Research brief: The Dataset is the Product — when a service business should own its data (piece 12 of 15)

Status: Research material — not finished article. Compiled May 2026. B&J CRM is in-development — frame as illustration only; no outcomes claims.

Thesis

Most small service businesses sit at Data Maturity Stage 1 or 2 — data exists, sometimes captured, almost never queryable. The economic question isn't "should we own our data?" (almost always yes above ~$1-2M revenue with repeat customers) but "at what point does owning structure pay back the cost of building it?"

The defensible 2026 answer: when you have (a) repeat customers with multi-event history, (b) pricing or scheduling decisions that recur, and (c) a vertical SaaS that captures your transactions but not your model of the business. Below that bar, a tidy QuickBooks + spreadsheet + single-source-of-truth CRM is sufficient. Above it, the dataset is the product.

The dominant working pattern in 2026

• 91% of companies with 10+ employees use a CRM; ~50% of <10-employee businesses do (industry-consensus)
• Industry-cited CRM project failure rate: 47% (Forrester) to 55% (Johnny Grow) to 70% (multiple aggregators); poor user adoption is the leading cause
• The "captured but unusable" gap: ServiceNow reports "91% of data in CRM systems is incomplete, 18% is duplicated, and 70% becomes outdated each year" (single-source caveat)
• McKinsey: knowledge workers spend "nearly 20%" of the workweek searching for information
• MuleSoft 2025: organizations use 897 applications on average; only 29% are integrated

Vertical SaaS as the system of record

ServiceTitan IPO'd December 2024 with first-day market cap $8.9B, gross retention >95%, NDR >110% — the vertical SaaS playbook validates. But every major vertical SaaS (ServiceTitan, Jobber, Housecall Pro, Clio, Karbon, Tekmetric) captures the transaction well and the customer-specific business model poorly. That gap is where Stage 4 / Stage 5 businesses justify their own data layer.

Honest caveats

• The "91% of CRM data is incomplete" figure traces to ServiceNow without primary methodology — illustrative, not load-bearing.
• ServiceTitan exit-friction figures ($24k-$39k contract buyouts) are aggregated from BBB/Reddit complaints by Projul — two steps from source. Don't cite specific dollar figures without verifying an original BBB complaint.
• Lovett Services 25% revenue growth / 5% net profit margin claim is vendor-published (ServiceTitan) — not independently audited.
• Tunguz's 0.97 vs 0.66 sales efficiency for vertical vs horizontal SaaS is from 2015 (n=54 public SaaS companies). Directional, not 2026 benchmark.
• Quebec Law 25 Axeptio "<5% SME compliance" figure is from a consent-management vendor — directional.
• The B&J CRM has not shipped. Frame as conceptual illustration of "modeling the actual business," not as a delivered case study.