocated is cost that had no driver to take it. The preview follows your row access.

A department or statistic name can’t contain a single quote ('), a backslash or a line break, because the name is written into the allocation steps’ filters. Save and a preview of stage 1 refuse one that does and name it.

Click Save. The rules take effect at the next Load, or when you click Apply changes. If you edited an allocation step by hand, your edit is kept, not overwritten, and the save names it. You can’t save while a load is running.

A board pack writes your results to a folder in a document account, ready to send. Open Actions > Board pack…, choose a Document account and a Folder, such as /board-packs, and click Save. Every Load then makes the pack, so the files in the folder are as current as the last load. Each run writes:

  • An Excel workbook, Board pack.xlsx, with the profit and loss by entity, a margin bridge, and top and bottom customers.
  • A PDF of the profit and loss dashboard.

Click Run now to make a pack immediately; its files appear in the folder when it finishes. The dialog shows the last run and its outcome, and the next run. Apply changes doesn’t make a pack. To make one every night, set up the nightly refresh (see Keep It Current); until then the dialog says Not scheduled yet. Only someone who can change the project sets it up.

A pack that can’t be made in full never fails the load:

  • If reporting is turned off for your workspace, the pack is the Excel workbook alone, and the run says Reporting is turned off for this workspace, so the board pack has the Excel workbook only.
  • If a dashboard someone else made already has the address Data Foundation needs for the profit and loss dashboard, that dashboard is left as it is and the PDF is not made. The message names the address; give the other dashboard another address, then run Load again.
  • If any step of the pack fails, the Status list shows an amber Board pack row naming the step and why. The load itself still succeeded.

A restore point is made as each Load or Apply changes starts, whoever starts it, including a scheduled load, and before each change to mappings or sources. It holds the model’s tables, the dimensions, and the settings the next load builds from: your sources, files and allocation rules. The last seven points taken for loads and the last five taken before changes are kept, so changes never push out the points taken for loads. Older points roll off.

To go back, open Actions > Restore…. Each point shows when it was made and, beneath it, how many tables restoring it brings back, for example Puts back 12 tables., along with any warnings. Choose a point and click Restore. Restore is offered once a point exists. Restoring first records the project as it is now as a point of its own, so you can go back again. It then puts the tables, the dimensions and those settings back as they were, and the next load builds from them. Restoring to the sources you already had doesn’t mark them as changed. A view, a dimension that can’t be saved, or a table with no saved copy is named in a warning, and the rest is restored. A restore point that has rolled off is refused, and you can’t restore while a load is running.

Actions > Clone… copies the project under a name you give. The copy keeps the template, your edits to the model, and your metrics and analysis paths, and loads the sample data. Its sources are not connected, so connect them in the copy before you load. You can keep working while the copy is made, and Open project takes you to it when it’s ready.

Save a project’s setup to reuse it for the next client or entity. Open Actions > Save as my template…, choose the document account, and name the file. A name ending in .zip is added if you leave it off. Saving over an existing file is refused, so choose another name.

The file holds the project’s structure, your edits to the model, metrics and analysis paths, the business names for entities and accounts, and the rows of the template’s configuration tables, such as the activity map of Cost to Serve or the policy ranges of Operational Transfer Pricing. It leaves out connections, files, the nightly schedule, runs, restore points, the board pack and who has access.

To use it, create a project, choose From a template file… under Start from, and pick the document account and path. The new project is private to you, with every source waiting to be connected. A release of the template later merges its changes around steps you edited. If there is no file at the path, nothing is created and you are told. Anyone who can read the document account can install from the file. To use it in another workspace, download the file and upload it there.

After a load, a project can still be waiting to be told three things: accounts the chart leaves without a type, codes more than one source uses, and columns in your files that matched nothing. Ask the assistant to resolve them: from an AI assistant over MCP, use the Resolve Data Foundation residue prompt. It reads what is waiting and proposes one answer for each item, with a reason.

The assistant writes nothing. You apply its answers where you would have answered yourself: accounts and codes under Codes waiting for your answer, and file columns in the file’s mapping. It never proposes an entity.

The Setup checklist has four closing items after Review your hierarchies:

  • Set up the nightly refresh is done once Refresh nightly is on.
  • Share with your team is done once someone besides the person who set the project up is a member.
  • Download the evidence pack is done when you download the pack after the last successful load.
  • Hand the project over to its owner is done once an Architect other than the person who set it up exists and the totals tie out. While it isn’t done, the checklist says what to do: make the client’s owner an Architect of this project, then remove yourself from its members.

