A delivery date your team can keep, with the evidence behind it.
PromiseNest checks stock, incoming supply and the plant schedule, date by date, before a date is given. When an order is short it lays out the options with their cost, and every promise is scored against what was delivered. For sales, customer service and supply planning teams.
1Reads
- The order: SKU, quantity, wanted date, customer
- Stock on hand by plant or warehouse
- In-transit shipments with ETA
2Decides
PromiseNest
- Which date to promise, from which site, and how sure to be
- Whether to split, expedite, move production or use another supplier when short
- Which orders get scarce stock first
3Produces
- Promise stored with its evidence and sources
- Confirmed date written to the sales order after approval
- Chosen option handed to the product or team that owns it
↺ Learns Promised against delivered, promise by promise back to 1 · Reads
The date the rep gave on Monday
A distributor asks for 1,200 cases by the 14th. The rep checks the lead-time table, sees seven days, and says the 12th. It sounds safe.
There are 800 cases in the warehouse and 200 on a truck. The other 200 need a production slot nobody asked for. The 12th becomes the 27th, the distributor's shelf runs empty, and the next order goes to someone else.
The rep at the next desk learned the opposite lesson. He pads every date by three weeks, and loses orders he could have won.
Why the usual fixes fall short
- Lead-time tables are set once and never checked against what is in stock, in transit or on the production plan.
- ERP availability checks look at stock but rarely ask whether the plant can make the rest, or whether the order is worth taking at this margin, and they never say when their data is stale.
- Nobody records promised against delivered by SKU, so the next promise is no better than the last.
Dates keep coming from a table, customers keep discounting them, and the business keeps paying to rescue promises it should not have made.
- RevenuePadded dates lose orders, and missed dates lose the next one.
- Customer trustOn-time, in-full slips, and penalties and fill-rate claims follow.
- MarginRescues cost expediting, split freight and overtime, and scarce stock goes first-come, whatever the margin or the account.
- Working capitalWithout dates they trust, teams hold extra stock as insurance.
What PromiseNest is, and how it fits
- What it is
- Order promising: a delivery date at order entry, built from stock, incoming supply and the plant schedule, with a confidence, its evidence, and options when the order is short.
- Who uses it
- Order desks and sales reps, customer service planners and supply planners, plus the sales and finance leaders who own on-time delivery and margin.
- What it does
- Runs three checks on every order: is it available (ATP), can the plant make the rest in time (CTP), and is it worth taking at this customer's margin and credit (PTP). It prices the options when short, abstains on stale data, and keeps a ledger of promised against delivered.
- How it fits your systems
- Sits beside order entry as an overlay on your ERP. It reads stock, in-transit, orders, credit and cost from SAP, Oracle or your ERP, and the make date from OpsNest. After approval it writes only the confirmed date back to the sales order. Connectors are configured per deployment.
- What is different
- The date comes from rules you can audit, never from a language model. It abstains when its data is old. Every option shows its cost and who would execute it, and every promise is scored afterwards.
PromiseNest in action
Follow one short order from entry to delivery: the three checks, the options, the approval and the ledger.
A simulated scenario on representative data. You choose the option and give the approval; the numbers follow your choices.
From data to decision to outcome
What goes in, what PromiseNest works out, what gets decided, what happens next, and how the result comes back.
- The order: SKU, quantity, wanted date, customer
- Stock on hand by plant or warehouse
- In-transit shipments with ETA
- Open production and purchase orders, and the plant schedule
- Customer terms, margin and credit
- Time-phased availability (ATP)
- Make-date check against the OpsNest schedule (CTP)
- Margin and credit gate (PTP)
- Freshness check on every read, with abstain
- Promise simulator with safe, balanced and aggressive dates
- Which date to promise, from which site, and how sure to be
- Whether to split, expedite, move production or use another supplier when short
- Which orders get scarce stock first
- Promise stored with its evidence and sources
- Confirmed date written to the sales order after approval
- Chosen option handed to the product or team that owns it
- Promised against delivered, promise by promise
- On-time rate and calibrated confidence per SKU
- Buffers tuned from late-delivery history
- "What should we have promised?" corrections
Capabilities, organized by the work they do
14 modules in 4 groups. Open a module for its inputs, outputs, users and an example.
