Skip to main content

How CryspIQ Works — From a File to an Answer

CryspIQ® takes data out of the systems you already use, gives every field a business meaning, checks it, and then lets anyone ask questions of it in plain language. This page is the whole journey in one place — read it first if you're new, then follow the links below into the detailed guide for each stage.

The stages happen in this order, because each one depends on the one before it:

Sources → Maps → Operations → Quality → Consume

Data is loaded (Operations) before it can be measured for quality (Quality), which is why Quality comes after Operations rather than before it.

The Five Stages​

StageWhat you do
SourcesConnect a system, or upload a CSV.
MapsSay what your columns mean, in CryspIQ's terms.
OperationsWatch the data move through.
QualityCheck it's good enough to trust.
ConsumeAsk a question, or watch a dashboard.

1. Sources — where data comes in​

Connect a system, or upload a CSV. You describe what a file contains once — the columns you expect — and every later file of the same shape is checked and loaded against that same description, so it needs no rebuilding. See the Sources guides for how to create a source and upload data.

2. Maps — what the data means​

Your column headings are yours; CryspIQ's are shared across every source. Mapping is where you say "this column is the customer's name" and "this one is the date it happened" — the step that turns a file into something answerable. You don't have to start from scratch: upload a file and Lumen, the CryspIQ® AI assistant, proposes how it fits your data model, asks when it isn't sure, and applies the high-confidence matches for you, leaving only the uncertain ones to confirm. See the Maps guides for how to create a map and apply data quality rules.

3. Operations — moving it through​

Behind the scenes, three engines pick the data up in turn: a connector pulls it in, the mapper applies your maps, and the load writes it into the warehouse. There isn't one single "processing" screen — Operations gives you three, one for each engine (Connector, Mapper, Load), and each shows whether that service is running and a live log of exactly what happened and where. Watch these screens while you're waiting for data to land, and check them first when it hasn't. See the Operations guides.

4. Quality — is it good enough to trust?​

Rules check each record as it's mapped, before it reaches the data model — so a problem is caught early rather than after a load has failed. A value that fails a check doesn't quietly continue — it's held in an open exception with the reason and the file it came from, for a steward to trace back to the source system and fix there, rather than patching it inside CryspIQ®; the exception closes itself when a corrected record with a trusted record ID arrives. The same happens when a record refers to something CryspIQ® doesn't recognise yet — a customer or product it hasn't seen in master data — it's held in the Parking Lot until that catches up, rather than being loaded against the wrong thing, and releases itself automatically once the missing record arrives. See the Quality guides.

5. Consume — getting answers​

Dashboards for the numbers you watch, a library of saved questions ready to reuse, an Enterprise Observatory that shows what data exists and who owns it, and the option to just ask a question in plain English — inside CryspIQ®, or from Microsoft Teams and Microsoft 365 Copilot, so you can ask from the tools you already use. Either way you get the same answer, because it comes from one governed interpretation of your data model. Every answer runs with your own permissions, so you only ever see data you're allowed to see, and it shows the evidence behind it — what it used, what it covers, and any related records still waiting in the Parking Lot. See the Consume guides.

Why Mapping Asks the Questions It Does — the 5W Model​

Any real answer needs five things pinned down. CryspIQ®'s own mapping targets go by more specific names — Person, Company, Asset, Location, an Event and a few others — but every one of them exists to answer one of these five plain questions. That's why mapping asks what a column means rather than simply accepting its heading:

QuestionWhat it means
WhoThe person or company it concerns — CryspIQ®'s Person and Company targets
WhatThe thing itself — a product, an asset, a service, an account
WhereThe place — CryspIQ®'s Location target
WhenThe date and time it happened
WhyWhat kind of event it was — and what it measured: an amount, a quantity, a percentage

Get those five pinned to the right columns and CryspIQ® can answer how — how many, how much, how often, how it's trending. That's the reason mapping asks what a column means rather than simply accepting your headings.

Good to Know​

  • Describe a source once. Later files of the same shape follow the same rules — no rebuilding required.
  • Nothing is silently dropped. Every failure is visible with its reason and held for a steward to fix or match correctly before it counts.
  • Answers improve with use, inside your organisation only. Asking questions in plain English gets more accurate the more your team rates and corrects answers. That learning stays inside your organisation and is never shared across customers.
  • Neither setup path locks you in. You can start with Guided Setup and open the full editor for a source later.

Learn More​