Home · Solutions · Procurement
Solution · ProcurementEach supplier's scorecard is built from the systems that already hold the facts
Supplier reviews that start from data, not from impressions
Robots assemble every supplier scorecard from the ERP, the quality system and the invoice history, publish it in Power BI, produce the review pack and track the agreed actions.
Executive summary
The numbers in your supplier review were exported by hand last week, and nobody can show the lines behind them.
A scorecard is a calculation, not a judgement call, which is why it belongs to robots.
The pack is ready on the first working day after quarter close, so the review discusses the quarter instead of waiting to be assembled.
Power BI semantic model and paginated report; SharePoint scorecard store and pack archive; Microsoft Planner action list
Business problem
Supplier management
A supplier review is a management meeting with a preparation problem. The facts that should decide it live in four places: delivery dates and receipts in the ERP, complaints and 8D cases in the quality system, blocked invoices and credit notes in accounts payable, price history in the category manager's own file. None was built to be read alongside the others, and none agrees on what a supplier is called or when a delivery counts as on time.
So the pack is assembled by hand, by the person who will also chair the meeting. Two and a half hours per supplier goes into exports, reconciliation and rewriting last quarter's slides; the analysis worth paying a category manager for is whatever is left. Quality sends a complaint count by email, finance answers about disputes if asked, and each plant has its own view of whose fault the late deliveries were.
At scale this decides which suppliers get reviewed at all. A base of 220 is managed as 60 that can be prepared and 160 that cannot, so the tail is judged on impressions. Actions land in minutes read once, and the next review opens by asking the supplier what happened to them. The evidence is not durable: it survives neither a comparison across quarters nor a handover to the next category manager.
How it works today
- PersonA category manager exports twelve months of purchase orders and goods receipts from SAP and cleans the list in Excel, plant by plant
- PersonQuality is asked by email for complaints and 8D cases; the count arrives in a different shape each quarter, and each plant counts rejections its own way
- WaitingFinance answers on blocked invoices and credit notes when the month-end queue allows, typically several days later
- PersonLast quarter's deck is copied, the charts repointed at the new workbook and the commentary rewritten
- Risk of errorThe supplier disputes a figure; the underlying lines are in a workbook nobody has open, so the point is noted and dropped
- PersonAgreed actions are typed into the minutes and mailed round; nobody owns that list between reviews
- WaitingThe next review opens by reconstructing what was agreed last time, from the minutes and from memory
Why the current process costs more than it appears
The bill that never reaches the budget.
- Preparation crowds out the analysis it was meant to serve. Two and a half hours of assembly buys the same slides as last quarter, while the questions worth asking, why one supplier's prices drift and another's complaints cluster on a single line, wait for spare time.
- Definitions drift between plants and quarters. Where on time means the confirmed date in one plant and the original request date in another, the score is not comparable, and the first ten minutes of every review go on method.
- Suppliers learn that numbers can be talked down. A figure with no traceable records behind it is an opinion, so the supplier who challenges hardest scores best, which nobody intended to reward.
- Actions decay between reviews. A commitment in a Word file has no owner, no date and no reminder, so part of every outcome is quietly rebooked into the next quarter.
- Coverage is capped by preparation capacity, not by risk. The unreviewed suppliers are not the safe ones, and a single-source part among them is where the next line stoppage starts.
Cost of inaction
Nothing here breaks, which is why it survives every reorganisation. Reviews take place, packs get built, suppliers are managed on whichever numbers were easiest to export, and the arithmetic above is only what shows up as salary. Beside it sits the cost of the 160 suppliers nobody scores.
What grows is the evidential exposure. A customer auditor, a resourcing decision and a claim against a supplier all need a history that can be produced and defended, and one kept in decks on personal drives cannot be produced at all.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A tier-1 automotive supplier, five plants in Poland, Czechia and Germany, 3,200 employees, SAP S/4HANA and Microsoft 365 E5; six category managers and four supplier quality engineers cover direct material, with quality data in a separate CAQ system.
220 direct-material suppliers, of which 60 are reviewed quarterly and the rest once a year; 20 scorecards a month, each drawing on twelve months of purchase orders, goods receipts, complaints, blocked invoices and price changes.
Exports from SAP, a complaint count by email from quality, a dispute list from accounts payable when there is time, and a deck copied from last quarter; actions live in minutes on a shared drive.
