# Item Count Recall vs Category Precision: Why Most Field Audits Measure the Wrong Metric

- **Canonical URL:** https://restorationintel.com/item-count-recall-vs-category-precision-contents-audits/
- **Section:** A-02 — Accuracy & Benchmarks
- **Field Brief PDF:** https://restorationintel.com/wp-content/uploads/2026/11/ri-item-count-recall-vs-category-precision-contents-audits.pdf
- **Markdown Twin:** https://restorationintel.com/item-count-recall-vs-category-precision-contents-audits.md

> **Direct answer:** Item count recall tracks whether every damaged object made it onto the schedule, whereas category precision tracks whether each logged line sits in the right carrier taxonomy, depreciation lifespan, and sublimit bucket — and auditing recall alone routinely masks 15 to 30 percent valuation distortion.

![Restoration QA reviewer inspecting a kitchen counter checklist beside packed numbered moving boxes](https://restorationintel.com/wp-content/uploads/2026/11/ri-item-count-recall-vs-category-precision-contents-audits-hero.webp)

Walk into most contents warehouses after a busy week of pack-outs and watch how managers spot-check accuracy: they open a box, count 22 wrapped items, check that 22 lines exist on the room sheet, and declare the inventory accurate. That spot check proves physical containment, not schedule precision.

In property claims estimating, an inventory line can be physically present yet financially wrong. Logging an upholstered heirloom dining chair as a generic folding kitchen chair, or bundling commercial espresso equipment into a residential small-appliances misc group, produces clean count parity on paper while distorting depreciation schedules, sublimits, and RCV pricing by thousands of dollars.

## Recall Failures vs. Category Precision Failures in Contents QA

| Metric | Definition | Where Crews Trigger Errors | Valuation Effect | Correction Play |
| :--- | :--- | :--- | :--- | :--- |
| Raw Count Recall | Detected items divided by true room count | Under-bed storage, closet top shelves, rear garage racks | Under-scopes total claim quantity; causes missing-item policyholder disputes | Mandatory cabinet/closet open-door protocol |
| Dollar-Weighted Recall | RCV of detected items divided by true room RCV | Unrecognized specialty hobby gear, optics, pro audio, shop tools | A single 1% count miss can remove 15%+ of room replacement value | High-value sweep pass before bulk pack-out |
| Category Precision | Lines assigned valid carrier category tree / total lines | Dumping mixed drawers into 'Misc Household' or confusing furniture vs decor | Applies inaccurate depreciation tables or carrier flat-cap rejections | Enforce taxonomy rules before export |
| Quality Tier Precision | Lines matching builder/average/custom quality grade | Defaulting designer, custom, or commercial lines to 'Average' preset | Systematic 20–45% undervaluation on replacement schedules | Pair material/brand cues with tier rules |

![Infographic: Item Count Recall vs Category Precision: Why Most Field Audits Measure the Wrong Metric](https://restorationintel.com/wp-content/uploads/2026/11/ri-item-count-recall-vs-category-precision-contents-audits-infographic.png)

## Why Unweighted Count Recall Flatters Sloppy Capture

Human crews and vision models alike excel at seeing large, unobstructed objects in the middle of a room: sofas, beds, dining tables, and countertop microwaves. Where recall breaks down is behind closed cabinet doors, inside stacked plastic totes, and along packed closet rods. When a QA audit skips closed zones or tests mostly open living rooms, recall metrics look artificially high.

Meanwhile, two different recall misses carry completely different economic weight. Missing three plastic food containers from a pantry lowers raw recall three times as much as missing one commercial-grade stand mixer — yet the stand mixer represents ten times the dollar value and often sits near a replacement-vs-clean decision line. Weighting recall audits by room zone and dollar tier exposes where omission risk actually lives.

- **Zone-stratified sampling:** Measure open-room recall separately from enclosed storage (closets, cabinets, garage bays).
- **Miscellaneous box audit:** Any schedule where > 8 percent of total RCV sits in generic 'misc' lines fails category precision regardless of recall score.

