Insight
Procurement intelligence: on-demand procurement analysis, verifiable figure by figure
Procurement intelligence means analysing your own purchasing data on demand: spend, prices actually paid, contract coverage. Insight turns your question into a query over the graph and the procurement database, computes it, and opens every figure onto the lines behind it, without filing a request with BI (business intelligence).
Insight
The question is never “how much did we spend”. It is “how much, with whom, under which contract, and how long has it been drifting”.
The question comes up on a Tuesday, before a renegotiation or a board meeting. It crosses several objects: a supplier, its legal entities, a contract, a rate schedule, twenty-four months of invoices. No ERP (enterprise resource planning) export covers it in one piece.
Book a demo
Customer casesNothing is lost. Everything can be checked, everything can be proven.
- The contract clause and the invoice line, highlighted side by side.
- Every extracted value stays linked to the exact place in the document where it was read.
- The same case produces the same decision, today as in six months: the rules are applied deterministically.
- No discrepancy is set aside in silence. Anything that matches no rule is raised, with its reason.
- The agent records what it did, in the order it did it: who, what, how much, when.
Procurement intelligence · Insight
Ask the question. It builds the analysis, and shows where every figure comes from.
Translates a question asked in plain language into a query over the graph and the procurement database, computes, and returns the useful form: a figure, a table, a curve, a dashboard. It reads, it does not write. Every value opens onto the lines that make it up, and every line onto its source document.
What Insight detects
- A category whose price has been drifting for months
- A contract coverage rate falling, by category and by entity
- Credit notes outstanding, with whom and for how long
- A single-source reference with no alternative on record
- A gap between the budget for a cost line and what was actually invoiced
- An increase driven by mix or volume rather than price
- A missing figure: declared as missing, never estimated
The delay between the question and the answer. A buyer preparing a renegotiation no longer waits for a slot with the data team.
The six families of intelligence, and the scope we own
The procurement intelligence market mixes six families of data under one word. We say which ones we cover, and which ones we do not: an acknowledged boundary beats a vague promise.
Spend intelligence
How much, on what, with whom, by entity
Pricing intelligence
Prices paid, history, gap to what was negotiated
Supplier intelligence
Performance, adherence to commitments, concentration
Category intelligence
Cost structure of a spend family
Market intelligence
Indices, commodity prices, macroeconomics
Competitive intelligence
Other companies’ purchasing strategies
We do not sell market data. We make yours queryable, and verifiable.
The last two rows are not a gap to be filled later. Commodity indices and your competitors’ purchasing strategies are third-party data, with their own publishers and subscriptions; we would rather say so than let you assume otherwise. Everything your own documents contain, however (a price paid, an indexation clause, a credit note never applied), we can find and prove.
How it answers without inventing
This is the first objection a finance director raises, and it is a fair one: an analysis that is wrong but well phrased costs more than no answer at all. Here is the mechanism, in order.
The question becomes a graph query
Not a text search. Graph relationships are explicit (supplies, governed by, priced by, charged to, belongs to group), so the joins are exact and the path taken stays readable.
The answer is grounded in connected facts
Not in similarity of wording. That is what separates an answer built on a graph from an answer built on a vector search, which brings together passages that look alike without knowing what is connected to what.
Two levels of proof
The figure opens onto the graph relationships that produced it, and each relationship onto the document line it came from: document, page, line, date.
What is not known is declared as such
A question about missing data gets “I do not have that information”, never an estimate. A plausible but unfounded figure is the worst possible outcome: it travels, and nobody remembers it was an assumption.
The query is kept
A useful analysis becomes a dashboard, versioned, replayable the following month on the same rules. The second time, the question is not asked again: it is re-run.
Two levels of proof, side by side
On the left, the aggregate as it appears. On the right, its breakdown down to the source invoice line. It is the same move as the clause and the line highlighted side by side for the Compliance agent, applied to an aggregated figure.
The figure returned
- Packaging category spend
- Q1 2026
- €412,300
What you read on the dashboard.
What sits behind it
- 1,284 invoice lines
- including invoice F-2026-0912, page 3, line 18
- €12.40 × 340
What you get when you open the figure, all the way to the source document.
No aggregate is returned without that second level. A figure whose lines cannot be shown is not a figure: it is a claim. And a claim cannot be put to a supplier.
Why grounding in a graph matters
Querying a language model without giving it anything to verify against does not produce an analysis; it produces convincing phrasing. Two public market benchmarks put numbers on the gap.
- Around 59% error
- On multi-hop questions spanning several entities, a language model queried with no retrieval mechanism gets it wrong in around 59% of cases, according to work presented at the Agent+Graph workshop at VLDB 2026.
- 30 to 40% more accuracy
- The 2026 market benchmark expects a graph-based approach to deliver 30 to 40% more accuracy than a plain vector search on relational questions.
That is why Insight does not query a document corpus but the graph built by Graph: a graph where suppliers are resolved under their group, references aligned, units converted, and where every edge carries the document it came from.
The questions actually asked
None of these questions can be answered with an ERP export: they all cross several objects, entities or periods. These are the ones we hear in demos.