Three checks behind every date
Every promise passes three checks on your own data: is it available (ATP), can the rest be made in time (CTP), and should we take it at this margin and credit (PTP). Each check is rule-based and repeatable: the same inputs give the same answer.
Available-to-promise (ATP)The first of the three checks: finds the first date that stock on hand, in-transit and open supply can cover the order, and adds your policy buffer.
- Challenge
- A lead-time table cannot tell you whether a date is available, makeable and worth taking, so the promise is a guess.
- Inputs
- Stock on hand, in-transit with ETA, open POs, near-term production output, your promise policy.
- Outputs
- A promise with date, confidence, evidence and sources; when short, the shortfall passes to the make-date check.
- Used by
- Order desk, sales reps.
- Value
- Dates grounded in actual supply, with the working shown.
- Works with
- Reads stock, in-transit and open supply from your ERP and from Stocki, TransNest and OpsNest, configured per deployment.
- In action · illustrative
- 800 cases on hand and 200 arriving on the 10th cover 1,000 cases by the 10th. With a 2-day buffer the promise is the 12th, at 0.90 confidence.
Capable-to-promise (CTP)For the part stock cannot cover, asks OpsNest one question: can we make this quantity by this date? OpsNest schedules it against the real floor and answers.
- Challenge
- A make date from a planning lead time ignores capacity, changeovers and QC release, so it is fiction.
- Inputs
- Shortfall quantity and wanted date; from OpsNest, line capacity, routings, changeovers, QC release time, the factory calendar and components.
- Outputs
- A feasible completion date, or a clear "not by then, here is the earliest", combined with the ATP date and its confidence.
- Used by
- Order desk, production planners.
- Value
- Make dates from the schedule, not a rule of thumb.
- Works with
- PromiseNest asks and never re-models the plant: the answer comes from the OpsNest CP-SAT scheduler. Without OpsNest, a conservative make-to-order rule you set applies.
- In action · illustrative
- The 200-case shortfall gets a slot on Line 1 on the 16th. After a 3-day QC release it can ship on the 19th, so the full order is promised for the 20th.
Profitable-to-promise (PTP): margin & credit gateThe third check: tests margin, credit and service terms before a promise goes out, and ranks orders that compete for scarce stock.
- Challenge
- Scarce peak-season stock goes first-come, and below-margin promises slip through.
- Inputs
- Customer terms, margin, credit headroom, SLA, account tier.
- Outputs
- A verdict that sends below-margin or over-credit promises to a person, the penalty exposure of an SLA breach, and a priority score that gives scarce stock to the best buyer first.
- Used by
- Sales manager, finance.
- Value
- Margin and credit are checked at the moment of the promise.
- Works with
- Margin floor, credit rules and priority weights are settings per company.
- In action · illustrative
- Expediting would leave a 17% margin against an 18% floor, so the gate sends it to the sales head.
Multi-site availabilityRuns the availability check at each plant or warehouse and picks the earliest date, with site priority as the tie-break.
- Challenge
- Reps only see stock at their default warehouse.
- Inputs
- Stock and supply by site, site priorities, transfer times.
- Outputs
- The best-site promise, with the other sites' dates shown for comparison.
- Used by
- Supply planners, order desk.
- Value
- Stock gets used wherever it sits.
- Works with
- Site priorities are a setting; transfer times come from TransNest.
- In action · illustrative
- Plant 1 can cover the 18th and the Pune depot the 13th, so the order is promised from Pune.
Decide the promise when the order is short
From a lookup to a decision: weigh the source, the lever, the risk and the customer, see the cost of each, and let a person choose.
Decision-to-promise simulatorRuns per-order what-ifs before you commit (split the shipment, expedite, move production, use an alternate supplier) and ranks them by date, cost and margin.
- Challenge
- Short orders get argued over on calls, with no shared view of source, date and cost.
