Head-to-head · updated 13 September 2026

DataShield vs Reltio (SAP): is master data management still a choice you make on its own?

Reltio is a good product. It runs multidomain MDM at a big scale. We do not match that scale. It covers customer, product and supplier data. Pfizer, Mars, and forty other brands you know all use it. Gartner made them a Leader in the 2026 Magic Quadrant for MDM. They ship real agent tools now. AgentFlow is one. It has named agents like Resolver and Unmerger. They also run their own MCP Server. Their hero line is "The future is agentic AI. The foundation is Reltio". They have earned the right to say it.

Then SAP bought them. The deal closed on 7 May 2026. Reltio is now "a core capability within SAP Business Data Cloud". Did you pick Reltio partly to stay off a big platform contract? That plan just changed. We sell a different shape of MDM. An agent interviews your business user. It builds the model with you. Then it generates draft objects in one step. The matching maths is open. You can read the weights yourself. Golden records only reach an agent after they clear a quality and consent gate. Each step lands on an audit chain. You can verify it yourself. Below is the honest split. That includes the rows where Reltio wins outright.

Get free Demo Center access Get an instant quote

The short version

Pick DataShield when

You would rather an agent built the model with you than wait on a services deal. Ours interviews a business user. It builds a draft spec. It walks you through a decision gate. Then it generates entity types, hierarchies, draft match configs, reference bindings and rules in one step. How the ontology works.
Your golden records have to live on hardware you control. Reltio is cloud only. We ship Docker images. We run self-hosted in your own cloud or data center. Or we run as a dedicated single-tenant server we operate.
You want to see how the match was scored. You want proof that no setup went live on a hunch. Ours is Fellegi and Sunter probabilistic linkage. It uses Jaro-Winkler, Levenshtein, Soundex and Double Metaphone. Nothing reaches production until it passes a frozen test battery through the real scoring path.
An agent will read those golden records. Someone will later ask you to prove what it was allowed to see. A golden must first clear six quality axes and a live consent basis. The audit log is a hash chain with signed checkpoints. Run the verifier.

Pick Reltio when

You need multidomain MDM across customer, product, supplier and location at once. You want the UI and the reference designs to match. This is their home ground. They are better at it.
You are an SAP shop. Reltio inside Business Data Cloud means one contract and one support line. It gives you a connection path we cannot offer.
You want named agent personas out of the box. Resolver, Unmerger, Profiler, Work Assigner and Address Enricher already exist. There is an Agent Builder for your own. We ship the surface those personas would call instead. That is 143 MCP commands for mastering alone, plus the guided onboarding engine. Their packaging is better today.
Most of your dark data is documents. AgentFlow Unstructured pulls PDFs and Word files into unified profiles. We do not go that deep.

Bottom line: Reltio is the broader suite. It has SAP's field team behind it. We are the better fit when the data stays in your building. We also fit when you want an agent to help build the model instead of a six-month programme. And we fit when each golden record an agent reads has to earn that read and leave proof.

Feature by feature

Competitor cells describe what Reltio's public site, press releases and SAP's own announcements say. This is as of the date above. If we have mischaracterised something, email support@myorg.ai and we will correct it, credited.

