A discovery toolkit that brings every blend tank, every batch, every margin lever into a single defensible view — for the plant manager, the procurement head, and the board.
OmniBlend Discover is the assessment layer of the OmniBlend platform. A self-contained toolkit that lets a plant manager, procurement head, or board bracket the value at stake from blend optimisation — in their plant's own vocabulary, against industry mid-segment benchmarks, with full traceability from shop-floor metric to a board-ready value pool.
Lube oil manufacturing sits between two worlds that don't speak the same language. The boardroom lives in ERP rupees per litre. The control room lives in SCADA tank temperatures and viscosity readings. Between them sits a decision void where margin disappears, batch by batch, without anyone being able to point to a number.
OmniBlend Discover is engineered so that the same data tells three different stories — without losing fidelity between them. Procurement reads in metric tonnes. Production reads in kilolitres. Finance reads in rupees-per-Cr. The toolkit speaks all three.
Where am I leaking margin and how do I prove it to my MD?
Six standardised loss categories — additive giveaway, spec giveaway, evaporation, FTR, off-spec, cycle-time — each carrying an Indian mid-segment benchmark. Compare your plant's reality to industry baseline in one screen. Drill into any blend to see exactly which category is bleeding.
What does this look like in tonnes, and what's it worth per metric tonne?
Every cost number is rendered in both KL and MT. Cost-per-MT and saving-per-MT shown as independent rows in every table. Use these numbers verbatim against vendor benchmark pricing when re-tendering base oil or additive package contracts.
Is this a ₹2 Cr opportunity or a ₹20 Cr one — and how confident are we?
Three scenarios per blend — Low (conservative), Expected (realistic), High (optimistic) — built on industry-best-practice gap-closure logic. Confidence range on every number. Range-bound cost analysis. An auto-generated executive brief that closes with a go / conditional-go / no-go recommendation.
The plant manager recognises every word. The CFO recognises every number. The board sees a defensible recommendation. That's the bar.
Every feature in OmniBlend Discover maps to one of these eight user goals. The toolkit was assembled from these cases first — code came second. Each card reads as a one-page story for a specific stakeholder, with the trigger, the action, and the artifact they walk away with.
The Discover tool ships with a 7-blend demo and a 10-blend industry master template covering the most common Indian lubricant categories — HD diesel 15W-40, PCMO 5W-30, hydraulic ISO 68, transformer IS 335, and others. Download the template, edit it in Excel to match your plant's reality, upload it back. Every numeric value is range-validated on import. Up to 10 blends supported.
The Workbench loads any blend and locks the manual baseline against historical data. Six sliders let you propose OmniBlend target values for additive giveaway, spec giveaway, evaporation, FTR, off-spec, and cycle-time. Save three scenarios per blend — Low, Expected, High — to bracket the range. Live cost recalculation as each slider moves.
Volume in production speak (kilolitres) and procurement speak (metric tonnes), with density-aware conversion per blend. Every cost table renders dual rows: one tagged KL, one tagged MT. Costs as ₹/KL and ₹/MT. Savings as ₹/KL and ₹/MT. No mental math. Procurement uses the MT row for vendor benchmarking; production uses the KL row for shop-floor reviews.
Click any blend in the Detailed scenario report. The page enters focus mode — every chart, KPI, heat-map cell, range-bound cost row, savings rate, and per-unit table reflects only that blend. A drill-down panel expands inline with cost detail, scenario summary, and loss-category breakdown. One button — "Back to full portfolio view" — restores the full picture.
Six metrics, plant-grade vocabulary, no data-science jargon. Each metric carries a tooltip with definition, Indian mid-segment benchmark range, OmniBlend best-practice target, and a worked rupee impact for the active blend. FTR is rendered with an explicit upward-better cue so the polarity inversion against the other five metrics is unambiguous.
The Analytics tab rolls up every saved scenario into four KPI cards (Expected pool, Confidence range, % of baseline cost, Verdict) and four charts (range bars, loss-category mix, doughnut composition, bubble scatter). A heat map visualises savings intensity across blends and scenarios. Range-bound cost analysis shows Worst / Expected / Best rupee outcomes side-by-side.
The AI Insights tab calls Anthropic's Claude API directly from the browser using a customer-supplied API key. Four scoped prompts — refinery profile, portfolio analysis, value-pool analytics, and a structured executive brief. The executive brief returns four sections: Headline finding, Evidence quality, Recommended next step, Key risk to manage. Plant-floor terminology throughout, anchored in Indian benchmarks, specific rupee values inline.
