Research gaps and source caveats — trust / referral networks / in-group reputation brief (May 2026 Brief #3)

Where this brief's claims stand on weaker empirical ground. Label claims accordingly when used in Candid content.

1. No replicated study of referral-conversion rates in Canadian residential construction. [[schmitt-skiera-van-den-bulte-2011-16pct-ltv-premium]] is banking; [[trusov-bucklin-pauwels-2009-wom-elasticity-053]] is an online social network. Cross-domain direction holds; magnitudes do not transfer at face value. APB SORCI 2024 ([[apb-sorci-2024-48-7pct-referrals-half-of-sales]]) gives in-domain volume but not conversion rates by tie strength.

2. No published independent study of BTA / EOS / Vistage / TAB outcomes in Canadian trades. Outcome figures from [[bta-contractor-growth-method-600-contractors]], [[eos-worldwide-scale-200k-businesses-800-1000-implementers]], [[vistage-tab-generalist-peer-advisory-not-trades-specific]] are organization-reported. Treat as marketing claims, not as evidence.

3. The "peer-coached contractor" persona is a synthesis, not an empirically validated segmentation. [[peer-coached-contractor-persona-synthesis]] should be read as a directional hypothesis worth field-testing in Candid's own sales-conversation data, not as a measured construct.

4. The credence-good frame is imported from healthcare, auto-repair, and durable-goods literatures. [[darby-karni-1973-credence-goods-framework]], [[marketing-services-as-credence-good-for-gc]], [[dulleck-kerschbamer-sutter-2011-liability-verifiability-credence-goods]]. Applicability to "marketing services for trades businesses" is plausible and theory-consistent, but in-domain published work is thin. We are extrapolating.

5. The Granovetter / Burt / Levin-Cross framework is US-derived. Canadian trades may differ (smaller market, more concentrated supplier base, statutory warranty structure with no US analogue). Cross-domain mapping flagged in [[burt-structural-holes-1992-2004-broker-advantage]] and [[brokers-in-ontario-trades-three-roles]].

6. APB SORCI 2024 referral data and Brief #1's 48.7% figure ([[apb-sorci-2024-48-7pct-referrals-half-of-sales]]) have not been independently re-derived in this brief. Taken as established within the Candid evidence base.

7. The Reddit / WhatsApp / Signal / Facebook-group digital infrastructure is rapidly evolving. Layer 5 of [[ontario-trades-info-ecosystem-five-layer]]. Any specific claim about volume or composition should be revisited within 12 months.

8. No Ontario-specific data on what percentage of HCRA-licensed builders also hold Gold Seal, RenoMark, or Guildmaster designations. Cross-tabulating these stacks would meaningfully sharpen the persona work — high-value future research project.

9. No data on Candid's own referral mix by tie strength. Building this dataset internally is recommended. If 80% of Candid's high-LTV clients arrive via strong-tie referrals ([[warm-intro-hierarchy-eight-levels-ontario-trades]] levels 6-8), that should be governing budget allocation now, not after the next year of media spend.

10. GuildQuality's annual proportion-of-members-winning rate is not publicly disclosed. [[guildmaster-guildquality-audit-verified-survey-signal]]. Winner list is published; denominator is not, so "what fraction of GuildQuality members are Guildmasters" cannot be computed from public data.