The checklist says who set the project up (Set up by).

A template builds its reports and models on the Common Finance Model, so they work the same whichever ERPs feed it. When you start a project from a template, such as Enterprise Profitability, it creates:

  • The Common Finance Model tables, empty or filled with sample data.
  • Entity, Account, Department, Customer, Product, Segment 1 and Segment 2 dimensions built from those tables. See Hierarchies.
  • The Load workflow and the workflows it runs to read your ERP and files, conform them, check readiness and build the template’s models.
  • The template’s own reports and models. Enterprise Profitability adds a profit and loss table with revenue, COGS, gross margin, direct and overhead operating expenses, contribution and operating margin, plus margin bridges year over year and against budget.
  • Metrics and an analysis path for asking questions about the results.

Three templates are available under Start from, each building on the same model:

Template Adds
Enterprise Profitability The profit and loss and margin bridges.
Cost to Serve The cost of serving each customer, and margin after it. See Cost to Serve.
Operational Transfer Pricing Each entity’s profit level against your policy range, and the true-up. See Operational Transfer Pricing.

From a template file… starts from a project you saved. See Save as My Template.

To explore before you connect your own ERP, tick Start with sample data when you create the project. The project runs a first Load straight away, so the Foundation tab shows a complete example, including a euro entity translated into US dollars. When you connect your own ERP or add your own files, the sample is removed and your data takes its place. If the removal is interrupted, the next Load finishes it. A folder of your own named sample is never removed: the status page notes that the sample holds that folder, so the other tables in it are not read, and asks you to move them to a folder of another name.

The Foundation tab is where you connect, load and check your data. A rail on the left lists four pages:

Page What it shows
Status The state of the project: Last load with the Load, Refresh and Refresh nightly buttons, What to do next, the Setup checklist, Data loaded, Periods, Readiness and the Actions menu.
Answers needed The questions waiting for you, such as codes more than one source uses, where you say whether they mean the same.
Source coverage Where sources overlap or leave gaps for a company and month, and which source counts from when.
Sources Each source in Load order, with Edit and Remove this source, the Group and Names cards, and Connect a system (or Connect another system), where you choose a kind of system, pick its connection, choose the companies and how many months of history to load, or add a file. The files you added are in the Files card.

Badges on the rail count what needs attention: the open questions on Answers needed, the coverage issues on Source coverage and the sources on Sources. Status shows no badge. The result of a Load or Apply changes and the status banner show above whichever page is open, so you see them on every page. The buttons on Status that point at a question or a coverage issue, such as Answer codes › and Choose sources ›, take you to that page. Closing Answers needed or Source coverage returns you to Status.

When you change your sources, a bar above every page says You have changed your sources. Press Apply changes to use them. It has Show what changed and Apply changes, and it is the one place to apply a change. With nothing waiting to apply, Load is the main button on Status. On Status, What to do next lists problems first and then items that need attention, numbered through both groups. On Answers needed, Answer all offers Accept suggestions, All the same and All different. On Source coverage, Save choice keeps which source counts from when.

The Setup checklist is: Connect your data, Load your data, Check that totals tie out, Review your hierarchies, Set up the nightly refresh, Share with your team, Download the evidence pack and Hand the project over to its owner. The current item is in bold. Business names, the files you added and their mapping state (✓ Confirmed or ○ Waiting for confirmation), and the readiness grid with one row per entity, period and check and its What to do, are on the same pages. See Business Names.

While a load runs, the tab refreshes on its own and keeps reading even if one read fails. If the status can’t be read, The Foundation status could not be read. appears with Try again, which reads it afresh; the tab also retries once by itself before showing the message, and the message clears when the server answers.

The Status page also notes:

  • A folder under /sources/ whose tables lack the columns they need isn’t added as a source. A note names each table that was left out and the column it lacks. Add the column, then press Load or Apply changes.
  • If a load succeeded but brought in no ledger lines to compare, an amber Model P&L ties to GL row says nothing tied out. Check that totals tie out stays pending on the setup checklist, not done, until something has tied out. Add a source with ledger lines, then press Load.

