Waterfall Enrichment Explained: How to Build a Multi-Vendor Data Pipeline

Waterfall enrichment is how serious B2B data operations handle the reality that no single vendor covers your entire addressable market.

The premise is simple: instead of routing every contact through one data provider and accepting whatever fill rate that vendor delivers, you chain multiple providers in a prioritized sequence. Vendor A runs first. Any records it can't resolve fall through to Vendor B. Vendor B's misses fall through to Vendor C, and so on, until you either have a verified result or have exhausted your cascade.

The result is materially higher coverage at lower blended cost per enriched record — and a data pipeline that doesn't collapse the moment one provider's coverage drops on a particular segment.

Why single-vendor enrichment fails at scale

Every B2B data vendor has a coverage model. They index contact records from a combination of web crawls, opt-in networks, user-contributed data, and licensed third-party sources. No vendor indexes the same universe of records.

The practical consequence: for any given ICP segment, one vendor might return a valid email for 45% of contacts. Another vendor — with completely different source infrastructure — might cover a different 35% of that same segment. In combination, those two vendors could reach 65–70% of the total addressable pool.

Run only one vendor and you're leaving the second coverage block unworked. That's not a data quality problem — it's a structural architecture problem.

Single-vendor approaches also create fragility. Vendor coverage degrades on specific industries, geographies, and company sizes. When you're sourcing SMBs in the Pacific Northwest or manufacturing companies in the DACH region, the delta between vendors can be extreme. A waterfall absorbs that variance.

Waterfall architecture: the core components

A production waterfall pipeline has four logical stages:

1. Input normalization. Before any enrichment runs, standardize input records: normalize company names, resolve domains, strip department prefixes from job titles, and deduplicate on email + company domain. Garbage in is amplified through every downstream stage.

2. Match and enrich (the cascade). Pass records to Vendor A. For records with no match or an unverified result, pass to Vendor B. Continue until the record is resolved or all vendors are exhausted.

3. Quality scoring per stage. After each vendor returns a result, score it. Verified deliverable emails score higher than catch-all or unverified results. Apply confidence thresholds and tag each record with its source vendor and confidence tier.

4. Deduplication and normalization. Merge enriched data back to your master record, resolve field conflicts (two vendors returning different job titles), and write a clean, source-attributed record to your CRM or outreach system.

Vendor selection criteria

Before mapping vendors to waterfall positions, evaluate them against these dimensions:

Coverage depth by segment. Run test batches of 500–1,000 known contacts in your ICP against each vendor. Track match rate, verified email rate, and catch-all rate by industry, headcount band, and geography. This is the only reliable way to rank vendors for your specific use case — published match rates are marketing, not measurement.

Data freshness and decay rate. B2B contact data decays at roughly 25–30% annually. Vendors with active re-verification pipelines and opt-in network refreshes produce lower bounce rates. Ask vendors directly: what is your median record age in the segments you're targeting?

Verification methodology. Some vendors return "verified" emails based on historical send data. Others run active SMTP verification. Others rely entirely on pattern matching. Understand the methodology; it determines how aggressively you can use the output.

Cost structure. Most vendors price on credits per record looked up or per field returned, not per successful match. Your blended cost per enriched record is a function of success rate times credit cost — compare on that metric, not list price.

API reliability and rate limits. Production pipelines need predictable uptime and rate limits that don't throttle mid-batch. Test API response times and document rate ceilings before you build dependencies.

How the major vendors compare

These are general market observations, not vendor endorsements. Run your own segment-specific benchmarks.

ZoomInfo has the broadest enterprise coverage and deepest firmographic data. It performs best on North American mid-market and enterprise contacts. Credit costs are high relative to competitors. Best placed as Tier 1 for enterprise ICPs.

Apollo has strong SMB and mid-market coverage, particularly for US-based tech and SaaS companies. Its data is community-sourced and refreshed frequently. Lower cost per credit makes it well-suited for high-volume Tier 2 enrichment.

Lusha performs well on European contacts and executive-level titles (VP+, C-suite) where other vendors have thinner coverage. Useful as a specialized Tier 2 or Tier 3 vendor for those segments.

RocketReach tends to have broader personal email coverage than most vendors, which can be valuable for certain outbound plays. Phone number coverage is above average. Useful as a Tier 3 for contacts where corporate emails aren't resolving.

Clearbit (now HubSpot Enrichment) has strong firmographic and technographic data, with solid coverage on US tech companies. Email coverage has historically trailed ZoomInfo and Apollo but the firmographic layer is valuable for account scoring and segmentation.

People Data Labs (PDL) provides raw data access at high volume with flexible API pricing. Coverage is broad but data is less frequently re-verified than premium vendors. Well-suited as a high-volume Tier 2 or Tier 3 where cost per credit matters more than peak accuracy.

