A configurator that cannot hand sales an invalid quote

Complex products break spreadsheets and they break most quoting tools. A rep picks a fast processor, a customer asks for the small memory option, and nobody notices the two do not fit until manufacturing rejects the order. The fixes are all bad: a thick compatibility document nobody reads, a senior engineer copied on every deal, or a quoting tool whose rules live in code that only the vendor can change.

The cost is not just the occasional wrong order. It is the week lost every time a quote bounces between sales and engineering, and the discount given away to keep a customer who waited too long.

A product configurator is a guided form that turns a product's options and constraints into choices a rep can make safely. On Neblex, the form is driven by rules you write in a visual builder, and the same rules are applied when a configuration is saved, so the platform enforces what the screen shows.

Attributes are the choices. Each one has a data type and a control: a drop-down, radio buttons, a multi-select with chips, a number, a date, free text, a read-only display value, a hidden value used only by your logic, or a button that runs a calculation on demand. Options on a picklist can carry their own cost, list price, and lead time. Option lists can also come from a data table you maintain, filtered by earlier answers, so a pump attribute can show only pumps that fit the selected motor.

Rules react to those choices. There are eight kinds, each with its own job: visibility, recommendation, violation, pricing, bill of materials, defaulting, computation, and lead time. A rule is an IF tree of conditions, grouped with AND and OR, using 14 operators such as equals, greater than, contains, starts with, is one of, and includes for multi-selects. The THEN side hides or requires a field, shows a message, raises a violation, recommends a value, sets a value, adjusts price, adds parts to the bill of materials, contributes a lead-time component, or switches specific options off.

Functions cover what a rule cannot say in one line. They are short scripts, written in the platform, that run in a sealed sandbox with no access to files, networks, or the database. A function can read every selection, query your data tables, call an external service you have registered, and return the same effects a rule can: values, visibility, violations, prices, parts, lead time. The sandbox has published limits: 32 MB of memory, 500 milliseconds per run, 25 data queries, 10 external calls.

Evaluation runs on every change. Neblex works out which rules and functions depend on which attributes, runs them in dependency order, and repeats until nothing changes, with a ceiling of 12 passes. If two definitions keep flipping a value between them, that is reported as a convergence violation instead of freezing the screen.

Lead time is aggregated rather than guessed. Each option and each rule can contribute a component, and the components combine in one of three modes: take the longest, add them, or override. The rep sees one number and the sidebar shows what drove it.

The save gate is the part that protects the quote. A configuration can be added to a quote only when three things are true: no violations, no switched-off options in the selections, and no required field left empty. The same check runs on the server when the line is saved, so a stale browser or a hand-edited request cannot slip past it.

• Visual rule builder. Conditions and actions are picked from your attributes and options, so a rule never refers to something that does not exist. A test panel evaluates the rule against a sample configuration and shows the result, each condition's outcome, the actions produced, and a trace of what fired. You can load a real configuration from an existing quote line as the test case.
• Functions with a real editor. Syntax errors are marked inline before you save, and the same test panel runs the function and shows its log. Dependencies are declared, so the platform knows the order to run things in and rejects a reference to an attribute that does not exist.
• Matt, the Neblex AI Assistant. Ask Matt how to express a constraint and it explains the rule, or drafts the function for you to review and paste in. Matt advises and drafts; it never changes your models on its own.
• Layouts. Tabs, sections, rows, and fields are arranged on a 12-column grid per model, with collapsible sections and status badges. Two products can have completely different configurator screens on the same platform.
• Promotion between environments. Export a model as a portable package: attributes, options, rules, functions, and layout, with every internal reference rewritten to a name. Import it into production and a pre-flight check lists anything missing before a single record changes. Existing items are updated, new ones created, nothing deleted.

• Eight rule types covering visibility, recommendation, violation, pricing, bill of materials, defaulting, computation, and lead time
• AND and OR condition trees with 14 operators, including list membership and multi-select includes
• Four ways to express a condition value: a fixed value, a script, a data query, or a query built by a script
• Option switch-off rules that hide or grey out incompatible options, and allow-only rules for the reverse case
• Picklists fed from data tables, filtered by earlier answers
• Per-option cost, list price, and lead time
• Sandboxed functions with published memory, time, query, and external-call limits
• On-demand buttons that run an expensive calculation only when the rep asks
• Dependency-ordered evaluation with a convergence check
• Live price, discount, lead time, and bill of materials in the sidebar as choices change
• Inline recommendations the rep can apply with one click
• A save gate enforced on the server, not just on the screen
• Rule and function versions with diff and rollback
• A conflict audit that reports 19 kinds of contradiction between rules, functions, and layouts before anything reaches a customer
• Where-used search across rules, functions, attributes, options, tables, parts, and layouts
• Portable model export and import with a pre-flight check
• Diagnostics for administrators: which rules and functions ran, how long each took, and what they changed

The product is a rack server with a processor family, memory, storage, cooling, and a regional power option. Here is how the rules shape one configuration.

1. 1The rep picks a region. A visibility rule with an allow-only action keeps just the 230 V and 400 V power options enabled for EU orders. Every other voltage is greyed out with the reason shown on hover.
2. 2The rep picks an ARM processor. A second visibility rule switches off the 128 GB and 256 GB memory options, because those modules only ship with the x86 boards. The memory field now shows three choices instead of five.
3. 3Cooling is a multi-select. Two small rules, one in each direction, make sure liquid nitrogen and air cooling can never both be selected.
4. 4A function sets a hidden flag. If the processor is the top model, or memory is 256 GB, or a data-centre GPU is chosen, the function sets a hidden RequiresLiquidCooling value to true. Three later rules check that one flag instead of repeating the whole condition.
5. 5Storage comes from a data table. The storage picklist runs a query against your parts table, filtered by the selected chassis, so only drives that physically fit are offered.
6. 6Price and lead time move as the rep works. A pricing rule adds a surcharge when liquid cooling is required. The GPU option carries a 21-day lead time; the platform's longest-wins mode makes 21 days the configuration lead time, and a message in the sidebar says why.
7. 7A bill of materials builds itself. A bill-of-materials rule adds the cooling loop parts; a function reads the accumulated parts and adds the correct number of rail kits for the chassis. The sidebar lists every part with quantity, cost, and list price, with cost hidden from users whose role is not allowed to see it.
8. 8The rep tries to add the configuration to the quote. One required field, the service tier, is still empty, and the summary at the top links straight to it. Once it is filled, the Add to Quote button enables. The line lands on the quote with its options, parts, price, and lead time.

Before this model went live, the administrator ran the conflict audit, which flagged that a defaulting rule was setting memory to 256 GB in the same condition where the ARM rule disables it. The fix took a minute, and it was found in the audit rather than by a customer.