In short
CRM (customer relationship management) refers to the systematic management of all of a company's relationships with its customers and to the software that maps these relationships. A CRM system stores contacts, interactions, quotes and orders in one place and makes them usable for sales, marketing and service. The aim is to understand customers better, retain them for longer and look after them more profitably.
The term is used in two ways in everyday life: once for the strategy, once for the tool. In the strategic sense, CRM describes a company's consistent focus on its customers. In the technical sense, CRM means the software with which teams maintain customer data and control processes. Both levels belong together: a CRM system without clear processes remains an address database, and a CRM strategy without a system remains an intention. A detailed definition is provided by Wikipedia on customer relationship management.
What a CRM system does
A modern CRM system covers five functional areas that build on one another. Not every company needs all five from the start, but the data basis should be set up so that the other areas can be added later without a migration.
Contact and company management
The basis of every CRM is the customer record. This includes people (contacts), organisations (companies or accounts) and the relationship between the two. A contact can be assigned to several companies, and a company can have subsidiaries. For each record, the system collects the history: emails, calls, appointments, notes, support requests. Data quality is decisive: duplicate checks, mandatory fields and validation rules prevent the data from becoming unusable over time.
Sales pipeline and opportunity management
The pipeline maps the path from the first contact to the deal in defined stages. In HubSpot the object for this is called a deal, in Salesforce an opportunity. An example from HubSpot: a deal pipeline has the stages "appointment scheduled", "qualified", "quote sent", "negotiation" and "won" or "lost". Each stage is assigned a probability of closing, from which the system calculates the weighted forecast. Sales managers can thus see how much revenue can realistically be expected in the quarter and where deals are getting stuck.
Salesforce separates the object model somewhat more strictly: a lead is a prospect who has not yet been qualified. On conversion, an account, a contact and an opportunity are created from it. This model enforces discipline because unqualified records do not dilute the actual pipeline. HubSpot, on the other hand, works with contacts and companies from the outset and uses a lifecycle stage instead of a separate lead object.
Service and ticketing
The service area manages requests after the purchase: complaints, technical questions, queries about invoices. Tickets are assigned to a contact, prioritised, distributed to teams and measured against service level targets. The added value compared with a pure helpdesk is that the service employee sees the sales history and, conversely, sales knows about open tickets before raising the next upsell.
Marketing automation
Marketing automation uses the CRM data to trigger communication: welcome sequences after a newsletter sign-up, follow-up emails after a download, reactivation emails for inactive customers. Workflows react to events (form completed, deal stage changed, order received) and to properties (industry, region, revenue class). Lead scoring evaluates contacts on the basis of their behaviour and only hands them over to sales once a threshold has been reached.
Reporting and dashboards
Reporting answers questions such as: which source delivers the customers with the highest customer lifetime value? How long is the average sales cycle? Where do we lose the most deals? Good CRM systems provide standard reports and allow custom analyses across all objects. For deeper analyses, many companies also export the data to a data warehouse or a BI tool.
Types of CRM systems
The literature distinguishes three types of CRM. In practice, the large platforms combine all three, but the distinction helps with the question of which focus a company needs first.
Operational CRM
Operational CRM supports day-to-day business: maintaining contacts, writing quotes, processing tickets, sending campaigns. It is what most users understand by "the CRM". Its benefit comes from automating recurring steps and from a shared view of the customer across departmental boundaries.
Analytical CRM
Analytical CRM evaluates the data collected: segmentation by purchasing behaviour, churn forecasts, basket analyses, customer value calculations. It provides the basis for decisions such as budget allocation, pricing or prioritising existing customers. Methodologically, the spectrum ranges from simple reports to machine learning models that predict churn risks.
Collaborative CRM
Collaborative CRM coordinates communication across channels and departments: telephone, email, chat, social media, field sales, partners. The aim is for a customer to get the same information regardless of the channel and not to have to explain the same thing three times. Technically, this involves integrating communication tools and a shared record to which all channels write.
