Three out of four CRM databases can't pass a basic accuracy check. According to Validity's 2025 State of CRM Data Management report (cited by Reach Marketing), 76% of respondents said less than half of their CRM data was accurate and complete. That's the foundation most teams are feeding into AI-powered lead scoring, personalization engines, and ABM platforms right now.

Patrick Shea, co-founder of AdDaptive Intelligence, made the case in a recent B2BMX Insights Q&A that this mismatch is getting worse. His argument: AI raises the stakes on data foundations because it can only work with the context you give it. "Human beings subconsciously add tremendous context to the data we consider," Shea said. "Without context, an off-the-shelf system can't THINK the way you can."

Identity Resolution Breaks Everything Else First

Shea was direct about which failure mode kills an ABM program fastest. "Resolution breaks all of it the fastest," he said. "If you're not identifying and reaching customers correctly, the rest won't even matter because even if you do everything else right, you're still talking to the wrong people and won't know why your marketing isn't working."

Operational data backs this up. A Vandfort analysis of more than 12 billion Salesforce records found that 80% of records created through API integrations were duplicates, compared to 19% for imports. If your MAP-to-CRM sync or enrichment tool is the primary way records get created, you might be manufacturing noise at scale before any scoring model touches the data. And only 23% of B2B marketers trust their identity-resolution data enough to act without manual review, according to Influencers Time. That number should concern anyone defending ABM ROI to a CFO.

"Closed Loop" Means Nothing Without Signal Reuse

Shea drew a sharp line between platforms that accumulate data and those that actually compound intelligence from it. His framing: most ABM platforms treat each campaign as a fresh start. The signals generated (engagement patterns, creative response rates, account-level behavior) die when the campaign wraps.

"Paid media is not simply about messaging your market; it's an additional opportunity for rich, deep engagement metrics generated by the interaction with the ads themselves," Shea said. The point isn't just reach. It's whether you can look at how different segments of your ICP responded to specific messaging, then use those signals to trigger next actions across email, CTV, social, or sales outreach.

For demand gen operators, this maps to a familiar problem: your LinkedIn campaign data, your Google Ads conversion signals, and your CRM stage data live in different systems with different identity keys. Harmonizing the output of one channel into a usable input for the next is where most programs stall. Shea acknowledged the complexity: "Input data is scattered (emails, company names, internal account IDs, HEMS, IP addresses, geofences, first-party, third-party, opted-in, double-opted-in), and output data is inconsistent."

Three Diagnostic Questions Before Your Next Platform Decision

Shea offered three questions any marketing director should ask their current vendor:

  1. What percentage of my ICP can you actually resolve and reach?
  2. What signals can we generate from our paid media campaigns?
  3. How do I take those signals and inject that data throughout my marketing stack to trigger next best activities?

These map directly to the coverage-vs-correctness distinction that keeps tripping up ABM reporting. "Match rate" is often ambiguous; practitioners increasingly recommend separating coverage (records assigned to an account) from verified correctness (assignments confirmed as accurate). A high match rate can still hide wrong account assignments that distort both targeting and attribution. The operational implication: before you buy another enrichment tool, audit where records are actually being created. API-driven syncs between your MAP and CRM may be your largest source of duplicates, and no amount of periodic cleanup fixes a pipe that's still leaking.

What This Means for Pipeline Defense

Gartner estimates poor data quality costs organizations at least $12.9 million per year on average. McKinsey research cited by Demandbase suggests personalization can produce a 10-15% revenue lift for companies that get it right. The gap between those two numbers is where most B2B teams are operating: paying the cost of bad data while chasing the upside of personalization built on top of it.

Shea's forward-looking take was measured: "The teams that pull ahead will focus on keeping their data clean and portable, realizing that automation doesn't replace resolution or fidelity." No promise of a silver bullet. Just the observation that teams investing in data governance now are building the substrate that makes every future AI tool more reliable, while teams that skip it are compounding errors at machine speed.

The uncomfortable part of that argument is that data foundation work doesn't produce a pipeline number this quarter. It produces fewer wrong numbers next quarter. For a demand gen leader defending budget to a board, "fewer wrong numbers" is a harder sell than "more pipeline." But if 76% of your CRM is inaccurate, the pipeline number you're defending might already be one of the wrong ones.