Vendor	Best Segment	Relative Cost	Waterfall Position
ZoomInfo	NA Enterprise, Mid-Market	High	Tier 1
Apollo	SMB / Mid-Market, US Tech	Low–Medium	Tier 1 or 2
Lusha	European contacts, VP+	Medium	Tier 2 or 3
RocketReach	Phone numbers, personal email	Medium	Tier 3
Clearbit / HubSpot	Firmographics, US Tech	Medium	Tier 1 (firmographic), Tier 3 (email)
People Data Labs	High volume, broad coverage	Low	Tier 2 or 3

A concrete email enrichment waterfall example

Here is a worked example for a US mid-market SaaS ICP (companies 100–500 employees, technology sector, targeting VP Sales, VP Marketing, and RevOps titles).

Input: 1,000 contacts with name, company name, and LinkedIn URL. No email addresses.

Tier 1 — Apollo:

• Match attempt: 1,000 records
• Verified email returned: 420 records (42%)
• Catch-all returned: 65 records (6.5%; held for secondary verification)
• No match: 515 records passed to Tier 2

Tier 2 — ZoomInfo (on the 515 unresolved):

• Match attempt: 515 records
• Verified email returned: 195 records (38% of this batch; 19.5% cumulative)
• Catch-all returned: 28 records (merged into catch-all pool)
• No match: 292 records passed to Tier 3

Tier 3 — People Data Labs (on the 292 unresolved):

• Match attempt: 292 records
• Verified or pattern-matched email returned: 87 records (30% of this batch; 8.7% cumulative)
• No match: 205 records (no enrichment found across all three vendors)

Secondary verification pass (catch-all pool):

• Total catch-alls collected across tiers: 93 records
• Run through dedicated email verification (e.g., NeverBounce, Zerobounce)
• Verified deliverable: 41 records promoted to usable pool
• Risky / unverifiable: 52 records flagged and sequestered

Final output on 1,000 input records:

• High-confidence verified: 420 + 195 + 41 = 656 records (65.6% overall coverage)
• Pattern-matched / lower confidence: 87 records (separate sending treatment)
• No email found: 205 records (routed to LinkedIn-only outreach or manual research)
• Unreachable: 52 catch-alls flagged as risky (suppressed from sends)

Without the waterfall (Apollo only): 420 verified + 65 catch-all = 485 records maximum, with significant bounce risk on the catch-all pool. The two-stage cascade adds 276 additional deliverable emails to the usable set — a 57% increase in reachable contacts from the same input file.

Cost optimization strategies

Waterfall enrichment is only cost-efficient if you control what you send to each tier.

Only send misses to the next tier. Never re-run a successfully enriched record through a more expensive vendor. If Apollo resolved the contact, stop. The cost savings of not running 420 records through ZoomInfo on the example above are significant at scale.

Apply ICP scoring before enrichment. Not every contact in your system deserves a full waterfall. Score accounts and contacts against your ICP before running enrichment. High-fit contacts get the full multi-vendor cascade. Low-fit contacts get a single-vendor pass or no enrichment at all.

Separate firmographic from contact enrichment. Firmographic enrichment (company headcount, revenue, tech stack, funding) is cheaper and more stable than contact-level email enrichment. Run a single-vendor firmographic pass to qualify accounts before spending credits on contact-level email resolution.

Batch over real-time where possible. Real-time API enrichment is convenient but expensive at volume. For prospecting workflows that aren't time-critical, batch enrichment runs at scheduled intervals reduce per-record costs and make it easier to control spend against monthly credit budgets.

Building the pipeline in Clay

Clay is the most commonly used orchestration layer for waterfall enrichment in modern outbound stacks, and for good reason: it supports direct integrations with most major data vendors and its "Claygent" AI agents can handle research steps that structured API calls can't.

A Clay waterfall setup looks like this:

1. Import your contact list as a Clay table.
2. Create a formula column that checks whether the email field is populated.
3. Add a data provider column (e.g., Apollo) configured to run only when email is empty.
4. Add a second data provider column (e.g., ZoomInfo) configured to run only when the first provider returned no result.
5. Add a third provider column following the same conditional logic.
6. Add an email verification column (NeverBounce or ZeroBounce) that runs against any populated email field, tagging results as valid, risky, or invalid.
7. Create a final "resolved email" formula column that pulls from whichever tier populated first, with the verification status attached.

The conditional logic — "only run this enrichment step if the previous step returned empty" — is the core of the waterfall. Without it, you're paying for every vendor on every record.

For data enrichment at scale, we build these pipelines with additional quality checkpoints: domain validation, catch-all detection, role-based email suppression (e.g., suppressing info@, sales@, support@ addresses that will never reach a named decision-maker), and source attribution on every enriched field.

Data normalization and deduplication

Enrichment pipelines that skip normalization create downstream problems in your CRM and outreach tools.

Field conflict resolution. When two vendors return different job titles for the same person, you need a deterministic rule: prefer the vendor with higher confidence score, or prefer the more recent data, or prefer the vendor with higher verified email match rate on that segment. Define the rule before you build the pipeline, not after.

Company name standardization. "HubSpot Inc.", "HubSpot", and "HubSpot, Inc." are the same company. If your deduplication logic treats them as three records, you'll build contact lists with multiple copies of the same people. Use domain as your canonical company identifier, not company name string.

