Unit Turn Data Study

Decisions needed

Trailing 12 months at a glance

Turn time by month · the progressive view

The trailing-12-month story: a stable fall (medians 15 to 21), a December bump, then February 2026, the worst month in the window (median 35 days on the highest volume in the dataset, 43 move-outs) with the hangover tail landing in March (P90 of 85 days). Since April: 21 → 15.5 → 14, a genuine improvement run, and June 2026 is the best fully-settled month in the entire dataset. Read the hatched July/August bars as incomplete, not as 9-day turns; the honest current-state claim is median mid-teens and improving. The walk queue held at 3 to 7 days all year, so the gains are coming from the work phase (21 median work days in February down to 7 in June).

Monthly detail, trailing 12 months

* Cohorts still finishing; only fast turns have closed, so these rows are biased low and will drift up. Cost is average recorded cost per turn, over the turns where cost was captured (n in parentheses): cost entry stopped during the March-to-June system transition and restarted in the native flow in July, so those months reflect capture, not true spend.

Distribution of days vacant, trailing 12 months (358 finished turns)

Pre-leased vs not · the prioritization question

Cohort: 182 turns with move-outs since March 12 (where daily snapshots make pre-lease status observable). 108 became pre-leased during the turn, 74 did not. Complexity is identical between groups (median scope weight 9 in both, similar cost), so the speed gap is prioritization, not easier units.

Raw examples · the pre-lease reaction, 2026 finished turns

The two highlighted rows are the management story: most pre-leased units get a 3-to-14-day response, then a tail nobody catches (Beaux House #510 waited 22 days to start after pre-leasing; Burnett Lofts #107 sat 57 days between pre-lease and work start) because nothing surfaces "pre-leased and not started."

A normal month of turns · raw examples

June 2026 move-outs, first 15 finishers, full date chains:

Even in the best settled month the spread runs 3 to 30 days, and the slow tail is dominated by late starts (Beaux House #400: walk done 06-05, work not started until 06-16), not slow work.

Walks

Walks are not slow; getting to them is. A 20-minute task with a 5-day queue is a scheduling habit, not a capacity problem.

Seasonality · benchmark context from the full dataset

Verdict: seasonality is real but moderate, and event-driven rather than smooth. Two patterns survive the year-by-year check: a December-through-March winter drag (the work phase itself stretches to a 22-day median in December vs 9 to 11 in summer, consistent with holiday staffing, vendor availability, and weather), and spike months that are weather/volume events rather than calendar constants (February spans 36 / 18 / 35 across the three years). Summer brings more turns but they run fast; winter brings fewer turns that each take longer. The dangerous scenario is a high-volume winter month, which is exactly what February 2026 was. Staffing implication: confirm vendor lineups and parts stock in November, not February.

Complexity

No true complexity score exists yet. Task scope weight works as a proxy today, and it tracks reality. Trailing 12 months:

Full-history benchmark shows the same gradient with more heavy volume: weight 9 runs 16 d / $343 (n=360); 10–12 runs 20 d / $513 (n=344); 13+ runs 36 d / $1,832 (n=43).

Recommended working tiers until tagging matures: Standard · weight ≤9, no red tag Heavy · 10–12 or orange tag Major · 13+ or red tag. That makes the technician leaderboard fair: a tech clearing hoarding units should not lose to one clearing standard turns.

Tracking it going forward

Already built, use today: the property reports tab (current-status KPIs, walk-pending aging, days-vacant distribution, a Pre-Lease Move-Ins at Risk card, oldest-sitting list, per-technician performance) and the supervisory report (portfolio scoreboard, turn times, people, cost, trends, PDF unit summary). Aging thresholds and hold reasons with timestamps are live; a hold deliberately does not stop the vacancy clock.

Gaps to close (route to Amber, none large):

1. Persist pre-lease onto the turn record before snapshot retention erases history (decision 3). Also powers the missing alert: "pre-leased and not started in X days."
2. Punch completion discipline: "unit ready" is only stamped when the punch walk is completed in SOS. Process first, software second.
3. Per-item walk timestamps are never written (column exists, 0 of 46,260 items). Trivial fix for true walk durations.
4. A dozen hand-created turns lack a move-out date (Belleview, Davis Square, M Streets) and can never age correctly. One-time cleanup plus a required field.

What to manage weekly

1. Punch-requested queue, oldest first. Today's bottleneck, entirely inside maintenance's control.
2. Pre-leased units against the move-in buffer. Not finished 3 days ahead of move-in = the day's priority list.
3. Unleased units past 21 days vacant. The drift candidates; nobody pushes them without a lease attached.
4. Walk-pending past 4 days. A 20-minute walk with a 5-day queue is a scheduling habit, not capacity.
5. Technician throughput within complexity tier once tagging starts.
6. Hold honesty. Only 4 units on hold right now; legitimate waits should say so, or slow turns hide behind them.
7. Winter readiness. December through March is when the process breaks, especially under volume. Confirm vendors and parts in November.