Five export buttons across two tabs. Excel workbook produces a six-sheet stakeholder report — Cover, Portfolio & Scenarios, Workbench Scenarios, Loss Categories, Range-bound Costs, Per-unit Costs & Savings — all with KL and MT columns where relevant. Two PDFs (portfolio report and analytics report) plus the executive brief PDF. Filenames stamped with refinery slug and timestamp.
OmniBlend Discover replaces vague modelling categories like "margin overdose" and "thermal loss" with the standard plant-grade vocabulary an Indian Group II/III blending operation uses on its monthly cost report. Every metric carries an industry benchmark range and a measurable cost driver.
| Metric | Polarity | Benchmark · Manual | OmniBlend target | Cost driver |
|---|---|---|---|---|
|
Additive giveaway
Top-treat dosed above recipe minimum, expressed as a percentage of additive package cost
|
↓ Lower better | 2.0 – 4.0% | 0.3 – 0.8% | Additive ₹/L |
|
Spec giveaway (KV + VI)
Finished property blended above the lower spec edge — KV100, VI, pour point centring
|
↓ Lower better | 1.5 – 3.0% | 0.3 – 0.7% | Base oil ₹/L |
|
Evaporation / Noack
Light-end vaporisation during blend tank heating · governed by ASTM D5800 Noack volatility
|
↓ Lower better | 0.3 – 1.0% | 0.2 – 0.4% | Batch volume |
|
First-Time-Right (FTR)
Share of batches passing QC at first attempt — no rework, no property correction
|
↑ Higher better | 92 – 96% | 98 – 99.5% | Rework ₹/batch |
|
Off-spec / downgrade
Batches failing QC and downgraded, sold as base oil, or rejected
|
↓ Lower better | 0.8 – 2.0% | 0.1 – 0.5% | Downgrade loss |
|
Cycle-time giveaway
Hours lost to slow QC cycles, manual property approvals, batch-hold pending re-sample
|
↓ Lower better | 5 – 12% | 1 – 3% | Tank-hour ₹ |
From signing in to shipping a board-ready brief — a single working session is all it takes. The toolkit is engineered for a discovery workshop with plant ops on one side of the table and the Pione team on the other.
These are the design decisions that won't show up in a sales deck but will be felt by every user. They're the difference between a tool the customer respects and one they tolerate.
No "margin overdose" or "thermal loss" — those are model-spec terms. Every label on every screen uses the standard Indian Group II/III plant report vocabulary: additive giveaway, spec giveaway, FTR, off-spec, evaporation, cycle-time.
Five of six metrics are lower-better. FTR alone is higher-better. The slider, the tooltip, and the mode preset logic all carry an explicit ↑ marker so the polarity inversion is impossible to miss.
Modes don't apply uniform reductions. They close 30% / 60% / 85% of the gap between baseline and industry best practice — so a transformer oil at 97% baseline FTR doesn't get the same target as a PCMO at 93%.
Every blend carries its own density. Cost-per-KL and cost-per-MT computed natively per blend — not by a portfolio-average shortcut. Procurement reads the MT row; production reads the KL row; they're looking at the same data.
Click a blend → the entire Analytics page filters to that blend, including the heat map, every chart, every cost table. A persistent banner shows what's focused. One button restores the portfolio view.
Density 0.7–1.05. FTR 50–100. Batch size 0.5–200 KL. Out-of-range values surface row-by-row in a specific error modal — the existing portfolio is left untouched until validation passes.
A self-contained toolkit that runs entirely in the browser. Customer data stays on the customer's machine. AI calls go directly to the customer's chosen AI provider using their own API key — no Pione middleware, no logged requests, nothing leaves the workstation unless the customer explicitly exports it.
Download the industry master template, edit in Excel, upload back. Same column structure, both directions. Test verified end-to-end in a real browser with real SheetJS — every numeric field round-trips losslessly.
The toolkit never shows a single Expected number without its Low and High companions. Worst-case cost is rendered alongside best-case. The Verdict KPI explicitly returns CONDITIONAL or NO-GO when the value pool is small or the band too wide.
Six loss categories, ten blends, one defensible number — the discipline of measurement, applied to the craft of blending.
At the close of the discovery workshop, OmniBlend Insights generates an editorial-quality report that walks the audience from headline finding to recommendations to next step. Visuals come from the saved scenarios. Commentary is drafted in plant-floor vocabulary. The CFO reads the cover; the implementation team reads pages four through eight. One artefact, three audiences.
↓ Download sample report