- Inputs
- Baseline date, order value and margin; source options from Stocki, TransNest and SourceNest; make dates from OpsNest.
- Outputs
- Ranked options, each with its date, extra cost, margin effect, confidence and the product that would execute it.
- Used by
- Customer service planners.
- Value
- A short order becomes a choice with its price, in one view.
- Works with
- Each lever's date and cost comes from the product that would execute it: TransNest for expediting, OpsNest for moving production, SourceNest for an alternate supplier.
- In action · illustrative
- DC North can ship in 7 days, DC South in 9, and make-plus-supplier in 11. Split ships 1,000 cases now and the rest by the 25th at 2% extra cost; expediting the rest brings it to the 23rd at 8%.
Promise agentRecommends the best option from the simulator and explains why, in plain words.
- Challenge
- Planners weigh date, cost and certainty in their heads.
- Inputs
- Simulator options, ledger history.
- Outputs
- An advisory card. On approval, the choice is recorded with its approver and handed to the product or team that owns the action.
- Used by
- Customer service lead.
- Value
- Recommendations that show their reasons. A language model may write the explanation; it never sets the date.
- Works with
- Executes nothing itself. The owning product (TransNest, OpsNest or SourceNest) or your dispatch team carries out the chosen lever.
- In action · illustrative
- "Split ranks first: most of the order arrives two days early and the rest 11 days late, at a quarter of the cost of expediting."
Risk-tiered promisesOffers a safe, a balanced and an aggressive date, each with its confidence.
- Challenge
- Every customer and every risk gets the same date.
- Inputs
- On-time history from the ledger, a supply risk score, your risk appetite.
- Outputs
- Three dates, padded by delivery history and risk pressure.
- Used by
- Sales reps.
- Value
- Reps choose the risk knowingly.
- Works with
- Risk appetite is a policy setting per company.
- In action · illustrative
- Safe: the 19th. Balanced: the 15th. Aggressive: the 12th, for a customer who will take the risk.
Customer-aware datesMatches the tier to the account and shows SLA, penalty, margin and credit beside the date.
- Challenge
- Strategic accounts and one-off buyers get the same treatment.
- Inputs
- Customer terms and tier.
- Outputs
- A recommended tier per order: strategic accounts lean safe, low-priority orders lean aggressive.
- Used by
- Sales reps, key account managers.
- Value
- Your most important customers get the dates you are surest of.
- Works with
- Customer terms and tiers come from SalesI or your ERP.
- In action · illustrative
- A national modern-trade account gets the safe date; a spot buyer gets the earliest credible one.
Show the date and keep score
Put the date, its confidence and its evidence where the order is taken, then score every promise against delivery so the next one is sharper.
The promise card: date, confidence, evidenceEvery promise is a specific computed date with a calibrated confidence and the evidence behind it, shown where the order is taken, or a clear abstain.
- Challenge
- "Sometime next month" or a padded date gives the buyer no reason to believe you.
- Inputs
- The promise result: stock, make and transit legs, confidence, cost and approval.
- Outputs
- A promise card inside SalesI or your order-entry screen, an abstain card when data is stale, and the breakdown attached to the sales order.
- Used by
- Sales reps, order desk, customers' buyers.
- Value
- The buyer trusts the date because they can see how it was built.
- Works with
- —
- In action · illustrative
- 10,000 cartons wanted by 15 Oct: 8,000 from stock, 2,000 off the line on 10 Oct after a 3-day QC release, plus 3 days in transit, so ready on 13 Oct, two days inside the window.
Promise ledger & shadow rolloutRecords every promise against delivery, recalibrates confidence and buffers each cycle, and stages the rollout from shadow to assisted to auto-confirm.
- Challenge
- Nobody measures promised against delivered, and switching on an unproven engine in front of customers feels too risky.
- Inputs
- Promises and delivery events.
- Outputs
- Per-SKU on-time rate, average days late, calibrated confidence, tuned buffers and an accuracy report; a shadow period measured on your real orders before any customer sees a date.
