SolutionsPricingWorkshopsBlogNewsAboutLet's Talk

Automatic Reporting System: How to Build and Deploy

An automatic reporting system pulls data from all sources, processes it, and distributes reports on schedule. A practical guide to building with n8n and APIs.

AutomushPublished 12 min read

An automatic reporting system pulls data from multiple sources, processes it according to predefined rules, and distributes the results on a fixed schedule or in response to a specific event. Instead of manually exporting information from each system, copying it into a spreadsheet, updating formulas, and sending by email, the system runs the entire process without human intervention.

We build these systems around n8n, the WhatsApp Cloud API, connections to CRMs and invoicing systems such as Morning (Green Invoice), and spreadsheets as a data layer or presentation layer. The system runs on our server or in the cloud, under continuous monitoring that detects failures and alerts in real time.

Why build an automatic reporting system instead of doing it manually?

An automatic reporting system saves time, eliminates copy errors, and ensures that every decision-maker sees the same information at the same time. A report built manually once a week becomes outdated the moment it goes out, and the preparation process takes between one and three hours of repetitive work.

When the system is automatic, it runs every day at a set time, or even every hour, without occupying anyone. Data is pulled directly from the CRM’s API, from the invoicing system, from a Google Sheet where information from other sources is centralized, or from any other source that exposes a programming interface. If there is an error in the source – for example, an empty field that should be filled, or a value that does not pass validation – the system alerts instead of sending an incorrect report.

The information is processed according to business rules: filtering by status, calculating totals, comparing to a previous period, classifying by region or product. The result is sent by email, uploaded to a shared folder, pushed to a Slack channel, or sent as a WhatsApp message via the business API. Each recipient sees only the part relevant to them, and the report arrives in a uniform format that does not change between runs.

What data can be pulled into an automatic reporting system?

An automatic reporting system can pull any data accessible via API, webhook, or scheduled export. The most common sources we connect are CRMs (lead, deal, and customer data), invoicing systems such as Morning that provide invoice and revenue data, spreadsheets that serve as a temporary database or manual collection layer, and support or ticketing systems.

When the source exposes an API, n8n pulls the data directly with an HTTP request. When there is no API but there is an option for automatic export to a shared folder or email, the system reads the file, parses it (CSV, Excel, JSON), and processes the content. If the source does not support either of these options, it is possible to build an intermediate layer: a spreadsheet into which data is entered manually or by copy, and the system reads from it.

It is important to remember that every data source comes with rate limits. The WhatsApp Cloud API, for example, allows a limited number of requests per second, and many CRM systems limit the number of records that can be pulled in a single request. The system must handle this with scheduled delays, division into batches, and handling of 429 (Too Many Requests) errors with retry.

How long does it take to build an automatic reporting system?

Building and deploying an automatic reporting system takes between two weeks and six weeks, depending on the number of data sources, the complexity of the calculations, and the readiness of existing systems. A simple report that pulls data from one source, calculates a sum or average, and sends by email – that is two weeks. A report that combines three sources, performs field matching, calculates performance metrics, and distributes different versions to different groups – that is four to six weeks.

The first stage is mapping: what data needs to appear in the report, where it comes from, what processing is required, and who the report is for. We examine the API of each source, check which fields are available, and identify gaps – cases where the data does not exist in the source system and needs to be supplemented another way.

The second stage is building: setting up a workflow in n8n that pulls the data, processes it (filtering, calculation, table joining), and produces the result. We build in a development environment, run on dummy data or on real data in a limited scope, and check that the result matches the expectation.

The third stage is deployment: moving to a production environment, setting a schedule (cron job), connecting to real distribution channels, and running the first time under supervision. We monitor the first run, make sure the report arrived, that the data is correct, and that there were no errors.

The fourth stage is monitoring: the system reports on every run – successful or failed – and alerts when something is wrong. We receive these alerts and handle them in real time.

Stage Duration Deliverable
Mapping and analysis 3–7 days Requirements document, list of sources and fields
Building and testing 7–20 days Active workflow in development environment
Deployment and launch 2–5 days System running in production, first report sent
Monitoring and adjustments Ongoing Stable system, fixes as needed

How are errors handled in an automatic reporting system?

An automatic reporting system must handle errors in a controlled manner, because failures happen: the API is unavailable, a field is empty that should be filled, a file did not arrive on time, or a rate limit was triggered. The system must identify the problem, try again if it makes sense, and alert if the failure repeats.

