The buyer’s lawyer already wrote the query. You just need to type it into the search box.

A DD Query Dossier is a 40 row artifact, one row per DD question, that sits in the data room alongside the T-12 and the rent roll. Each row has five columns: question_id, english_query_string, matched_event_count, iso8601_ts_range, thumbnail_gallery_href. The english_query_string is literally the sentence the buyer's lawyer wrote in their DD email ('any package theft in the lobby in the last 90 days'), typed verbatim into Cyrano's natural language search box. The matched_event_count is the number of clips the model found. The thumbnail_gallery_href points to a folder of 480x270 JPEGs. It replaces the email thread because the buyer's rep can read the row in 10 seconds instead of a lawyer paragraph, and they have evidence attached, not a claim.

Exactly, and that is the gap. LoopNet, Crexi, CBRE DealFlow, Marcus & Millichap, Ten-X, and Zillow Commercial are inventory marketplaces. Their product is a filter on cap rate, unit count, submarket, year built. Their audience is buyers looking at 30 to 50 listings at once. None of them publish DD workflow content because DD historically happens in email after the LOI, not on the listing page. But the buyer's decision to even submit an LOI is influenced by how the seller looks prepared to answer DD. A one page dossier preview in the OM signals a seller who will close in 45 days, not one who will spend 14 days hunting for a DVR password.

The broker runs the query interface. The seller plugs in the Cyrano unit on HDMI in the office closet before the broker sends the OM. After that, when the buyer's lawyer emails the 40 line DD question list, the broker pastes each English question directly into the Cyrano search box, captures the result count and the thumbnails, and exports a row to the dossier. Typical throughput is 40 DD questions handled in one afternoon, versus a three day back and forth where the seller chases property managers and ex employees for DVR admin credentials.

DD question lists are short English sentences written by lawyers who have never seen the DVR. Common examples: 'any person loitering near the pool gate after 11pm', 'packages left unattended at the lobby for more than 30 minutes', 'vehicle dwelling in the parking lot after hours', 'tailgating at the pedestrian gate', 'tampering near the meter room'. Every one of those is directly typable into Cyrano's search box. The model matches the phrase against 90 days of indexed footage and returns a ranked clip list. Nothing reformats the lawyer's sentence into SQL or tile labels.

Yes, but the dataset is thinner. If the unit goes in on day 1 of the 21 day inspection window, the buyer has at most 21 days of indexed footage to query against. That is enough to answer time-of-day patterns but not seasonality. If the seller installed Cyrano 60 to 90 days before listing (the pre-listing tap playbook), the buyer queries a full operating quarter and the answers are stronger. Most mid market sellers we see now deploy at the same time the OM is drafted, so the dossier is populated by the time DD starts.

The dossier has a specific non applicable flag. If the DD question is 'confirm current property tax assessment', the row is marked question_type=financial, english_query_string=null, and the row links to the tax bill PDF instead of a clip gallery. The 40 row artifact is not a replacement for the entire DD answer set. It is the subset of DD questions that resolve to observable behavior at the property, which on a typical multifamily list is 10 to 18 questions out of 40.

The pre listing tap page is about the seller generating a measured indicator (incidents per 1000 resident days) for the OM. This page is about the DD back and forth after the OM is out. Different audience, different moment, different artifact. The pre listing page ships a summary number in the OM. This page ships a typed query log with evidence. Both can run on the same Cyrano unit, but the pre listing page serves the underwriting model and this page serves the lawyer email exchange.

Yes. The dossier is a static CSV plus a thumbnails folder. At close, the file transfers to the new owner as part of the data room archive. The Cyrano unit also stays plugged into the HDMI port and continues to accrue indexed events under the new owner, so the dossier becomes the first 90 day archive of a running log that now extends forward. A buyer's rep who wants to benchmark month 1 post close against month 1 pre LOI can pull both from the same dataset using the same English queries.

It is not a replacement for forensic review. It is a replacement for the lawyer paragraph that currently answers the DD question. A forensic review is commissioned after a specific dispute arises, takes two to four weeks, costs $5,000 to $15,000, and produces an affidavit. An English query is a 10 second answer to a DD checklist item during the inspection window, costs zero incremental dollars, and produces a count plus thumbnails. The dossier is upstream of forensics, not a substitute for it.

The dossier is designed to make that request unnecessary. Most sellers cannot produce DVR admin credentials on a five year old recorder, and the buyer's rep has no standard schema for reading a raw DVR anyway. When the dossier exists, the buyer's first move is to ask for an additional query ('also run: vehicle dwell at the side entrance between 1am and 4am'), which the broker runs in one minute and appends as row 41. The DD exchange becomes query based, not credential based.

Plate level identification ('the blue sedan with partial plate 7ABC'), specific individual re identification across multiple days, and audio based events all perform worse than object level queries. The dossier flags these rows as model_confidence=low and routes to a conventional timeline scrub. Everything at the object level, person with backpack, unattended package, vehicle dwell, loitering, tailgating, tampering, performs well at DD grade precision because these are exactly the classes the on device model is trained on.