About 150 minutes of assembly per scorecard, mostly reconciliation rather than judgement, and packs finished the evening before the meeting; roughly 160 suppliers are never scored.
Robots pull the source records on a fixed calendar, apply the definitions and weights procurement owns, store every score with the lines that produced it, publish in Power BI, export a pack per supplier and turn agreed actions into Planner tasks.
In the modelled case the pack is ready on the first working day after quarter close, assembly falls to reviewing flagged exceptions, and both sides can trace the numbers they argue about. Nothing here was measured at a client.
Proposed solution
A scorecard is a calculation, not a judgement call, which is why it belongs to robots. On a fixed calendar an unattended UiPath robot reads records that already exist: purchase orders, confirmed dates and goods receipts from SAP S/4HANA, complaints and 8D cases from the quality system, blocked invoices and credit notes from accounts payable, and price history. Each metric has one written definition, and definitions and weights sit in SharePoint lists procurement edits under version history.
Every score keeps the evidence that produced it: delivery reliability of 91.4% is not a number on a slide, it is 218 receipts of which 19 arrived after the confirmed date. Before publication the draft reaches the category manager as a UiPath Action Center task completed in Microsoft Teams: exclude a disputed event with a reason, add commentary, release. Scores cannot be overwritten by hand, and exclusions stay visible to the supplier.
Publication has two audiences. Internally the scorecards are one Power BI report over the scorecard store, with row-level security so a plant buyer sees their own plants and the procurement director sees everything. The supplier gets their page as a paginated report exported to PDF and filed in a SharePoint folder shared with a named contact; a live guest view is possible where licensing allows. The review is a Teams meeting whose agenda the robot posts with the pack attached, and whatever is agreed becomes a Planner task chased into the next quarter's pack.
UiPath Orchestrator time triggers, queues and audit; UiPath Integration Service connectors for Microsoft Teams and Microsoft OneDrive & SharePoint; UiPath Action Center actionable notifications in Teams; Power BI row-level security and paginated reports; Microsoft Planner tasks through Microsoft Graph
The metric definitions and data-quality gates, the weighting model and its versions, the scorecard store with its evidence lines, the release step, the Power BI report and supplier page, the review pack, the action loop in Planner
SAP purchase-order, goods-receipt, invoice-block and price extracts through UiPath SAP activities (BAPI/OData); the quality system through a database view or scheduled export; pack generation through the Power BI export API, which needs a Fabric capacity workspace
How the automated process works
- AutomationA time trigger in Orchestrator starts the run after quarter close; the robot reads purchase orders, confirmed dates, goods receipts, invoice blocks and prices from SAP, and complaints and 8D records from the quality system
- SystemData-quality gates run first: unmapped supplier numbers, receipts with no confirmed date and out-of-scope plants are held back and reported rather than scored
- AutomationEach metric is calculated on its written definition, weighted by the model in the SharePoint list, and stored with the lines that produced it
- PersonThe draft reaches the category manager as an Action Center task in Teams: exclude a disputed event with a reason, add commentary, release
- AutomationThe Power BI report refreshes; row-level security shows each viewer their own suppliers and plants, with four quarters of trend beside the current score
- AutomationThe pack is exported per supplier as a PDF, filed in SharePoint and attached to the review invitation, and the agenda is posted in the category channel
- AutomationAgreed actions become Planner tasks with an owner and a due date; the robot reminds owners before the date and lists what is still open next quarter
Human-in-the-loop model
Automation handles
- Reading the source records on the calendar, across every plant and system, with no export request to anyone
- Calculating each metric on its written definition and applying the agreed weights, identically for every supplier
- Storing the evidence behind each score, refreshing the report and producing each supplier's pack
- Creating, reminding and reporting on the actions agreed in each review
People decide
- Which events are excluded from a score, and with what reason on the record
- What the metrics mean and how they are weighted; the model belongs to procurement, not to the robot
- The commentary, the ranking and what happens to a supplier who does not improve
- Which suppliers belong in the quarterly programme and which are reviewed yearly
Before and after
Systems and integrations
The stack is deliberately short: one engine, one execution layer, one place where a person decides.