In n8n we build error handling with Error Trigger and Stop and Error nodes. When an API call fails, the system checks the error code: if it is 429 (Too Many Requests), it waits and tries again. If it is 401 (Unauthorized), it means the token has expired and needs to be refreshed or alert. If it is 500 (Server Error), it is a failure on the source server side, and the system tries again after a minute.

When the data itself is problematic – for example, an empty amount field or a negative value where that is not possible – the system applies validation checks. If the check fails, the record is marked, and the report includes a warning or is omitted entirely, depending on the policy that was defined.

The alerts themselves are sent through the same channels used in the business: a WhatsApp message via the business API, a message in a dedicated Slack channel, or email. The alert includes the workflow name, the exact error, and the time it occurred. We receive these alerts and usually handle them within an hour.

When the system cannot send the report itself – for example, the email address is invalid or the mail server rejected the message – it saves the report in a local folder and alerts. That way the data is not lost, and it can be sent manually or the address can be corrected and run again.

Should you build on spreadsheets or on a database?

An automatic reporting system can use spreadsheets (Google Sheets) or a database (PostgreSQL, MySQL) as an intermediate storage layer. The decision depends on the volume of data, the required speed, and who needs direct access to the information.

Spreadsheets are suitable when the volume is small (up to a few thousand rows), when there is a need for occasional manual editing, and when business stakeholders want to see the raw data without entering a separate system. Google Sheets exposes a simple API, n8n connects to it directly, and you can build formulas within the sheet that calculate additional metrics. The downside is performance: reading and writing thousands of rows takes time, and there is a limit on the number of requests per minute.

A database is suitable when the volume is large (tens of thousands of rows and up), when complex queries are needed (JOIN, GROUP BY, subqueries), and when speed is critical. PostgreSQL, for example, easily handles millions of records, and n8n connects to it with a Postgres node. The downside is that business stakeholders cannot see the data directly – you need to build them an interface or export to a spreadsheet.

In practice, we combine: raw data is stored in a database or pulled directly from the API at runtime, and the final result is written to a spreadsheet that serves as a view or sent directly as an attached file. That way you get the advantages of both worlds.

If you are building a reporting and BI system that combines many sources and produces multiple reports, it is worth investing in a central database and building the spreadsheets as a presentation layer only.

How do you define who receives which report?

An automatic reporting system needs to know who receives which report, in what format, and at what frequency. This definition is stored in a separate table or spreadsheet that contains all the rules, and the system reads from it on each run.

Example: a table with the following columns – report name, email address or WhatsApp number, filter (for example, only deals from the central region), format (PDF, Excel, text message), and frequency (daily at 08:00, weekly on Sunday). The system goes through the table, checks if it is time to send, applies the filter, generates the file or message, and sends.

If someone needs to stop receiving a report, simply delete their row from the table. If you need to add a recipient, add a row. There is no need to touch the workflow itself, which means the business owner or manager can make the change without contacting us.

When more complex permissions are needed – for example, a regional manager sees only the data from their region, and a general manager sees everything – we build logic that uses a “role” or “region” field in the recipients table, and applies dynamic filtering on the data before sending.

How much does it cost to set up an automatic reporting system?

The setup cost of an automatic reporting system depends on the number of data sources, the complexity of the processing, and the number of reports. A simple system with one source and one report costs between $2,000 and $4,000. A medium system with three sources, data processing, and several different reports costs between $5,000 and $9,000. A complex system with five or more sources, advanced business calculations, and deep integration with existing systems costs $10,000 and up.

These costs include mapping, building, testing, deployment, and training. They do not include ongoing monitoring, which is provided as part of a monthly maintenance package.

Beyond the setup cost, there are monthly operating costs: storage on n8n Cloud (if not hosting on a private server), API calls to external systems that charge by usage, and our monitoring and maintenance hours. For an average system, monthly operations cost $400–$800.

It is worth comparing this cost to the cost of manual work: if someone invests three hours a week preparing reports, that is 12 hours a month. If their hourly rate is $40, that is $480 a month, or $5,760 a year – just in direct cost, not counting errors and delays.

If you want to understand how to calculate return on investment in automation, read our guide on automation ROI calculation.

What happens when you need to change an existing report?

