A useful owner list should survive the handoff
A useful owner list should survive the handoff. If the next person can't tell why two records were joined, the cleanup has left them another research task.
- Sasha Deneux
The Take: Keep the merge decision with the record
CompStak's September 29 announcement puts its commercial real estate data inside AI assistants through a new connector. The company also describes entity resolution connecting properties, owners and transactions behind those results. That second detail deserves attention when you're designing your own sourcing workflow.
Connecting a dataset makes its conclusions easier to reuse. An ownership match can travel from a research sheet into a portfolio summary, then into a proposed contact record. If the match was weak, the next tool may inherit it without the question that originally surrounded it.
I'd make the join itself reviewable. Which records were combined? What evidence supported that decision? Could another researcher separate them again without losing either source?
A shared mailing address might be useful evidence for further research. It might also be a service provider's office. The workflow needs somewhere to preserve both possibilities before it turns a group of LLCs into a single owner's portfolio.
That doesn't mean an analyst has to retype every record. Let software standardize formatting and propose matches. Give the reviewer the source comparison and the reason for the proposal. Keep a rejected match so it doesn't return as a fresh suggestion next week.
The owner-portfolio research guide takes that distinction into the property search. I want the handoff to preserve what the team actually established, including relationships it couldn't establish yet.
The Teardown: Clean the list without losing the evidence
Use a small practice export that your team is permitted to process. Include an obvious duplicate, two unrelated entities with similar names, and an unresolved relationship. Write down the expected treatment before asking an assistant to clean it.
Keep the source rows. Save the original file, retrieval date and a stable identifier for each row. Work on a separate table. A parcel identifier needs its jurisdiction beside it; the same identifier text in another county isn't enough to establish the same property.
Choose what you're matching. Repeated observations of a property, legal entities and contact people need separate identifiers. One entity can hold several properties. One person can have a documented role in several entities. Neither relationship makes those properties or entities duplicates.
Normalize into extra columns. Standardize spacing, punctuation and address formatting while preserving the raw values. Ask AI to propose candidate matches with the fields it compared. A similar name can enter the review queue without becoming an accepted merge.
Review the evidence as a pair. In a fictional example, two industrial-property rows share a tax mailing address. Their title-holding entities have different registration numbers in the same jurisdiction. Keep those entities separate. If a filing identifies a common manager, record that role as a relationship; it doesn't settle beneficial ownership. If the evidence is missing, assign the next check.
Make the decision reversible. Save the accepted or rejected match, supporting record, reviewer and date. Keep earlier values when a source changes. When you split an incorrect group, check whether its portfolio summary and contact research also need correction.
Test the handoff. Give another researcher the cleaned export without narrating it. Can they open the source, explain an accepted match and find the unresolved rows? Rerun the same practice file to check that rejected matches stay rejected and repeated imports don't create fresh duplicates.
Track incorrect merges separately from duplicates left for review. Both matter, but an incorrect merge can attach a property or contact to the wrong entity. The owner-list deduplication guide includes a practice download for testing those decisions before a live handoff.
Signal
Two recent releases and one supplier standard to keep beside the workflow:
September 29: external data moves into the assistant. CompStak's connector launch describes access through its One Connect offering. Why it matters: include the source identity and retrieval date in your output contract, so a useful answer remains checkable after export.
September 30: a warehouse needs its own operating context. CBRE's European logistics analysis discusses power capacity, building suitability and network decisions. Why it matters: use it to frame questions for relevant European assets. It doesn't establish the condition or sale intentions of an owner on your list.
Effective September 2: traceability belongs in supplier expectations. Cushman & Wakefield's AI charter calls for documented, explainable outputs and clear responsibility. Why it matters: ask a proposed sourcing supplier to demonstrate a disputed merge and its correction, alongside the attractive finished list.
From NextAutomation
We build sourcing workflows around acquisition criteria, property records and the decisions your team needs to review. Book a conversation with a redacted example of a list that's difficult to hand off. The free off-market sourcing system pack includes a source map and verification checklist to help organize that work.
Keep the uncertain relationship visible. Leave the next researcher enough evidence to disagree.
Next week: carrying a lease obligation from the source clause into the operating review.
- NextAutomation Team