# Which business data should update automatically Source: https://baqueiro.es/en/automatizacion/business-data-update-automatically/ Published: 2025-11-19T09:18:00+00:00 Updated: 2025-11-19T09:18:00+00:00 Author: Baqueiro Desarrollo Web Language: en-US Usage: Attribution required: quoting allowed with a visible link back --- **TL;DR** - Copying data by hand between systems creates errors, delays and work that adds nothing. - The key is telling apart what should flow on its own and what needs human sign-off. - Automating master data (customers, products, prices) usually has the biggest impact. ## The invisible work that kills efficiency Every week, someone on your team copies data from one system to another. They update prices on the website because they changed in the ERP. They add customers to the CRM who already exist in the invoicing system. They reconcile stock between the physical shop and the online one. This manual work is invisible until it goes wrong. An outdated price means a lost sale. A duplicate customer, confusing communication. Incorrect stock, orders you cannot fulfill. The good news: most of these flows can be automated. The bad news: not all of them should be. Knowing which ones should and which should not is a strategic decision, not just a technical one. ## The single source of truth principle Before automating anything, answer one question: where does the official version of this piece of data live? - The ERP holds the official version of prices and stock. - The CRM holds the official version of contacts and the commercial relationship. - The website holds the official version of published content. Each system is the authority in its own area. Automation flows from the source of truth toward the systems that consume that data. Do not attempt two-way sync until it is genuinely essential: when 2 systems can edit the same data, conflicts are guaranteed. If every tool in your company currently works like an island, that map of sources of truth is the first step: without it, any automation just copies errors faster. ## Data that should almost always update automatically These are safe candidates for automation: **Product master data.** Name, description, SKU, categories, images. If they change in the ERP or the PIM (product information management system), they should be reflected automatically on the website, marketplaces and catalogs. **Prices and availability.** Critical in e-commerce. If the website shows a lower price than the real one, you lose margin or end up in a dispute with the customer, because the price you advertise is binding on you toward the consumer. If it shows a higher one, you lose the sale. Visible stock should reflect reality in near real time. **Customer data from the point of entry.** When someone signs up on the website or buys for the first time, that information should reach the CRM without anyone copying it. **Order status.** Received, being prepared, shipped, delivered. The customer should see every change on the website (and get it by email) without anyone updating it by hand. **Financial transactions.** Confirmed payments, issued invoices, collection status. If the payment gateway says “paid”, the ERP should know. ## Data that needs human validation Not everything should be automatic. Some data needs a person to give the go-ahead: **Data with high commercial impact.** Strategic price changes for key accounts, contract amendments, special discounts. Automation can apply them technically, but someone should confirm they are intentional. **Data with legal implications.** Changes to terms and conditions, the privacy policy or other legal texts. They can be published automatically once approved, but the approval must be human. **Data with ambiguous context.** A customer who “does not renew” in one system: have they churned or moved to another product? Blind sync can misclassify them and trigger, for example, a win-back campaign aimed at someone who is still a customer. **Data where quality is critical.** Product descriptions, marketing content, technical information. They can flow automatically from a content management system, but someone must check they are correct before they go live. ## The decision matrix: automatic or manual? To decide, we cross 2 dimensions: volume (how often it happens) and decision complexity (how much judgment it requires). - **High volume, low complexity:** automate. For example, syncing stock every hour. - **High volume, high complexity:** semi-automate. The system proposes and a person approves. For example, invoices that are generated automatically but reviewed before being sent. - **Low volume, low complexity:** it can be automated, but it may not justify the development and maintenance cost. - **Low volume, high complexity:** manual. Do not try to automate decisions that require judgment if they only happen a few times a month. If you want to put numbers on this decision before you start, work out how many hours a month the manual task takes today and what the errors it causes cost you, then compare that with the cost of building and maintaining the integration. ## Putting it into practice: sync patterns Once you decide to automate, there are several technical patterns to choose from: **Real-time sync (webhooks).