- Used by
- Supply planners, sales leadership, CFO.
- Value
- Promise accuracy becomes a number you manage, and trust is earned on a measured track record.
- Works with
- Rollout stages: shadow (measure only), assisted (a person approves each promise), then auto-confirm above a confidence gate for the order types you agree. Everything below the gate goes to a person.
- In action · illustrative
- SKU-A: 7 of its last 9 promises on time, calibrated confidence 0.73, buffer 2 days. In the first weeks PromiseNest promises beside the reps and is only measured.
"What should we have promised?"Looks back at delivered orders and learns a per-SKU correction from the gap.
- Challenge
- Learning from your own promises without simply confirming them.
- Inputs
- Ledger history, at least three deliveries per SKU.
- Outputs
- A per-SKU date correction that skips promises which only confirmed themselves.
- Used by
- Supply planners.
- Value
- Corrections drawn from delivered results.
- Works with
- Trains on a schedule. A planner reviews each correction before it changes live promises.
- In action · illustrative
- SKU-A lands two days after its promise on the median, so the correction suggests promising two days later.
Trusted beside your ERP
Keep the ERP in charge, know your data is ready before go-live, and prove every change on your own past orders before it ships.
ERP overlay: read, decide, write back the dateA stateless overlay that keeps your ERP's own availability check and adds what it lacks: a profit and credit gate, a make date from the floor, confidence with evidence, and a learning loop.
- Challenge
- Teams fear that a new system means migrating data and disrupting the system of record, and a date the ERP confirms can still be unkeepable or unprofitable.
- Inputs
- Stock, in-transit, orders, credit, cost and price from your ERP through OData, CDS views or BAPI; the make date from OpsNest.
- Outputs
- The ATP, CTP and PTP result, the confirmed date written to the sales order after approval, and the outcome recorded in the Promise Ledger.
- Used by
- ERP and IT teams, order desk.
- Value
- Nothing ripped out or duplicated; your ERP stays the system of record.
- Works with
- Works with SAP, Oracle and other ERPs. Connectors are configured per deployment.
- In action · illustrative
- On SAP, the order's availability check runs as usual. PromiseNest adds the margin gate and the OpsNest make date, and the approved date lands on the sales order line.
Data readiness & freshness checkGrades the inputs that drive accuracy before go-live, and checks how old every piece of data is before a date is given.
- Challenge
- "Our master data isn't clean" is the usual reason accuracy quietly fails, and a stuck feed produces confident-looking dates that are wrong.
- Inputs
- Five input families (stock, time including QC release per SKU, capacity, money, policy) and the last sync time of each source.
- Outputs
- A green, amber or red readiness scorecard per input before go-live. At run time, data older than 24 hours cuts confidence by a fifth; older than 7 days, or missing, and the engine abstains, holds a conservative date and asks a person to decide.
- Used by
- Order desk, planners, ERP teams.
- Value
- You know the accuracy ceiling before you commit, and stale data never turns into a confident promise.
- Works with
- —
- In action · illustrative
- The stock feed stopped nine days ago. PromiseNest gives no date from it, holds the order date plus 30 days, and flags the order for a person.
Accuracy model & eval gateFour guards behind every date: deterministic computation, confidence calibrated on your own outcomes, abstain on thin data, and an eval gate that replays every logic change on your past orders.
- Challenge
- "An AI picked our ship date" makes any CEO nervous: black-box numbers nobody can defend to a customer.
- Inputs
- Your past orders and deliveries, a golden set of cases, every proposed logic change.
- Outputs
- Accuracy, abstain precision and calibration scores per change. A change that loses accuracy on your past orders does not ship.
- Used by
- Supply chain head, CFO, IT.
- Value
- A confidence number you can defend to a buyer and to the board.
- Works with
- No language model touches the date. When one is configured, it writes the agent's explanation only.
- In action · illustrative
- A new buffer rule is replayed on last quarter's orders before release. It promises earlier but misses more often, so the gate blocks it.
What the intelligence actually does
Five levels, from seeing what is happening to learning from it.
