I ran 50 mid-market contacts through both PDL's Enrich API and Apollo's bulk enrichment endpoint over three days in late March 2025. Same input: full name + work domain. The result: 34% of those contacts came back with a different job title between the two providers. Not just wording differences — actual seniority mismatches. "Director of Revenue Operations" in Apollo, "Revenue Operations Manager" in PDL. "VP of Marketing" in PDL, "Head of Marketing" in Apollo. For outbound personalization tokens, that's not a minor inconsistency. That's the difference between "Hi [First Name], saw you recently stepped into a VP role" landing well and landing embarrassingly wrong.
Here's why this happens structurally, and which source you should actually pipe into your sequence tokens versus your routing rules.
Why the Same Person Gets Two Different Titles
The divergence isn't random — it comes from three structurally different data pipelines.
Apollo operates primarily off LinkedIn scrape snapshots with a self-reported correction layer. When a person updates their LinkedIn title, Apollo picks it up on their next crawl cycle. That cycle varies, but in my experience it's 60–90 days for mid-market contacts who aren't heavily engaged on the platform. Apollo also surfaces title data from email signature inference and CRM sync partnerships, which means you sometimes get the displayed title from an older email thread rather than the current LinkedIn title.
PDL aggregates from a wider set of sources — public social profiles, resume databases, ATS leakage into job boards, professional directory filings — and runs identity resolution across them before settling on a canonical record. The upside: PDL's title for a contact is often the intersection of multiple signals, so it's less likely to be a stale outlier. The downside: when those signals disagree internally, PDL tends to resolve toward the most common observed title across sources, which may lag an actual promotion by four to six months if the person hasn't updated everywhere simultaneously.
The structural asymmetry: Apollo skews toward current LinkedIn but inherits its lag. PDL skews toward cross-source consensus but can be anchored by older signals it overweights. Neither is strictly more accurate. They're wrong in different directions.
What 50 Contacts Actually Showed Me
I pulled a cohort from a Clay table — 50 mid-market contacts in SaaS (100–500 person companies), all in Marketing, RevOps, or Sales functions, all in North America. I ran them through PDL's Person Enrichment endpoint and Apollo's bulk export simultaneously. Then I manually verified current titles against LinkedIn for all 50 on the same day.
The numbers:
| Metric | Apollo | PDL |
|---|---|---|
| Exact title match (vs LinkedIn) | 28/50 (56%) | 24/50 (48%) |
| Seniority level correct (VP/Dir/Mgr) | 38/50 (76%) | 35/50 (70%) |
| Both providers agreed (same title) | 33 contacts | — |
| Disagreement cases | 17 contacts | — |
| Apollo correct on disagreement | 9/17 (53%) | — |
| PDL correct on disagreement | 6/17 (35%) | — |
| Both wrong on disagreement | 2/17 (12%) | — |
A few observations from digging into the disagreement cases:
The 9 Apollo wins were all recent promotions (within 90 days) where the person had updated LinkedIn but hadn't updated resume aggregators or other secondary directories. Apollo caught the new title; PDL was still anchored to the previous role.
The 6 PDL wins were older contacts with inconsistent LinkedIn behavior — people who had a title on LinkedIn that looked like a soft inflation (think "Head of" for what their company ATS showed as "Senior Manager"). PDL's cross-source consensus deflated those titles back down accurately.
The 2 cases where both were wrong: one was a founder who listed a board advisory role as a primary title on LinkedIn; the other was a contractor with a client-facing title that didn't reflect their employer-of-record seniority.
Which Source to Use for Personalization Tokens vs Routing Logic
This is where the structural difference actually matters for how you build sequences.
For personalization tokens in copy — things like {{current_title}}, {{seniority_level}}, or anything you're injecting into email body text or subject lines — Apollo is the safer default for recently active contacts. Its crawl bias toward current LinkedIn means you're less likely to congratulate someone on a role they left six months ago. If you're using Clay, pipe Apollo's title first; use PDL as a fallback for contacts Apollo doesn't return a title on.
For routing and scoring logic — ICP tier assignment, territory rules, BDR vs AE handoff thresholds — don't use either title field raw. This is where you want to normalize. Take both values, run them through a simple seniority classification layer (even a regex pass against a seniority taxonomy works), and route on the normalized output rather than the raw string. A disagreement between "VP" and "Director" shouldn't flip routing logic if both are senior enough for your ICP.
Here's the basic normalization approach I use in Clay's formula fields:
// Seniority normalization — works as a Clay formula or in a JS enrichment step
const title = pdlTitle || apolloTitle || "";
const t = title.toLowerCase();
if (t.includes("chief") || t.includes("cxo") || t.includes("c-suite")) return "C-Suite";
if (t.includes("vp") || t.includes("vice president")) return "VP";
if (t.includes("director") || t.includes("head of")) return "Director";
if (t.includes("manager") || t.includes("lead")) return "Manager";
if (t.includes("senior") || t.includes("sr.")) return "Senior IC";
return "IC";
This isn't sophisticated NLP, but it collapses "VP of Marketing," "Vice President, Marketing," and "VP Marketing" into the same routing bucket. The PDL vs Apollo disagreement stops mattering for your Salesforce assignment rule.
Where Each Provider Fails Predictably
Knowing the failure patterns in advance saves you the manual verification step I had to do.
| Scenario | Apollo Failure Mode | PDL Failure Mode |
|---|---|---|
| Recent promotion (< 3 months) | Often correct | Likely stale |
| Quiet LinkedIn user | Stale or missing | Better cross-source coverage |
| Founder / multi-role exec | Over-indexes on primary LinkedIn title | May surface board roles as primary |
| Non-US contact | Variable crawl coverage | Better for EU/APAC aggregation |
| Contractor or fractional | Client-facing title surfaced | Employer-of-record title may win |
| Small company (< 50 employees) | Title inflation from self-reporting | May under-resolve due to fewer signals |
The contractor and fractional scenario is worth calling out specifically. I had three fractional CMOs in the cohort. Apollo returned their client-company title for all three. PDL returned their LLC or consulting entity title for two of them. Neither was the right input for a "VP of Marketing at [Company]"-style personalization token without verification.
If your ICP includes fractional executives, interim leaders, or people who list advisory roles prominently — neither provider's raw title output is usable for personalization without a verification step. Maigret or a manual LinkedIn scrape via Phantombuster is more reliable for those edge cases.
What I Actually Use
For high-volume sequences where I'm not verifying manually, I use Apollo as the primary title source for personalization tokens and fall back to PDL when Apollo returns null or a title that fails a basic sanity check (I flag titles over 60 characters as suspicious — they're usually scraped job description fragments).
For routing and CRM field population, I normalize both using the seniority layer above and write the normalized value, not the raw string. The raw PDL and Apollo titles go into reference fields that reps can see but that don't drive automation.
For cohorts that include non-US contacts or people with sparse LinkedIn activity, I weight PDL more heavily. Its cross-source aggregation holds up better in those cases than Apollo's crawl-dependent output.
If you're using Clay specifically, Wiza, RocketReach, and Ziwa are worth testing as spot-check sources on the disagreement cases — particularly Ziwa if you want an API-first option that can sit alongside PDL and Apollo in a waterfall without requiring a full platform contract. But neither replaces the core PDL + Apollo pairing for volume enrichment.
The uncomfortable truth: neither provider is accurate enough to trust raw title output for personalization at scale without at least a seniority normalization pass. The 56% exact-match rate from Apollo and 48% from PDL mean you're injecting wrong titles into roughly one in two or one in three emails if you don't normalize. That's not a data quality complaint — it's a workflow design constraint you need to build around.
Top comments (0)