Question 1

Is Eisberg really replacing 8+ tools?

Accepted Answer

Yes — and the savings show up in your renewal cycle. The math: ingestion + transformation + catalog + observability + reverse-ETL + an agent framework + your warehouse averages $1.2M-$3M/year for a mid-market data org. Eisberg's outcome-priced platform tier plus a single governed substrate routinely lands at 30-50% of that — with measurable savings replacing a maintenance burden. Bring an invoice list to the demo; we model it on the spot.

Question 2

How is this different from a 'unified data platform' pitch?

Accepted Answer

Microsoft Fabric, Databricks, and Snowflake all sell 'unified data platforms.' What they mean is 'unified inside our perimeter, using our proprietary format, billed by our consumption meter.' Eisberg unifies on open Apache Iceberg in YOUR bucket. The substrate is portable. The compute layer is replaceable. The customer can leave at any time and their data stays with them. That's not a feature; it's a different architectural philosophy.

Question 3

What about the tools you DON'T replace?

Accepted Answer

Four categories we deliberately don't build: BI authoring (Tableau, Power BI, Looker, Sigma, Omni, Mode, Hex, ThoughtSpot, Qlik) — we feed them. CDP (Segment, RudderStack, mParticle) — we feed them. Product analytics (Mixpanel, Amplitude, PostHog, Heap) — we ingest their event firehose. Marketing attribution (Triple Whale, Northbeam, Rockerbox) — we feed them. Each is a knife-fight we'd lose; partnering serves the customer better.

Question 4

Can I migrate incrementally?

Accepted Answer

Yes. Most customers connect Eisberg to their existing Snowflake or Databricks first, run Eisberg's intelligence layer (Knowledge Layer + ontology + agents) against the warehouse in place, prove the savings + capabilities, then graduate to Iceberg-on-your-cloud as the system of record on your schedule. Parallel running is the default — never a forklift cutover.

Question 5

What's the tribal-knowledge claim everyone else seems to be missing?

Accepted Answer

Look at where business knowledge actually lives. It's not in your data warehouse. It's in the Slack thread that explains why this metric is computed weirdly, the Confluence page that documents the close process, the pull-request comment where the engineer captured an edge case. None of that surfaces in your warehouse, your notebook tool, your catalog, or any other platform — at most they link to it. Eisberg ingests the bodies, extracts the facts, binds them to your business ontology, and cites them in every agent answer. Nobody else does this — and once you have it, you cannot pitch a platform without it.

Layer	Modern data stack	Eisberg
Ingestion / ELT	Fivetran + Matillion + Airbyte + Hightouch Three tools, three contracts, three integration ceremonies. Each charges per-source per-connector.	One Lifecycle ABC + 21+ adapters + Connector SDK + Airbyte adapter (700 sources via one build). Reverse-ETL native. Outcome-priced.
Storage	Snowflake / Databricks proprietary tables Your data lives in their account. They charge you to read it back. Egress fees discourage portability.	Open Apache Iceberg in your own cloud bucket. Customer KMS keys. Any compliant engine can read it. Eisberg never holds your data.
Query engine	Snowflake compute / Databricks DBSQL / ClickHouse Cloud One engine billed per second. Wrong workload? Wrong cost or wrong latency.	Multi-engine router picks per workload. Embedded for sub-second. Distributed for medium. GPU-native for heavy aggregation. MPP-on-Iceberg for real-time at scale. Customer never picks an engine.
Orchestration	Airflow / Astronomer / Prefect / Dagster DAG authoring, custom operators, environment drift, on-call pages.	Unified Job spine with durable execution. Any autonomous unit of work — human, agent, scheduled, triggered — runs as one Job with budget, observers, self-healing, audit trail.
Transformation / modeling	dbt / dbt Cloud / SQLMesh / Coalesce YAML you maintain. Tests that pass while data drifts. Re-run hell.	Job spine + OSI v1.0 semantic layer with native import/export from dbt / Cube / MetricFlow / ThoughtSpot / Sigma. Metrics travel into Eisberg without translation. Compounding self-healing.
Catalog / metadata	Atlan / Collibra / Alation / DataHub A second source of truth that goes stale immediately. Crowdsourced glossaries, abandoned.	Native catalog + ontology auto-linker. Your business ontology emerges from watching the workload in two weeks — not from six months of consulting.
Observability / quality	Monte Carlo / Bigeye / Datafold / Soda / Anomalo Another contract, another integration, another alert channel.	Column-level lineage emitted at write-time. HMAC-chained audit trail. Native data-quality engine with self-healing pipelines. One platform, one alert plane.
Reverse-ETL / operational	Hightouch / Census / Polytomic Yet another vendor for write-back. Per-destination pricing.	Native reverse-ETL with policy enforcement on every write. Reverse-ETL pipelines compile as Jobs and inherit Birth Certificate scoping.
Agent / AI orchestration	CrewAI / LangGraph / Microsoft Agent Framework / Sierra / Decagon / Lindy Agent frameworks + vertical CX products with their own state management. Frameworks (CrewAI Flows, LangGraph checkpointers, Microsoft Agent Framework) ship durable execution in 2026. Vertical products (Sierra Agent Data Platform, Decagon, Maven) ship persistent memory. They are not stateless dispatchers — that framing was from 2024.	Complementary primitives, not replacement. Agent frameworks consume Eisberg's governed substrate via the standard agent protocol; Eisberg signs each agent action with a Birth Certificate at the compute layer where the action actually runs. Where Eisberg is genuinely differentiated: cryptographic agent identity, customer-owned data in open format, seven-sub-policy governance plane (role + attribute + row-level + masking + purpose + consent + temporal), and a multi-engine query router. The honest direct competitors at the platform layer are Microsoft Agent Framework, Salesforce Agentforce, and ServiceNow AI Agents.
Governance / access	Immuta / Privacera / Okera / Satori Per-tool RBAC. Compliance officer maintains a spreadsheet of who can read what.	One seven-policy plane composing RBAC + ABAC + RLS + dynamic masking + purpose limitation + consent + temporal access on one engine. Fail-closed by construction. Nobody else has this stack.
Knowledge Layer	— None. Nobody else has this.	Every query, every action, every classification, every approval captured as institutional memory. The platform gets smarter from operating, not from training. Cross-customer compounding under k≥3 anonymity.
Tribal knowledge capture	— None. Tribal knowledge lives in Slack threads and Confluence pages — invisible to your data stack.	Slack + Teams + Confluence + Notion + GitHub + Google Docs + meeting transcripts ingested as governed facts. Bound to your business ontology. Cited in every agent answer. The Slack thread that explains your AR aging shows up in the variance investigator's report.
BI / visualization	Tableau / Power BI / Looker / Sigma / Omni / Hex Tableau / Power BI / Looker / Sigma / Omni / Hex — pick one.	We don't build this. We feed your existing BI tool a governed Iceberg substrate via the OSI v1.0 semantic layer. Sigma, Omni, Mode, Hex are preferred partners.

8 tools. 8 contracts. 8 integrations. 8 renewals. Or one platform that learns as it goes.

The Knowledge Layer + Tribal Knowledge capture is the wedge.

Eisberg learns your business

The stack does not

The honest objections we hear.

Bring an invoice list.