What matters	DataShield	Reltio	Edge
Standing up the model	An agent does it with you. The guided onboarding tool interviews a business user. It reads the datasets you already loaded. Then it builds a draft spec. You then work a decision gate, in order. Each proposal carries a what, a why, an impact and the alternatives. You confirm or override. An override propagates. So every choice that depends on it reopens for a fresh look. There is no accept-all button, on purpose. One single step then does the rest. It generates the entity types, hierarchies, draft match configs, reference bindings and rules. Each object lands with an honest gate state. The generator never promotes a config on its own. It never runs resolution either. That's 15 commands, and no implementation services.	Named agent personas ship ready to use. An Agent Builder covers your own. Those agents work the data once the model exists. Third-party summaries still put a Reltio implementation at six to twelve months. That is where the systems integrator comes in.	◆
Multidomain MDM	You model your own domain. We do not ship a template you have to bend to fit. There are three wired matching modes per entity type. A People, Places and Things taxonomy is a table row, not a fixed list. Hierarchies are scoped to a parent. A child config resolves inside its parent's scope at a base threshold. Across parents, it uses a stricter one. Relationships carry two-way time validity. Domains carry an owner and a compliance framework. Rules gate golden publication. What we do not have is a packaged Customer 360.	The category leader. Customer, product, supplier, location and more. Intelligent 360 sits on top. A Gartner MQ Leader placement in April 2026 backs it up.	◇
Entity resolution method	Fellegi and Sunter probabilistic record linkage. Comparators: Jaro-Winkler, normalised Levenshtein, Soundex, Double Metaphone, n-gram, date and numeric proximity. LSH blocking with a bucket cap. EM-trained m and u parameters run on an honest holdout. A trained polarity inversion is a hard refusal. There are five survivorship strategies. Most trusted reads measured overturn rates, not a hand-ranked source list. Population Stability Index drift monitoring proposes a retrain. It never acts on its own. An absolute-identifier tier routes identifier conflicts to review, instead of merging blind.	Strong and long-standing. They have shipped ML-assisted entity resolution for years. The scoring internals are not published in the detail ours are. So you cannot read the weights.	◈
Ingestion and binding	There are six on-ramps into mastering. Inline, URL and batch ingest. Storage watch sources. A resumable file repo. And a catalog bind of a live PostgreSQL table, with no rows leaving it. Bind an object to an entity type once. Then mastering stays current on a set cadence. You pick auto, agent-review or steward-review as the remaster policy. Differential loads mark deleted source records as stale. Updated ones go back for re-evaluation. A weekly full pass reconciles deletes. That is PostgreSQL today, not your whole estate. We are not an ETL product, and we will not sell you one.	A cloud platform with a long connection bench. Partners have wired it into SAP, Salesforce and the rest. We found no public detail on how a bound source stays current. We also found none on what review policy sits on the refresh. Worth asking.	◈
Stewardship and trust	Stewards review patterns, not items. A counterevidence scan keeps each proposal honest. Trust has three rungs. A pattern earns auto-disposition only after 50 approvals at a 98% rate, with zero dissents. One human call counts as one approval, however many items it covers. A single dissent demotes it. At the top rung, a deterministic sampler sends 10% of auto-disposed items back to a human. So you can prove why any item was sampled. Standing rules are born staged. They need a second release choice to go live. A blast-radius dry run uses the writer's own gate code. So a preview cannot say yes and then refuse.	A stewardship UI business users will sit in all day. Bulk review and resolution of match conflicts shipped since the 2025.1 release. A Work Assigner agent routes the queue. Their screens are better than ours.	◈
Agent admission gate	A golden record must earn the right to be read by an agent. It needs a composite score of at least 0.70 across six quality axes, with published weights. Each axis must clear its own floor. It also needs a live consent basis. Consent is a hard veto and fails closed. It covers the six GDPR Article 6(1) bases across nine purposes. No recorded basis means no permission. One rule enforces the gate in both the SQL and the app path. The golden consumption view masks data in the view itself, not in a setting. The reader role has no grant on the base table.	Role-based access control, plus data masking rules added in the 2025.1 release. We found nothing in their public docs that gates an agent read on data quality. Nor on a lawful basis.	◆
Deployment	Self-hosted in your own cloud or data center, or a dedicated single-tenant server we operate. Docker images ship for Auth, Ontology, Corpus and Lighthouse. Guardian verifies a signed deploy manifest.	Cloud only, on AWS, Google Cloud and Azure. It is now delivered through SAP Business Data Cloud too. We found no self-hosted or on-premises option in their public docs.	◆