A buyer
- How much have we spent with this supplier, across all entities, over 24 months?
- What average price do we pay for this reference, by site, and what is the gap to contract?
- Which categories have been drifting fastest over the past six months?
- Which references am I single-sourced on, with no alternative on record?
- What is the real total cost of this contract, ancillary fees and surcharges included?
- What should I prepare before next month’s renegotiation?
A finance department
- What is my contract coverage rate, by category and by entity?
- How many credit notes are outstanding, with whom, and for how long?
- If the index this contract is tied to rises by three points, what is my exposure?
- What share of spend was committed off contract this quarter?
- What is the gap between the budget for a cost line and what was actually invoiced, and where does it come from?
- What explains the increase in this category: volume, price, or mix?
Native dashboards, and export to your BI
Insight delivers in two ways, depending on where your reporting already happens. Both start from the same reconciled data.
- Native dashboards
- A useful analysis does not stay a conversation: it becomes a dashboard inside Zylio, versioned and replayable. The following month, the same rules produce the same figure on up-to-date data, and the previous version stays available, which lets you explain a variance instead of absorbing it.
- Clean datasets for Power BI or Tableau
- For organisations that already report in a BI tool, Insight pushes reconciled datasets: supplier entities resolved, references aligned, units converted, spend tied to its contract. You copy nothing and you migrate nothing.
Keep your reporting tool. Power BI and Tableau present clean input very well; the hard work sits upstream, in the reconciliation, and that is the part we do. In the meantime, Insight answers the questions that do not justify building a report, and there are far more of those than there are reports.
What Insight does not do
Insight reads, it does not write. No writes to the ERP, no commitment made, no supplier contacted. No data leaves your environment and none is used to train a model: the graph and the database it queries are specific to your organisation and hosted in France.
What Zylio does not do
The Insight cluster
Five pages go into what this one summarises: the questions to ask, the export to your BI, the two most requested analyses, and the comparison with a consulting engagement.
- The questions to ask about your spend, and what it takes to answer themTwelve buyer and finance questions, and the data each one demands.
- Connecting your procurement data to Power BI, without copying itReconciled datasets pushed to the reporting tool you already use.
- Contract coverage rate: how much of your spend is really under contractHow it is calculated, by category and by entity, and what it reveals.
- Price-volume-mix analysis: where the increase really comes fromSeparating the price effect, the volume effect and the mix effect on a category.
- Procurement analysis in-house or by a consulting firm: what changesTurnaround, scope, repeatability: the honest comparison.
Frequently asked questions
What is procurement intelligence?
The analysis of your own purchasing data (spend, prices paid, supplier performance, contract coverage), obtained on demand rather than at the end of a reporting project. At Zylio it works on your documents and your graph, never on market data bought from a third party.
How does it differ from procurement analytics?
Procurement analytics refers to the tooling: the dashboards and reports a data team models in advance. Procurement intelligence refers to the ability to answer a question nobody planned for, by crossing several objects. And the order matters: certainty about the past first, because a wrong figure is not fixed by a forecast.
Where does the data come from?
From your own documents, already read by the agents: contracts, amendments, unit price schedules, rate cards, purchase orders, goods receipts, invoices, credit notes. Graph connects them into a graph where suppliers are resolved and units converted; Insight queries that graph and the procurement database.
How do I verify a figure?
Every value opens onto the lines that make it up, and every line onto its source document: page, line, date. It is verifiable, not declarative. And what is not known is declared as such, never estimated.
Do we have to replace our Power BI?
No. Keep your reporting tool: Insight pushes clean, reconciled datasets to it, and in the meantime answers the questions that do not justify building a report.
Can Insight modify data?
No. It reads, it does not write: no writes to the ERP, no commitment, no supplier contact. No data leaves your environment and none is used to train a model.
Measurable impact in every environment
More than 5 million procurement documents analysed
Between 1 and 7% of margin recovered
on the scope analysed
From 15 to 45% of time given back to teams, per FTE
depending on the scope and on data maturity
Zylio fits into your existing ecosystem.
The ERP runs the process. Zylio handles the exception and recovers the value that escapes it: invoices without a purchase order, line-by-line price discrepancies, duplicates and overbilling, off-contract spend.
IA
Your data under high security.
- Certifications
- SOC 2 Type II · ISO 27001
- Hosting
- Hosted in France
- Encryption
- End-to-end AES-256 encryption
- Access
- Enterprise SSO · multi-factor authentication · Zero Trust approach
The five Zylio agents
- CaptureGather and consolidate all your procurement data.
- CompareCompare supplier offers, objectively.
- ComplianceEvery invoice checked against your agreements.
- MatchingPurchase order, goods receipt and invoice matching, automated.
- ExceptionInvoices without a purchase order, multi-order invoices, duplicates: handled.
Procurement intelligence: the two intelligence layers
See what this looks like on your own data
Twenty minutes, on a spend category of your choosing. We show you what the agents detect, with the evidence behind it.
- No commitment, on your own data
- Result in 3 weeks
- 20 minutes, no sales pitch
- Your data stays hosted in France
- No change of tool or process