The setup checklist tracks your progress:

  • Connect your data and Load your data tick themselves once a source is attached and a load succeeds.
  • Check that totals tie out ticks itself when the tie-out checks pass.
  • Review your hierarchies: click Open dimensions to review the Entity, Account and Department hierarchies, then tick I have reviewed them. You confirm this one yourself.
  • Set up the nightly refresh, Share with your team, Download the evidence pack and Hand the project over to its owner close the list. See Hand the Project Over and Managing Project Access.

The Actions menu holds Allocations…, Board pack…, Clone…, Restore… and Save as my template…. See Allocation Rules, Board Pack, Clone a Project, Restore Points and Save as My Template.

Data Foundation conforms every source, ERP or file, into the same fifteen tables, so reports, metrics and models built on them work the same whatever the source. Each row also records its source system and when it was read. The Dataset column is the name you choose when you add a file.

Dataset Table Holds
Entities cfm_entity Legal entities with parent, local and group currency, and country.
Entity crosswalk cfm_entity_map Source entity codes and the entity each belongs to. See Entity Crosswalk.
Currency rates cfm_fx_rate Monthly average, closing and historical rates by currency pair. See Exchange Rates.
Fiscal calendar cfm_fiscal_calendar One row per entity fiscal period, with start and end dates and whether it is closed.
Chart of accounts cfm_account Accounts with parent, type (Revenue, COGS, Opex, NonOperating, Asset, Liability, Equity), statement, normal balance and rate type.
Departments cfm_department Departments and cost centers with parent and profit center.
Products cfm_product Product master with parent, family, group, unit of measure and standard cost.
Customers cfm_customer Customer master with parent, group, region and country. Customer names are not masked.
Segment values cfm_segment Segment values with name and parent. See Hierarchies.
General ledger cfm_gl_journal Journal lines in transaction, local and reporting currency.
Trial balance cfm_trial_balance Period debits, credits and net by entity, account and department, in local and reporting currency.
Sales lines cfm_sales_line Invoiced sales lines with quantity, gross, discount, net and COGS amounts.
Inventory balances cfm_inventory_balance Period-end inventory by product and warehouse.
Budget cfm_budget Budget and forecast by scenario, version, account and department.
Allocation drivers cfm_stat_driver Statistical drivers such as headcount, square feet or units.

A Foundation project comes with project elements for the columns these tables share: entity, account, department, profit_center, product, customer and period. Each carries a description, and a type and dimension where one applies, so every table that has one of these columns shows the same meaning. Data Foundation adds any of them the project lacks and leaves one you have edited as it is. This holds for every ERP and file source.

Periods are written YYYY-MM, and transactions are placed by posting period. Budget is optional: an ERP without budgets still loads everything else.

A trial balance file can carry debit and credit columns, a net column, or both. With only debit and credit, the net is debit minus credit, and a blank side counts as zero. With only a net, a positive net becomes the debit and a negative net becomes the credit, as a positive amount. A department column is optional.

Symptom What to do
Test connection says the ERP refused the credentials Check the connection’s credentials, and that the integration user can sign in to the ERP.
Test connection shows Missing permission for a dataset Ask your ERP administrator to add the permissions it names. The connection guide for your ERP lists what the connection needs.
An ERP error message is shorter than the ERP’s own PlaidCloud removes credentials, tokens and keys from provider errors before showing them, so some detail is left out. Run Test connection to see which dataset is affected.
Data loaded is a problem for a period An ERP read for that period failed. Run Load again; periods that loaded are skipped.
Accounts missing from chart is a problem Add the missing accounts to the chart of accounts in the ERP or your file, and run Load again.
Currency translation is a problem An entity’s currency has no rate into the reporting currency for that period, or no single reporting currency was found. Add the rates to a Currency rates file, check that the connection can read the ERP’s exchange rates, or set the project variable reporting_currency. Then run Load again.
Currency translation needs attention Balance sheet, inventory or budget rows have no closing, historical or planned rate for the period. The profit and loss is not affected. Add the rates, including planned rates for budget months, and run Load again.
Codes waiting for your answer needs attention Click Answer codes ›, answer each code, and click Apply changes.
Which source covers which months needs attention Click Choose sources ›, pick the source, click Save choice, and click Apply changes.
Intercompany balances is red or amber Check the pairs in Intercompany pairs. Book the missing side, add the trading partner to the lines, or add the missing rate.
Group views are off Ask an Architect to tick Everyone on the project may see the group results under Group.
Unmapped entities is a problem Add the entity to the entity master, or map its code to an existing entity in the Entity crosswalk file, and run Load again.