Role-based email suppression. After enrichment, filter out role-based and shared mailbox addresses (info@, contact@, team@, legal@, etc.). These addresses technically "exist" and won't bounce, but they route to inboxes monitored by teams, not to the individual decision-maker you're targeting. They generate spam complaints and kill deliverability.

Source attribution. Tag every enriched field with its source vendor and enrichment date. This matters for compliance documentation, for troubleshooting data quality issues, and for tracking which vendors are performing well in your specific segments over time.

Quality scoring at each stage

Treat enriched data as a spectrum, not a binary. A record with a verified email from ZoomInfo's actively-refreshed database is not equivalent to a pattern-matched email from PDL that was last confirmed 18 months ago.

Define confidence tiers and route them differently in your outreach tooling:

• Tier A — Verified deliverable (SMTP-confirmed): Full send cadence, standard volume.
• Tier B — High-confidence unverified (major vendor, fresh record): Standard send cadence, monitor bounce rate closely.
• Tier C — Pattern-matched or aged (low-confidence vendor output): Reduced send volume, external verification pass before use, sequester on first bounce.
• Suppressed — Catch-all unverifiable, role-based, or invalid: No outbound send. Route to LinkedIn-only plays or manual research queue.

This tiering integrates with lead scoring frameworks to ensure that your highest-confidence contacts get your highest-effort outreach, and that marginal data doesn't contaminate your sending reputation. See our post on B2B data quality costs and contact verification at scale for deeper coverage of how bad data propagates through outbound systems.

Connecting enrichment to outbound execution

A waterfall enrichment pipeline that sits in isolation is infrastructure, not a revenue system. The output needs to route cleanly into your outbound systems.

High-confidence enriched contacts should enter your sequencing tool with ICP score, enrichment confidence tier, and source vendor attached as custom fields. That metadata drives:

• Cadence selection (high-confidence vs. lower-confidence contact sequences)
• Personalization depth (accounts with rich enrichment get more specific first lines)
• Deliverability throttling (lower-confidence tiers get lower daily send caps per inbox)
• Bounce-triggered suppression (first hard bounce from a Tier C record pulls all same-domain records into a review queue)

The OppZo case study shows this architecture in practice — enrichment output feeding directly into segmented outbound sequences with per-tier deliverability controls.

FAQ

How many vendors do you actually need in a waterfall?

Three is the practical sweet spot for most mid-market B2B ICPs. The coverage gains from adding a fourth vendor are typically marginal (5–8% additional fill on records that three vendors couldn't resolve), and the operational complexity of maintaining four API integrations, four billing relationships, and four sets of data terms grows non-linearly. Start with two vendors, measure your fall-through rate, and add a third only if you're seeing significant unresolved volumes on segments that matter to your pipeline.

What is a good overall email coverage rate to target?

For a US mid-market SaaS ICP, 60–70% verified email coverage on a cold contact list is a realistic and good outcome from a 3-vendor waterfall. If you're seeing below 50%, audit your input data quality (bad name/company pairings, outdated LinkedIn data) before adding more vendors. Coverage above 75% is possible for narrow, well-defined ICPs in sectors with strong data vendor penetration (US tech, finance) but is not the norm across the board.

Should you use catch-all emails in outbound sends?

With caution and at reduced volume. Catch-all domains accept all email addresses — valid or not — at the SMTP level, so you can't confirm whether a specific address is real. Run catch-all results through a secondary verification service and separate them into their own sending pool with lower daily volume limits. Monitor bounce rate on that pool independently. If it exceeds 3–4%, pause and scrub the pool before continuing.

How do you handle GDPR and data privacy in a multi-vendor waterfall?

Each vendor in your waterfall must have a documented lawful basis for the data they're providing. Review each vendor's DPA (Data Processing Agreement) before building a dependency. For European contacts specifically, document your legitimate interest basis, ensure your outreach meets local requirements for B2B cold contact, and honor opt-out requests across all active enrichment sources — not just the vendor where you sourced the original record.

Can you run a waterfall in real-time, or does it have to be a batch process?

Both are architecturally possible, but real-time waterfall enrichment is expensive and adds latency to contact creation workflows. The most common approach is a hybrid: near-real-time enrichment from a single primary vendor on new contacts as they enter the system, with a nightly or weekly batch waterfall job that resolves any records the primary vendor missed. This balances cost, speed, and coverage without requiring synchronous API calls to three vendors on every new contact event.

---

Waterfall enrichment is a standard architectural pattern in serious B2B data operations for a reason: it eliminates the structural ceiling that single-vendor approaches impose on your reachable market. The implementation complexity is real — vendor evaluation, conditional pipeline logic, normalization rules, confidence tiering — but the delta in reachable pipeline is durable and compounds over time as you refine which vendors perform best in your specific ICP segments.

If you're building or auditing your enrichment infrastructure, our data enrichment service covers the full pipeline: vendor selection, waterfall architecture, verification integration, CRM normalization, and quality monitoring. Talk to our team to walk through your current setup and identify where coverage is leaking.