## How Category Misclassification Triggers Depreciation and Sublimit Traps

In estimating platforms like XactContents and Symbility, category assignment is not cosmetic labeling — it is the lookup key for useful-life depreciation tables and special policy limit checks. Misclassifying a solid wood dining table as laminated particleboard, logging collectible firearms accessories under general sporting goods, or grouping prescription medical devices with bathroom sundries distorts actual cash value (ACV) holdbacks and misleads desk examiners reviewing sublimits.

For cleanability triage, wrong categories also break cleaning scope budgets. [TPA contents cleaning SLAs](https://restorationintel.com/tpa-contents-cleaning-sla-by-category/) and carrier fee schedules price soft goods, electronics, hard surfaces, and textiles on completely different labor formulas.

- Check sublimit-sensitive categories separately: jewelry, watches, furs, silverware, firearms, business property, and collectible art.
- Require material + construction qualifier tags (solid hardwood, veneer, stainless, wool) on any furniture or rug line over $300.

## Building a Dual-Metric QA Scorecard for Estimators

Instead of a pass/fail percentage, score internal audits with a two-part tuple: **Recall / Precision** — for example, 97% Count Recall (99% Dollar Recall) / 94% Category Precision. A crew scoring 99% recall with 78% precision needs taxonomy coaching or cleaner pre-export mapping rules, whereas a crew scoring 96% precision with 84% recall is leaving doors closed on site.

Teams evaluating computer vision or mobile barcode capture tools should demand the exact same split in pilot tests, as outlined in our companion brief on [designing a blind room audit](https://restorationintel.com/designing-a-blind-room-audit-test-contents-inventory-workflow/).

## Field Execution Checklist

- Report both count recall and dollar-weighted recall when auditing any pilot or live residential loss.
- Cap generic 'Miscellaneous' lot lines at strictly defined low-value thresholds with mandatory interior overview photos.
- Audit sublimit-triggering categories (jewelry, electronics, collectibles, BPP) for 100 percent category precision prior to carrier upload.
- Verify quality-grade presets (economy, average, high, custom) against visible material and brand evidence on major assets.
- Tie taxonomy mapping errors back to estimator template defaults so fixes persist across future jobs.

## Related Guides & Verified Authorities

- [A-01 Companion Guide](https://restorationintel.com/how-restoration-teams-measure-contents-inventory-accuracy/)
- [A-03 Companion Guide](https://restorationintel.com/designing-blind-room-audit-test-contents-inventory-workflow/)
- [F-08 Companion Guide](https://restorationintel.com/taxonomy-mapping-ai-object-labels-xactcontents-carrier-trees/)
- [Contents Cleaning SLAs by Category](https://restorationintel.com/tpa-contents-cleaning-sla-by-category/)
- [Which Xactimate Line Items Do Adjusters Actually Approve?](https://restorationintel.com/which-xactimate-line-items-adjusters-approve/)
- [Verisk XactContents Platform Overview](https://www.verisk.com/products/xactcontents/)
- [ANSI/IICRC S320 Standard for Professional Contents Restoration](https://iicrc.org/s320/)
- [Symbility Claims Support Portal](https://support.symbilityproperty.com/en/support/home)
- [Verisk Property Claims Solutions](https://www.verisk.com/solutions/claims/)

## Frequently Asked Questions

### Which matters more to adjusters: recall or category precision?

Adjusters notice missing big-ticket items (recall failure) when policyholders complain later, and they notice bad category precision immediately during line-by-line depreciation and LKQ review. High-performing packets need both.

### What causes low category precision in field inventories?

Time pressure on site leads crews to batch unrelated objects into generic 'kitchen items' or 'garage misc' boxes without separating electronics, specialty tools, or sublimited property.

### Can AI object detection solve category precision automatically?

Computer vision identifies broad object classes well, but still requires clean room context, narration, or estimator review to map raw visual labels into insurance-specific quality tiers and XactContents codes.

### How should mixed junk-drawer boxes be scored in audits?

Batching true low-value mixed hardware or stationery into a documented bulk lot with overview photos is acceptable, provided high-value tools, jewelry, or electronics are extracted first.