Audit evidence	SHA-256 hash chain with Ed25519-signed checkpoints, and the checkpoints are themselves chained. Verification returns clean, attested damage, or tampered, and names the break. Every merge, unmerge, pin and override also lands on a before-and-after trail. It happens in the same step as the change. Ed25519 audit-signing keys stay on your instance, in a key directory you control. Kid-keyed rotation fails closed on an unknown key. KMS or HSM custody is a built seam in Auth, not a shipped connection. Try the verifier.	AgentFlow lists audit logging and role-based access control. We found no published cryptographic tamper evidence. Ask them whether a log entry can be silently edited.	◆
Agent authorization	Every governed tool call passes three checks before dispatch. A scope ceiling check. An authority tier check. And a fresh access check. Then it hits metering, then a sealed audit write. The call fails closed. Delegation is RFC 8693 token exchange, with an enforced scope ceiling.	Role-based access control on the platform. Plus the Reltio MCP Server, for "real-time, secure access to governed data". We found no per-call decision point in their docs. Nor mid-session revalidation.	◆
Break-glass	Scoped, time-boxed emergency access for agents. It is gated by admin role, IP allowlist and step-up. It auto-revokes, and nobody can quietly delete it from the log. Mastering has its own emergency brake too. Guarded operations run a canary before they ramp up. Scope-keyed pause and kill flags stop a single run. A global kill stops the whole fleet.	Not described in their public docs.	◆
GDPR erasure	Crypto-shred of per-subject key material. Plus ISO 27560 consent receipts. And a consent-validity record per subject and purpose, sitting under the masked golden views. Only the derived consent score crosses the masking boundary. So the facts themselves never reach the view reader. Actor identities in the chain are HMAC-committed. So the audit still verifies after the subject is gone. Golden version history survives deletion as a tombstone, with its reason attached.	Their 2025.1 release added data masking rules for PII inside the platform. The erasure mechanism, and what it does to history, is not spelled out.	◆
Tokenization	Datasets are tokenized at ingest. Agents query tokenized data over MCP. Detokenization is a privileged, audited step. Tokens are deterministic and join-preserving, so records still match across datasets. Matching runs on real source values. A tokenized-at-rest matching field makes the run refuse. It throws a typed error, rather than merge everything blind. Derived phonetic and n-gram fields inherit the PII verdict of the field they came from. So a name cannot be rebuilt from its Soundex.	Masking rules, yes. Reversible tokenization is not vocabulary they use.	◆
MCP and agents	More than 200 MCP tools span Ontology, Auth, Corpus and Lighthouse. 143 commands are for mastering alone. That is about a third of the whole agent surface. Tool tokens carry scope ceilings. Per-call usage is metered and attributed to the agent. There is no REST API for golden records. It is MCP, a masked SQL view, and a DuckDB dimension.	A shipped Reltio MCP Server, plus AgentFlow. Named agents include Resolver, Unmerger, Profiler, Data Explorer and Segmenter. There is also an Agent Builder for custom ones. This is a real surface, not a slide.	◈
Unstructured data	Corpus indexes documents for retrieval with local embeddings. We extract text from PDF, DOCX and EML. Those facts do not get pushed into golden profiles. We run the other direction instead. Goldens get projected into RAG grounding documents. This only happens if they are agent-ready. They render from the already-masked view. Their chunks get hard-deleted when a record falls out of the gate.	AgentFlow Unstructured ingests PDFs, Word files and HTML to enrich unified profiles. A folder of documents becomes profile attributes. That is a feature we do not have.	◇
Change stream and lineage	A value-free CloudEvents 1.0 outbox has 21 event types and a closed, contract-tested key set. No attribute value can ride an event. That's why it needs no erasure cascade. Delivery is exactly-complete, in commit order. Event lag shows on each mutating response. Golden to source-dataset lineage is traversable through typed edges. An edge is visible only when you can read both ends.	A cloud platform with an event and connection story we could not inspect in public detail. Ask what leaves the platform on a change, and whether values travel with it.	◈
Vendor independence	We are small and independent. Design-partner terms include source escrow. So a small vendor is not a single point of failure.	Owned by SAP since 7 May 2026. SAP says the portfolio stays available standalone "for the foreseeable future". That is a sentence, not a contract term.	◆
Maturity signals	Auth, Guardian and Lighthouse are live in production (Guardian and Lighthouse since April 2026). SOC 2 is not yet certified, and we say so. The matching engine is checked against roughly 370,000 real public records. These span three City of Chicago datasets, not only synthetic ones. We publish no record-count ceiling, and we will not invent one.	Founded 2011. $185M ARR at the end of FY26. Over 200 big customers, 43 of them Fortune 500. Gartner MQ Leader. ISO 27001, plus alignment to HITRUST CSF and NIST 800-53. Customer, partner and ARR figures are theirs, from their own press pages.	◇
Pricing	Published model, scoped instant quote, no sales wall.	Quote only. Billed by consolidated profile count, and invoiced annually in advance, according to third-party summaries. Reltio publishes no figures. Post-acquisition, the number may arrive inside an SAP negotiation.	◆