An automatic reporting system needs to be flexible, because requirements change: you add a new field to the CRM, change the definition of “active customer”, or want to see the data in a different breakdown. A change to an existing workflow takes between an hour and a few days, depending on complexity.

A simple change – for example, adding a column to a report or changing a recipient’s email address – is one to two hours. We enter the workflow, add the new field to the query or table, check that it appears in the report, and run.

A medium change – for example, adding a new data source or changing calculation logic – is one to three working days. You need to understand the new source, build the connection, make sure the data fits with the rest, and check the result.

A complex change – for example, splitting one report into several different reports by role, or adding a new business calculation that requires joining several tables – is one to two weeks. This is already a small project, with re-mapping, building, and testing.

It is important to document every change: what changed, when, and why. We keep this documentation in a shared spreadsheet or task management system, so that if there is a problem you can go back and understand what changed.

How do you ensure the data in the report is correct?

An automatic reporting system produces data without human supervision, so it must include validation checks that ensure the result is logical. These checks run on every execution, and if something does not pass – the system alerts.

Common checks:

  • Totals: Is the total amount greater than zero? Is it unusually low or high compared to the previous period?
  • Counts: Is the number of records logical? If last week there were 200 leads and this week there are 2, that is probably an error.
  • Empty fields: Are there critical fields that are empty? For example, a customer without a name or a deal without an amount.
  • Dates: Are the dates in the expected range? A deal from 1990 or 2050 is probably an entry error.
  • Duplicates: Are there duplicate records? For example, the same customer appears twice due to a merge error.

If a check fails, the system can act in three ways: alert and send the report anyway with a problem marker, alert and not send the report, or try to fix automatically (for example, fill an empty field with a default value) and alert about the fix.

Beyond automatic checks, it is worth having someone in the business review the first report the system produces, compare it to the last manual report, and make sure the numbers match. This happens once, at deployment, and after that the automatic checks are sufficient.

Can reports be run on demand and not just on a schedule?

An automatic reporting system can also run on demand, not just on a fixed schedule. This is useful when you need an urgent report for a meeting, or when you want to check the current status without waiting for the next scheduled run.

In n8n you can run a workflow manually from the interface, or expose it via a webhook that allows activation from anywhere. For example, you can send a WhatsApp message with the command “today’s sales report”, and the system recognizes the message, runs the workflow, and returns the report within a minute.

Another way is a button in a shared spreadsheet or in a simple interface we built, that triggers the webhook when clicked. The user clicks, waits a few seconds, and the report arrives in their email or appears in the spreadsheet.

It is important to note that on-demand execution still consumes resources – API calls, processing time, and sending. If there is a rate limit on one of the sources, too frequent execution can reach the limit and cause failure. Therefore it is worth limiting the number of manual executions per day, or building a queue that delays a second execution if the first is still running.

Frequently asked questions

Can an automatic reporting system work with a CRM that has no API?

Yes, but through a workaround. If the CRM allows automatic export to a file (CSV or Excel) to a shared folder or email, the system can read the file and process it. If even that does not exist, you can build an intermediate layer: a spreadsheet into which data is copied manually each day, and the system reads from it. It is not ideal, but it is better than full manual report preparation.

What happens if one of the data sources is unavailable when the report runs?

The system detects that the source is unavailable (usually error code 500 or timeout), tries again after a minute, and if the second attempt also fails – alerts. Depending on the configuration, it can send a partial report with a note that a certain source is missing, or skip the send and try again on the next run. We receive the alert and check if it is a temporary failure or a problem that requires handling.

Can different reports be sent to different people from the same run?

Yes, and this is one of the main advantages of an automatic system. The system goes through a recipients table, applies filtering by role or region on the data, generates a dedicated report for each recipient, and sends. A northern region manager receives only northern data, a general manager receives everything, and each sees the report in the format they preferred – PDF, Excel, or text message.

How long does it take to run one report?

A simple report with one source and a few hundred rows runs in 10 seconds to a minute. A complex report with several sources, thousands of rows, and advanced calculations can take three to five minutes. The time depends on the speed of the sources' APIs, the volume of data, and the complexity of processing. If the report takes more than five minutes, it is worth checking if it can be optimized – for example, pulling only the necessary fields instead of everything.

Interested?reporting and business intelligence solutions

From the newsroom

All news

Which of your processes could run on its own?

A free 30-minute discovery call, no commitment. We map your processes together and tell you honestly whether and how automation can help.

Try a call with agent now