Well-known CRM systems at a glance
The market ranges from tools for sole traders to platforms for corporations. The following table classifies five widespread systems by target group, strength and operating model. We deliberately leave out prices because they change frequently and depend heavily on the number of users and modules.
| System | Target group | Strength | Operating model |
|---|---|---|---|
| Salesforce | Mid-sized companies to corporations, complex sales organisations | Very deep data model, extensive customisability via Apex and Flows, large partner ecosystem | Cloud (SaaS) |
| HubSpot | Small and medium-sized companies, marketing-driven teams | Easy to use, marketing automation and CRM from a single source, free entry tier | Cloud (SaaS) |
| Microsoft Dynamics 365 | Companies with a Microsoft 365 environment, mid-market and enterprise | Close integration with Outlook, Teams and Power Platform, can be combined with ERP modules | Cloud, hybrid or on-premises |
| Pipedrive | Small sales teams, focus on the pipeline | Visual deal pipeline, quick introduction, low complexity | Cloud (SaaS) |
| Zoho CRM | Small and medium-sized companies, price-sensitive teams | Broad suite with many add-on modules, extensive functionality per licence | Cloud (SaaS) |
There are also open-source alternatives such as SuiteCRM or EspoCRM that can be self-hosted. They are interesting if data sovereignty is a priority or if deep changes to the code are planned, but they require operating know-how of your own.
CRM in e-commerce
In online retail, a large part of customer data is created in the shop and not in the CRM. Orders, baskets, returns and newsletter sign-ups run via the shop system, ERP and email tool. The CRM only becomes valuable when this data comes together. Together with the ERP, it forms the data core on which marketing and service work.
Connection to Shopware via API and webhooks
Shopware 6 offers two ways of getting data into a CRM. Customers, orders and products can be read and written via the Admin API. Via webhooks or the app system events, the CRM is actively notified as soon as something happens: a new registration, a completed order, a change of status. Ready-made Shopware extensions exist for HubSpot and Salesforce; for special requirements, such as B2B price lists or multi-tenant operation, a custom middleware is often the cleaner solution.
The direction of the data flow is technically important. As a rule, the shop is the leading system for order data and the CRM the leading system for communication and sales status. If fields are synchronised in both directions, clear rules are needed as to which system wins in the event of conflicts. Without such a rule, endless loops or overwritten data arise.
Customer data and order history
With the order history connected, a service employee can see in the CRM what the customer bought when, whether a return is open and how much revenue they have brought in so far. Sales in B2B recognises whether a customer has not ordered for months and can follow up in a targeted way. Marketing can base product recommendations on actual purchases instead of assumptions.
Segmentation
Segmentation divides the customer base into groups by characteristics: first-time buyers, regular customers, customers with large baskets, customers with many returns, inactive customers. A proven model is the RFM analysis by recency (last purchase), frequency (purchase frequency) and monetary value (revenue). The segments control campaigns, discount logic and service priority and are a central tool for customer retention.
CRM and AI agents
Since 2025, the major CRM providers have been opening their systems to AI agents via the Model Context Protocol (MCP). An MCP server provides CRM functions as tools that a language model can call, such as "Find all open deals with company X" or "Create a note on contact Y". How the interplay of agent, shop, ERP and CRM works technically is explained in our article MCP servers explained.
HubSpot: remote MCP server generally available since April 2026
HubSpot released its hosted MCP server to all customers on 13 April 2026. Authentication runs via OAuth 2.1 with PKCE, so no API keys need to be stored in agents. The server allows read access to CRM records, marketing content and activity history as well as creating and changing contacts, companies, deals, tickets, products and activities. Source: HubSpot Developer Changelog.
Salesforce: hosted MCP servers and Agentforce
With the Summer '26 release, Salesforce made hosted standard MCP servers generally available. According to the Salesforce Developer Blog, this allows MCP-compatible clients such as Claude, ChatGPT or Cursor to be connected to a Salesforce org. There are pre-built servers for SObject operations, Data 360 queries and Tableau analyses, and Agentforce agents themselves can also be provided as MCP tools. Agentforce is Salesforce's platform for custom AI agents that access CRM data, Flows and Apex.
What this means for companies
In practice this means: an AI agent in customer service can combine the order status from Shopware, invoice data from the ERP and the conversation history from the CRM in a single process. The prerequisite is the same as for any integration: clean data, clear permissions and a defined scope of what the agent may read and what it may write. Write permissions should be tightly limited and logged at first.
