AI Smart Detection

LOP understands the structure and column layout of your packing-list files for you. When your file is unfamiliar — unusual headers, an unmapped required field, a combined-dimension cell — AI Smart Detection reads the headers and a few sample rows and proposes the correct column mapping, which you review before importing. On files LOP already recognizes with confidence, it doesn't need to run at all.

Throughout, the AI only ever suggests a mapping. It never changes your numbers — every value is parsed by LOP's deterministic import engine, so you can always see and check exactly what will be imported.

When AI Smart Detection Runs

You don't turn AI Smart Detection on or off. LOP decides when it's useful:

• On a clean, confidently-detected file — where the headers are recognized and every required field (name, length, width, height) is mapped — AI Smart Detection doesn't run. There's nothing to improve, and the import is instant.
• When detection is uncertain — unfamiliar headers, or a required field LOP couldn't map on its own — AI Smart Detection runs automatically and fills in the mapping for you to review.
• On demand — an "Improve with AI" button is always available on the mapping step for spreadsheet imports. If the automatic mapping looks off, click it to have the AI take a fresh look.
• Never on SAP or Oracle exports — fixed-format files (SAP IDoc/BAPI, Oracle WMS) are mapped exactly from their known field codes, by design. The AI is never involved, so a canonical field is never mis-guessed. See ERP/WMS Import.

What Structural Analysis Detects

Before any AI is involved, LOP analyzes every uploaded file's structure instantly:

• Header row — finds the real header row even if it isn't in row 1. Title rows, summary rows, and metadata above the header are identified and skipped.
• Footer rows — totals, averages, and summary rows at the bottom are excluded from import.
• Decimal separator — European format (comma as decimal: 1.200,50) and US format (period as decimal: 1,200.50) are detected automatically.
• Combined dimensions — a single column holding values like 1200x800x600 is identified and split into Length, Width, and Height (see below).
• Merged cells — Excel files with merged header cells are expanded automatically.
• Transposed layouts — detects when data runs across rows instead of down columns.

Reviewing the Mapping

When a mapping is shown, each column is labelled with its detected property (Name, Length, Width, Height, Weight, Quantity, …). Columns the AI proposed carry a small AI badge, and a colour indicates how confident the detection is:

Combined-Dimension Cells

If your file packs Length, Width, and Height into one cell, LOP detects it and splits it for you. The common forms — 1200x800x600, 1200 × 800 × 600, 1200*800*600 — are recognized automatically. Other separators (such as 120/80/60 or 120-80-60) and cells that carry a unit inside them (120cm x 80cm x 60cm) are handled by AI Smart Detection.

Units

LOP determines dimension and weight units in order of reliability:

1. Header hints — a unit written in the header (Length (mm), [cm], Gewicht (kg)) is used directly.
2. AI detection — the AI infers the unit from the header text and sample values.
3. Magnitude inference — with no unit anywhere, LOP judges from value ranges: 1.2, 2.4, 0.8 look like meters; 1200, 2400, 800 look like millimeters.
4. Default — if nothing settles it, millimeters is used.

You'll see a label next to the unit dropdown showing how the unit was chosen — for example "AI detected: meters", or an amber "Inferred from values — verify" when it was guessed from magnitudes. You can always override it with the dropdown.

Per-column units. A single column can carry its own unit that differs from the file default — for example a Height (cm) column in an otherwise-millimeter file. When that's detected, that column is converted from its own unit while the rest of the file uses the default.

Per-unit vs. total weight. If a weight column looks like a per-line total (its value tracks weight-per-item × quantity across the rows), LOP recognizes it and converts back to a per-unit weight, so a stacked total isn't mistaken for a single heavy item.

Import Verification Checks

After parsing — but before you import — LOP screens every file for physically implausible values and surfaces anything suspicious in the preview. These are advisories: they never block an import. They're there so a silent data or unit error doesn't reach your load plan unnoticed.

• Implausible density — LOP compares each item's weight against its volume. A carton the size of a washing machine that weighs 200 grams, or a shoebox that weighs 400 kilograms, usually means a wrong unit somewhere (dimensions read in the wrong unit, or a weight in the wrong unit). The item is flagged for you to check.
• Weight that looks like a total — when a weight column appears to hold per-line totals rather than the weight of a single unit, LOP points it out so you can confirm whether it should be divided by quantity.
• Unit / magnitude mismatch — when the dimension values are the wrong size for the unit you've selected (for instance, values that look like millimeters while the unit is set to meters), LOP flags the likely mismatch before a box ends up a thousand times too big or too small.

Ask AI to Review the Warnings

When the preview shows verification warnings, an "Ask AI to review these warnings" action appears. It sends the current mapping, the warnings, and the offending rows to the AI, which proposes a small set of column changes — each with a short reason ("values are cm-scale, not mm"; "weight tracks unit × quantity, so it's a per-line total").

You then choose:

• Apply suggestions — LOP returns you to the mapping step with the changes in place so you can review them, then re-run the preview yourself. Nothing is imported automatically.
• Dismiss — nothing changes.

Any column you edited by hand is left untouched — the review never overwrites your own decisions, and it tells you when it skipped a column for that reason. If the AI decides the warnings point to a data problem rather than a mapping one, it says so instead of inventing changes.

Mapping Profiles

When you import from the same system repeatedly, save the mapping once and reuse it. After a successful preview, choose Save profile and give it a name (for example, "SAP Weekly Outbound").

The next time you upload a file with the same columns, LOP recognizes the template — regardless of column order or accented characters — and offers to apply the saved profile in one click. Applying a profile is instant and uses no AI: it's a saved, deterministic mapping. Saving again on a recognized template updates the existing profile rather than creating a duplicate. Profiles are shared across your organization.

Supported Header Languages

LOP recognizes column headers in English, Turkish, and German directly, including common abbreviations and regional variants (Qty, Brüt Ağırlık, Bruttogewicht). Headers in French, Spanish, Italian, Dutch, Polish, and other languages are handled by AI Smart Detection, which reads the header text and sample data to map them.

Supported File Formats

Tips for Best Results

• Use descriptive headers. "Length (mm)" is recognized every time; "Col A" may need the AI, or a manual pick.
• Keep data clean. Remove images, charts, and decorative formatting before uploading.
• One item per row. Each row should be one item type with its dimensions and quantity.
• Check the preview. The verification warnings are your last, quickest safety net before a plan is built on a bad unit.

Next Steps

• Importing from Excel or CSV — the standard import workflow
• ERP Import — importing from SAP, Oracle, and other enterprise systems
• Item Properties — what each imported field controls