◆ DataShield leads◇ Reltio leads◈ comparable

Reltio claims are drawn from reltio.com, Reltio press releases and SAP's own newsroom. We checked these last on 13 September 2026. We link them below rather than work from memory.

Three things you get here that you won't get from an enterprise MDM suite

An agent that builds the model with you

Most MDM buying starts with a statement of work. Ours starts with a conversation. An agent asks your business user what they sell. It asks who their customers are, and which system wins when two disagree. It builds a draft model. Then it makes you check it, decision by decision. The generated objects land as drafts. So nothing masters anything until you promote it. How the ontology works.

Authority that can change mid-flight

A steward leaves on a Friday. Their agent is still 20 minutes into a 40-minute merge run. With DataShield, the next governed tool call gets re-checked against current authority, and fails closed. Not the next token refresh. The next call. How Auth does it.

Golden records that earn their reads

A record only reaches an agent once it clears six quality axes. It also needs a live lawful basis for that purpose. Miss either one, and the read excludes it. You also get told which class it fell into. The proof sits on a hash chain with signed checkpoints. That's the property EU AI Act Article 12 and HIPAA §164.312(b) reviewers ask about. Try the verifier, no signup.

Where Reltio is genuinely stronger

We would rather you heard this from us. Reltio has been building MDM since 2011, and it shows. It has multidomain breadth. It has a stewardship UI real business users will sit in all day. It has pharma reference designs. It has a bench of systems integrators who have done this fifty times. Their own press pages put them at $185M ARR at the end of FY26. That's over 200 big customers, 43 of them Fortune 500. Gartner named them a Leader in April 2026. Their agent story is shipped, not sketched. There is AgentFlow, eight or more named agents, an Agent Builder, and an MCP Server. AgentFlow Unstructured does something we simply do not do. It turns a folder of PDFs and Word files into enriched profile attributes. If your problem is breadth, buy breadth.

Here is the push-back, and it has two halves. The first is where the product now lives. Reltio is cloud only. Since 7 May 2026, it's a piece of SAP Business Data Cloud. SAP says standalone purchase continues "for the foreseeable future". Read that phrase again, slowly. The second half is what those named agents actually sit on. They work your data once the model exists. Building the model is still a project, with a partner attached. Our bet runs the other way. The agent helps you build the model in a conversation. Each generated object lands draft and gated. No config goes live without a green test battery. No golden reaches an agent until it clears a quality score and a lawful basis. Then a pulled access lands on the next tool call. An examiner can check the log without taking anyone's word for it.

We found none of that last part in Reltio's public docs. That is the gap we sell into. We are happy to lose the rest.

Questions worth asking both of us

These are the questions we would want answered if we were the ones buying. Ask them on every call, ours included.

Can you cryptographically prove an audit log entry wasn't deleted?

DataShield: yes. Each record commits to the one before it. Checkpoints are Ed25519-signed and chained. The verdict is clean, attested damage, or tampered. Deletion reads differently from truncation, and differently again from an edit. Mastering adds its own trail. Every merge, unmerge, pin and override is written before and after. That happens in the same step as the change. Run it at /verify. Reltio: AgentFlow lists audit logging, and the platform holds ISO 27001. We found no tamper-evidence method in their public docs. Ask them to show one.

What happens to a revoked agent mid-session?

DataShield re-checks authority on every governed tool call. So a pulled access lands on the next call. Reltio has role-based access control and an MCP Server. But there is no public detail on token lifetime, scope model, or how they revoke access. Ask how long a compromised agent keeps reading golden records after you pull its access. The answer is usually the token TTL. And the token TTL is usually too long.

How does GDPR erasure interact with the audit trail?

