Why Your CRM Data Is Only as Good as Your Data Foundation

A pattern from many CRM projects: a client calls about the CRM. Reps do not trust the forecast, the AI features underwhelm, adoption sags after month three. We review the tenant, and the CRM configuration is usually fine. What is broken sits underneath: duplicate accounts, three versions of the customer master, opportunity amounts that disagree with order intake, and nobody who owns any of it.

TL;DR: Most CRM data problems are foundation problems. Fix them in this order: decide one owner per attribute, repair the join keys between CRM and ERP, clean one domain and show the result, then let a platform such as SAP Business Data Cloud and SAP Master Data Governance carry the maintenance. Configuration alone will not solve it, and AI makes the cost of skipping it higher.

The symptoms are in the CRM; the cause is underneath

The forecast meeting is where it shows. Sales presents pipeline from Sales Cloud V2, finance presents revenue from the ERP, the two decks disagree, and the meeting turns into an argument about whose number is right instead of what to do. Most companies we meet have a version of this meeting.

The CRM gets the blame because the CRM is where people look. Trace any single disagreement, though, and you land on foundation issues: an account that exists twice because two subsidiaries onboarded it separately, a won opportunity that was never matched to its order, a currency conversion done differently in two reports. None of that is a CRM setting.

Fix it in this order

Sequence matters more than tooling. What works, in order:

1. Ownership before cleansing. Decide who owns each attribute before anyone touches a record. Cleansing without ownership is mopping with the tap open.
2. Join keys before completeness. Perfect address data on records that cannot be joined to their orders is decoration. The keys (account to business partner, opportunity to order, product to material) are what analytics and AI stand on. Fix those first.
3. One domain before the landscape. Clean accounts and opportunities against orders, ship the pipeline-to-cash view, and hold the meeting where both decks carry one number. That result funds domain two. A two-year cleansing programme with no visible output funds nothing.
4. Platform before heroics. Foundations decay through maintenance: the person who built the pipeline leaves, releases break it, entropy wins. This is the burden SAP Business Data Cloud is built to take on, with data products that SAP maintains through releases, so quality work you did stays done. The integration patterns post covers the mechanics.

Who owns what: a starting point

On the integrated SAP stack, the standard integration between Sales Cloud V2 and S/4HANA Public Cloud already settles many of these decisions, which is one of its underrated benefits. A split that works for most mid-market clients:

Attribute	Owner system	Direction
Sold-to customer, billing address, credit, payment terms	S/4HANA (business partner)	ERP to CRM, read-only in CRM
Materials, prices, conditions	S/4HANA	ERP to CRM
Contacts, roles, relationships	Sales Cloud V2	CRM only, or CRM to ERP where needed for documents
Leads, opportunities, activities	Sales Cloud V2	CRM to analytics
Orders, deliveries, invoices	S/4HANA	ERP to CRM for visibility, to analytics for reporting
Prospects without a sales history	Sales Cloud V2	Created in CRM, become a business partner at the first order

The last row is where most duplicates come from. Agree when a prospect becomes an ERP customer and who checks for an existing record at that moment.

Where SAP’s master data tools fit

Ownership rules need enforcement, and SAP has been moving here. SAP Master Data Governance governs master data and its policies; in May 2026 SAP announced it as a core component of Business Data Cloud, together with SAP HANA Cloud (SAP News). In the same month SAP completed the acquisition of Reltio, a master data management provider whose AI-based entity resolution merges related records into one reliable record. SAP says Reltio will become a core capability within BDC and should help customers “unify, cleanse and harmonize data across sources” (SAP News, March 2026).

For a CRM team, that addresses the duplicate-account problem at its source instead of in every report. Two caveats: several of the announced capabilities were planned for general availability in the second half of 2026, and the commercial model varies. Check what your contract and region include before you design around it. The sequence above still applies; tooling makes ownership enforceable, it does not decide it for you.

The mid-market angle

Large enterprises solve this with a data governance office. A 200-person Swiss distributor cannot and should not. The realistic mid-market model is one named owner per domain (a person, not a committee), the standard SAP data products as the maintained backbone, and modelling effort only in the last mile where your industry logic lives. Once it stands, that model runs on a fraction of a full-time role, which is why the platform choice matters more in the mid-market than in the enterprise: there is no team to absorb what the platform does not do.

It also helps when the partner who configures the CRM also knows the ERP side. Most foundation problems sit exactly at that seam, which is why we deliver Cloud ERP and CRM with one team.

Why this is urgent now

For years, weak foundations cost reporting time: annoying, but survivable. AI changes the cost curve. Every AI feature you switch on, from lead scoring and forecasting to Joule in Sales Cloud V2, reads the foundation and amplifies it: clean data in, leverage out; noise in, confident noise out. Switching on AI features over an unowned foundation means paying AI prices for amplified errors.

If you are starting from zero, read what BDC is first. The order of operations stands on its own: ownership, join keys, one domain, platform. The CRM was rarely the problem.

For Swiss companies starting this clean-up, our page on SAP CX consulting in Switzerland shows how we combine data foundation and CX delivery in one team.

FAQ

Where should CRM master data live: CRM or ERP?

The ERP owns the commercial truth: customers with credit and billing data, materials, prices. The CRM owns the relationship layer: contacts, activities, opportunities. Problems start when both sides maintain the same attribute; the working rule is one owner per attribute, replicated read-only in the other direction.

Do I need to clean all data before starting a BDC project?

No. Pick the one domain your first model needs (usually accounts and opportunities against orders), clean the join keys there, and let the visible result fund the next domain. A full-landscape cleansing programme with no visible output for a year is how data initiatives stall.

How does bad CRM data show up in AI features?

Directly. Lead scoring trained on duplicate accounts learns noise, forecasts built on stale opportunities inherit optimism, and a Joule summary of a fragmented account history is a fragmented summary in fluent language. AI does not clean data; it amplifies the quality underneath.

Does SAP offer master data tools for duplicate accounts?

Yes. SAP Master Data Governance governs master data and its rules, and in May 2026 SAP made it a core component of Business Data Cloud. SAP also completed the acquisition of Reltio, whose AI-based entity resolution merges related records into one golden record. Check which of these capabilities are available for your contract and region before planning around them.