Stock, in-transit, open supply and the plant schedule on one dated timeline per SKU and site, with the age of every source.
The first date supply covers the order, the date the plant can make the rest, and a confidence calibrated on your own history, or an abstain when the data is stale.
Ranked options when short (split, expedite, move production, alternate supplier), each with date, cost, margin effect and confidence, at safe, balanced or aggressive risk.
After approval, writes the confirmed date to the sales order and hands the chosen lever to the product or team that owns it. Auto-confirm above a confidence gate is switched on only with your agreement, after a measured shadow period.
Promised against delivered per SKU recalibrates confidence and tunes buffers each cycle, and "what should we have promised?" corrections are learned from delivered orders.
Runs automatically
- Availability and make-date checks, with their confidence
- Freshness check, and abstain on stale data
- Pricing and ranking of the options
- Ledger rollups, calibration and buffer tuning, on a schedule
Always waits for a named person
- Choosing an option for a short order
- Any promise built on stale or missing data
- Below-margin or over-credit promises (sales head)
- Every date written to the sales order, until auto-confirm is agreed for that order type
What it helps you work toward, and how we measure it
We agree the baseline with you in writing before a pilot starts. These are the outcomes we measure.
Promises kept
Dates built from actual supply and the plant schedule, then checked against delivery.
Share of promises delivered on or before the promised date, per SKU, from the promise ledger.
Confidence you can take at face value
A stated 0.8 should come true about eight times in ten.
Calibration of stated confidence against delivered results (Brier score), per SKU.
No dates from stale data
Old feeds trigger an abstain and a person, instead of a guess.
Abstains raised, against promises later missed because of data problems.
Short orders decided on cost
Every rescue option shows its date, cost and margin effect before anyone commits.
Extra cost per rescued order and margin after cost, against a baseline agreed in writing.
A SKU with 7 of its last 9 promises on time carries a calibrated confidence of 0.73, that is (7 + 1) ÷ (9 + 2). If the next two promises land one on time and one two days late, it moves to 0.69, and the late-delivery history sets a 2-day buffer on remainders.
Integration, control and deployment
ERP & system integration
Reads stock, in-transit, open supply, credit and cost from SAP, Oracle or your ERP through OData, CDS views or BAPI, and the make date from OpsNest. Writes only the confirmed date back, after approval. Connectors are configured per deployment.
Data sources & connectivity
Every read carries its sync time. Each promise is stored with its evidence, sources and confidence, or an abstain.
Deployment & implementation
Container-based service with its own database and scheduled jobs, for your cloud or your own servers.
Configuration & business rules
A promise policy per company: safety buffer days, the confidence gate, margin floor, credit rules and risk appetite.
Roles, access & approvals
Sign-in with roles for reps, planners, customer service and sales heads; every approval carries the approver's name.
Security, governance & audit
Tenant isolation enforced in the database on every table, with role-based access on every action.
Human oversight & exceptions
Recommend first: nothing reaches the customer until a person approves, and stale data makes it abstain, not guess. Auto-confirm is earned per order type after a measured shadow period.
Scale & ongoing optimization
Rule-based date logic with no model in the date path, calculated per SKU and site, and replayed on your past orders before each change ships.
Questions about PromiseNest
Something else? Ask a specialist.
Do we have to replace SAP or our ERP's availability check?
What runs automatically?
Does AI decide the date?
Where does the make date come from when stock is short?
What if the early promises are wrong?
Discuss your order-promising challenge
A delivery date is a decision. PromiseNest makes it from actual supply and the plant schedule, says how sure it is, refuses to guess on old data, and keeps score afterwards.
Make-to-stock and make-to-order businesses selling through distributors or to large accounts, where on-time, in-full is a commercial promise.
Works alongside
OpsNest
When a line stops or a vendor slips, the plan and the POs keep up.
Explore OpsNest →
Stocki
Know what your stock costs you each month, and keep the right stock in the right place.
Explore Stocki →
TransNest
See the late truck before your customer does, and pay only the freight your contract allows.
Explore TransNest →