DataShield crypto-shreds the per-subject key. That destroys each ciphertext for that subject at once. It also issues an ISO 27560 consent receipt. Actor identities in the chain are HMAC-committed, so the evidence still verifies. Golden version history survives as a tombstone, with its reason attached. The change stream carries no values at all. So there is nothing to purge downstream. Reltio added masking rules in 2025.1. Masking is not erasure. Ask for the mechanism, not the workflow. Ask what it does to match history.

Do I need an implementation partner to stand up the model?

Not with us, and that is the part of this comparison we care most about. An agent runs the interview. It builds a draft spec. Then it hands you a decision gate, in order. Each proposal tells you what it wants to do, and why. It also tells you what it affects, and what else you could pick. Override one, and every choice that depended on it reopens. Once the gate is clear, one single step does the rest. It generates the entity types, hierarchies, draft match configs, reference bindings and rules. Each match config lands as a draft. The only route to live is train, then test, then promote. Reltio ships agents that work the data. Third-party summaries put a Reltio implementation at six to twelve months. Ask both of us who writes the first model, and how long that takes.

What does the SAP acquisition mean if I buy Reltio now?

SAP closed on 7 May 2026. They call Reltio "a core capability within SAP Business Data Cloud". SAP's own March release says the portfolio stays available standalone "for the foreseeable future". We will not tell you that is a disaster. It is not. SAP has money and a field team. Reltio will get both. But did you choose Reltio to stay independent of a platform vendor? Then ask for the standalone commitment in writing. Get a term on it. Also ask who sets the roadmap for a non-SAP customer in year three.

Can we run Reltio, or DataShield, inside our own infrastructure?

Reltio is cloud only, in everything we could find. It is built on AWS, Google Cloud and Azure. Now it is delivered through SAP's cloud too. DataShield ships Docker images for Auth, Ontology, Corpus and Lighthouse. It runs self-hosted in your own cloud or data center. Or it runs as a dedicated single-tenant server we operate. Does a residency rule or a board call say the master data index stays inside your walls? Then that is the whole conversation.

Is DataShield a replacement for a full multidomain MDM programme?

Sometimes, and we will be straight about which times. Do you need customer, product, supplier and location live at once? Do you want a stewardship UI for two hundred business users? Or a packaged 360 model to start from? Then Reltio is the better buy. We do not ship a packaged 360, on purpose. We fit the team that wants one or two domains modelled properly. They host it themselves. Matching they can read comes with it. They get an agent read path that gates on quality and consent. Our public benchmark shows zero false merges on every trap class. That includes 40 different people who share an exact phone number. Plenty of buyers in that shape have been quoted enterprise MDM, and walked away.

Does DataShield have SOC 2?

Not yet, and we will not imply otherwise. Auth is live with a public threat model and a verifier anyone can run. Guardian and Lighthouse have been in production since April 2026. Reltio holds ISO 27001. It says its security program follows HITRUST CSF and NIST 800-53. That is more than we can claim today. Design-partner terms include source escrow. So a small vendor is not a single point of failure. Details on the security page.

Reltio's current positioning: "The future is agentic AI. The foundation is Reltio". Product lines include Intelligent 360, Multidomain MDM and AgentFlow. - reltio.com, 13 Sep 2026
SAP completed its acquisition of Reltio on 7 May 2026. That made it a core capability within SAP Business Data Cloud. - SAP News, 7 May 2026
"The Reltio portfolio will also remain available as a standalone offering for the foreseeable future," with a flexible model to buy it separately or bundled with SAP products. - SAP News, 27 Mar 2026
Reltio's own release claims $185M ARR for the fiscal year ended 31 January 2026, over 200 large enterprise customers and 43 Fortune 500. Introduces AgentFlow and the Reltio MCP Server. - Reltio press release, 10 Feb 2026
Reltio AgentFlow ships named agents. These include Resolver, Unmerger, Profiler, Work Assigner and Data Explorer. There is also an Agent Builder, and AgentFlow Unstructured for PDF, Word and HTML ingestion. - reltio.com/agentflow, 13 Sep 2026
Named a Leader in the 2026 Gartner Magic Quadrant for Master Data Management Solutions. - Reltio press release, 8 Apr 2026

Other head-to-heads

MDM

Get free Demo Center access

You've seen the proof

Ready for a number? Scope your deployment and we'll price it against your own economics.

Get your quote →