** If the source system supports webhooks, updates are almost instant: as soon as the change happens, the sending system pushes the data to the receiver. **Scheduled sync (polling).** The receiving system checks for changes at set intervals (every 15 minutes, every hour, every night). It is simpler to build but introduces a delay. It suits data that does not need to be accurate to the second. **File-based sync.** Exporting and importing CSV, XML or JSON. It works well for large volumes or when there is no API available. It forces you to track what has changed since the last sync. **Message queue sync.** For scenarios involving many systems, a messaging infrastructure (RabbitMQ, Amazon SQS) is used, where each system publishes events and subscribes to the ones it cares about. It is more robust, but also more complex. Tools such as Make or Zapier handle the first 2 patterns well in many cases, as long as they are set up with care: with error handling, alerts and someone who knows what each scenario does. ## Conflict resolution What happens when a piece of data changes in 2 systems between one sync and the next? There are several strategies: **Single authority (recommended).** One system always holds the correct version. If there is a conflict, the authority wins. For that data, the other systems are read-only replicas. **Last write wins.** The most recent change prevails. It is simple, but it can lose legitimate changes if one system’s clock is off or if the sync is running late. **Manual resolution.** When there is a conflict, the record is flagged for human review. It is the safest option, but it needs an operational process to work through that conflict queue. **Conflict-free replicated data types (CRDTs).** Advanced techniques where changes can be applied in any order and the final result is always the same. They are complex and mostly used in real-time collaboration tools; syncing the systems of a small or medium business rarely needs them. ## Specific cases by sector **E-commerce.** Stock and prices from the ERP to the website: automatic. Product descriptions: semi-automatic (they flow from the PIM but are approved before going live). Orders from the website to the ERP: automatic, with exceptions handled manually. **B2B SaaS.** User data from the app to the CRM: automatic. Usage events for billing: automatic. Plan changes: semi-automatic (the system proposes and the account manager approves if there is a negotiation involved). **Professional services.** Customer data from the web form to the CRM: automatic. Project data from the CRM to the project management tool: semi-automatic, because the initial setup tends to vary from client to client. ## Frequently asked questions ### How much does it cost to automate data sync? It depends on the systems involved and the complexity. Between popular platforms (HubSpot, Salesforce, WooCommerce, SAP) you can often rely on existing connectors or on Make and Zapier, with costs in the hundreds of euros plus the subscription. Integrations with legacy or custom systems require development and usually run into the thousands. But not doing it has a cost too: hours of manual work every week, plus the errors it causes. ### What happens if the sync fails? Design for failure: logs, alerts when a sync does not complete within the expected time, and a manual contingency procedure. Never assume the integration will always work. The question is not whether it will fail, but when. And write down how it works: [a poorly documented automation](https://baqueiro.es/en/automatizacion/hidden-cost-poorly-documented-automations/) gets very expensive the day it breaks. ### Should I sync everything or only specific fields? Only what you need. Every synced field is a potential point of failure. If your CRM has 200 fields but the website only uses 20, sync those 20. ### How do I know my sync is working properly? With regular checks: compare a sample of records in both systems, set up an alert when the error rate goes above a threshold, and measure latency, that is, how long a change takes to propagate. ## Conclusion Data that flows between systems on its own frees up time for work that actually adds value. But automating the wrong sync, or doing it badly, can create more problems than it solves. The key is to be deliberate: which data, in which direction, how often and what happens when something fails. Neither “everything should be synced” nor “nothing can be automated”. **You might also like:** - [How to document technical decisions so they don’t get lost](https://baqueiro.es/en/ingenieria-sin-atajos/how-to-document-technical-decisions/) - [How to automate your ERP reporting without changing systems](https://baqueiro.es/en/codigo-que-trabaja-por-ti/automate-erp-reporting-without-switching-systems/) If you want to check which of your business data is out of sync between systems, or plan integrations that keep it consistent without manual work, Baqueiro can help. [Let’s talk](https://baqueiro.es/en/contact/).