A closing binder that leaves the table finished

The organisation

Ashcroft Title & Settlement is an eleven-person agency handling residential and small commercial closings across four North Carolina counties. Around ninety transactions a month. Two closers, four processors, a title examiner, a post-closing clerk, an escrow accountant and two principals.

The agency's product, at the end of every transaction, is a closing binder: the complete executed package, indexed, delivered to the lender, the buyer and the agency's own file.

The problem

A residential closing package runs to about forty-one documents. Some are executed at the table by three or four parties. Some are recorded and come back stamped days later. Some — the lender's instructions, the title commitment, the survey — never go near the table at all.

The binder has to contain all of it, in the order the lender's post-closing checklist gives, with an index. The lender wants the package within twenty-four hours of funding, and its checklist differs from the order documents are actually signed in, which in turn differs from the order the recordable set needs to be in for the register of deeds.

So three orders, one pile. The pile was assembled at the table, scanned as one job afterwards, and then reordered on screen by a post-closing clerk working from the checklist. That reordering happened whenever the clerk got to it, which in a busy month was the following week.

Late binders had a direct cost. The lender's own audit deadlines meant a package delivered outside twenty-four hours was flagged, and repeated flags affect an agency's standing on a lender's approved list. Ashcroft was averaging about six rejected or flagged packages a month, most for lateness or for a document filed out of the checklist order.

What they tried first

The agency first hired a second post-closing clerk. Throughput improved for a quarter and then settled back, because the constraint was not hands — it was that the correct order could only be reconstructed after the fact by someone reading forty-one documents to work out what each one was.

It then tried scanning at the table in checklist order, with the closer sorting the pile before scanning. This transferred fifteen minutes of sorting to the closing table, in front of the parties, and closers hated it. It also failed on every transaction where a document came back from recording later, because those arrive after the pile has been scanned and there is no clean way to insert them into a flat scan.

A third attempt was an index maintained as a spreadsheet alongside the pile. It was accurate when maintained and it was maintained about two-thirds of the time.

What changed

Each transaction type is a preset carrying the lender checklist's order. Documents are placed into the preset's slots as they are executed, at the table, rather than sorted into order afterwards — the slot is chosen once, by the person who knows what the document is, at the moment they know it.

Dividers are generated from the slots and the index from the dividers, so the index is derived from the binder rather than maintained beside it. A document that arrives late — a deed back from recording — goes into its slot, and the index and the numbering update from that.

The recordable set is extracted as its own document straight from the binder, in the register's required order, without a second assembly. That was the change that surprised the agency most: it had always treated the recording package and the binder as two separate jobs that happened to share paper.

Signatures are captured where they are captured, and the binder is flattened on export so the delivered package cannot be edited downstream.

How it was put in

One closer and one processor ran a single lender's transactions on the preset for six weeks. That lender was chosen because it had the strictest checklist and the shortest fuse, on the same reasoning Halvorsen used with its fussiest carrier.

The preset was built by the post-closing clerk — the person who had spent two years reconstructing the order by hand and therefore knew it better than the checklist did. It included four ordering decisions that were not in the lender's written checklist but had been the subject of previous rejections.

Rollout to the second lender took two weeks. The remaining lenders were done in a single afternoon each, from the first preset, once it was clear that a preset is mostly a reordering of the same forty-one things.

The result

Binders are delivered the same day as funding. The twenty-four-hour requirement stopped being a target and became slack.

Flagged or rejected lender packages went from about six a month to fewer than one. The agency was moved up a tier on one lender's approved list during the following review, which the principals attribute in part — not wholly — to the delivery record.

The second post-closing clerk was not made redundant. The role changed to title curative work, which is chargeable, and which the agency had previously outsourced.