GDPR aspects of using a CRM
A CRM processes almost exclusively personal data. The requirements of the GDPR therefore apply in full, regardless of whether the system runs in the cloud or in your own data centre.
Data processing
With a cloud CRM, the provider processes data on behalf of the company. For this, a data processing agreement (DPA) under Art. 28 GDPR is mandatory. With providers based outside the EU, the question of third-country transfers is added: the legal basis can be standard contractual clauses or certification under the EU-US Data Privacy Framework. Many providers now offer EU data centres as an option; this does not replace the DPA but reduces the transfer risk.
Deletion periods and data minimisation
Customer data may only be stored for as long as a purpose exists. Retention periods under commercial and tax law apply to contract data, but there is no such basis for pure prospect data without a contract. A CRM therefore needs a deletion concept: automatic deletion or anonymisation of inactive contacts after a defined period, blocking instead of deletion where retention obligations are ongoing and a process for access and erasure requests under Art. 15 and 17 GDPR.
Consent for marketing
Promotional emails require consent, which must be documented verifiably in the CRM: time, wording, source, ideally via double opt-in. One exception is advertising to existing customers for similar products under § 7(3) of the German Unfair Competition Act (UWG), which is subject to strict conditions. The CRM should keep consents separately for each channel (email, telephone, post) and transfer withdrawals immediately to all connected systems.
Selection criteria for a CRM system
Selection rarely fails because of functions; it usually fails because of processes and acceptance. The following checklist helps clarify the requirements before comparing providers.
- Processes first: is the sales process described with stages, responsibilities and handovers before it is cast into the system?
- Data model: do the standard objects (contact, company, deal, ticket) fit the business, or are custom objects needed, for example for projects, contracts or devices?
- Integrations: are there ready-made connections for the shop system, ERP, email and telephony, and how well is the API documented?
- Automation: can workflows be mapped without programming, and are there limits on number or complexity?
- Rights and roles: can records be protected at team, region or field level?
- Data protection: DPA, server location, deletion concept, consent management and the ability to export your own data.
- AI capabilities: is there an MCP server or comparable interfaces, and how is agent access logged?
- Migration: how is legacy data imported, cleaned up and taken over with its history?
- User acceptance: how much effort does daily maintenance take, and are there mobile apps and email integration?
- Total cost: licence, implementation, training, integrations and ongoing customisation, considered over several years.
- Exit scenario: how completely can data be exported if a switch becomes necessary?
Common mistakes when introducing a CRM
The most common mistake is treating the CRM as an IT project. A CRM changes ways of working in sales and service; without the backing of management and without a recognisable benefit for users, data maintenance does not happen and the system becomes outdated. The second mistake is taking over unchecked legacy data: duplicates and outdated contacts poison the new database from day one.
The third mistake concerns integration. Anyone who reconciles shop, ERP and CRM manually via export and import creates inconsistencies and effort. An automated interface with clear data ownership per field is cheaper from the outset than cleaning up later. And finally: a CRM without defined metrics delivers reports that nobody reads. A few clear metrics such as pipeline value, win rate and response time in service are better than a hundred dashboards.
Frequently asked questions about CRM
What is the difference between CRM and ERP? A CRM manages the relationship with the customer: contacts, sales opportunities, communication, service. An ERP manages internal resources: warehouse, purchasing, accounting, production. In e-commerce the two interlock, because an order in the shop affects both the customer record in the CRM and the warehouse and invoice in the ERP.
Does a small company need a CRM system? As soon as more than one person maintains customer contacts or prospects are looked after over several weeks, a CRM is worthwhile. To get started, the free or inexpensive tiers of HubSpot, Pipedrive or Zoho are often enough. What matters is less the tool than the discipline to record every contact and every activity.
How can a CRM be connected to Shopware? Customers, orders and status changes can be transferred to the CRM via the Shopware Admin API and webhooks. Ready-made extensions exist for HubSpot and Salesforce; for individual requirements such as B2B price lists or several tenants, a custom middleware is usually developed that controls data ownership per field.
Can AI agents access a CRM directly? Yes. HubSpot has offered a hosted MCP server with OAuth 2.1 since April 2026, and Salesforce hosted MCP servers and Agentforce since the Summer '26 release. This allows AI agents to read, create and change records. Permissions should be narrowly defined and all agent access logged.