Inputs
- SAP S/4HANA purchase orders, goods receipts and invoice blocks
- complaints and 8D records from the quality system
- info-record price history
- premium-freight log
- SharePoint lists holding definitions and weights
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Action Center
Target systems
- Power BI semantic model and paginated report
- SharePoint scorecard store and pack archive
- Microsoft Planner action list
Human touchpoints: Action Center draft review in Teams; the category channel; the review meeting itself; the supplier's own page or PDF pack
Technologies used
pull the source records on schedule, run the scoring, retry, log and audit
Areads definitions and weights, writes the scorecard store, files the packs
Acategory managers review, annotate and release each draft scorecard
Aposts the agenda, the scorecard and action reminders to the category channel
Athe live scorecard, the supplier's own page and the exported review pack
Aagreed actions become tasks with an owner and a due date, read back by the robot
Apurchase orders, confirmed dates, goods receipts, invoice blocks, price history
AIllustrative economic model
Numbers you can check against your own data.
Two and a half hours per scorecard is assembly, not thinking: four extracts, reconciling supplier numbers across plants, rebuilding last quarter's charts, writing commentary. €35 an hour is a fully loaded cost for a category manager or supplier quality engineer in Central Europe, above a clerical rate because these are not clerks. Nothing below was measured at a client, and the model leaves out what the reviews decide: dual sourcing, price recovery, freight avoided.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- The pack is ready on the first working day after quarter close, so the review discusses the quarter instead of waiting to be assembled
- Every score opens to the receipt, complaint or credit note behind it, which turns the meeting from argument into agreement
- The whole base can be scored, not only the sixty prepared by hand, so the tail is managed by exception rather than by memory
- Definitions stop drifting: one written rule per metric, one weighting model with a version history, the same arithmetic in every plant
- Agreed actions carry an owner, a date and a reminder, and the next review opens with their status rather than a reconstruction
The management view
- Supplier performance stops being a quarterly artefact and becomes a figure the procurement director can open on any day
- Sourcing and resourcing decisions rest on a comparable history instead of the last incident anyone remembers
- A customer or certification audit is answered from the record: scores, evidence, exclusions and reasons, actions and closure
- Supplier and plant count can grow without the review programme shrinking to fit the preparation capacity
Board-level KPIs
Security and governance
An auditor should be able to reconstruct every decision.
- The robot's accounts in SAP and the quality system are read-only; nothing here writes back to a transactional system, which keeps the security review short
- Connection secrets sit in Orchestrator's credential store backed by your key vault; scores, evidence and packs they produce stay in your Microsoft 365 tenant, with UiPath Automation Cloud in its EU region
- Row-level security decides who sees which suppliers, and supplier access goes to a named contact rather than a link that can be forwarded
- Weighting changes, metric redefinitions and event exclusions carry an author, a date and a reason, so a published score can be reproduced as it stood and the exclusions appear on the report the supplier sees
Why now
Every new plant and every dual-sourcing decision adds suppliers that ought to be reviewed, while the programme is capped by how many packs ten people can build; assembly alone runs at the modelled €1,750 a month before anything is decided
Price indexation and renewal negotiations go differently when delivery and quality history is on the table rather than asserted from memory, and that history exists only if somebody kept it
The reporting layer no longer needs a project: Power BI Pro comes with Microsoft 365 E5, row-level security and paginated reports are standard, and a pack per supplier is an API call once the report sits in a capacity workspace
Relevant executive roles
The programme covers the supplier base rather than the share that could be prepared, and every score is defensible in front of the supplier
Delivery and quality per supplier and plant become one figure with one definition, available between reviews and not only four times a year
Complaints, 8D cases and their actions sit beside delivery and commercial performance, so escalation rests on the whole picture
Supplier decisions and price movements are evidenced, and ten experienced people stop spending six working days a month on assembly
Common questions and objections
They should, and today they win, because a figure with no records behind it is an opinion. Every score here opens to the receipts, complaints and credit notes that produced it, and a wrong event is excluded with a reason that stays on the report.
That report is one input, not a scorecard, and it answers a slightly different question in each plant depending on which date it measures against. The work here is agreeing definitions once, adding quality, price and dispute data, and keeping the evidence.
No. The inputs are structured records and the weighting is a policy decision procurement should own. A deterministic calculation can be reproduced, explained to a supplier and audited, which is what a sourcing score has to survive.
When this is not the right solution
- Fewer than about twenty suppliers under formal review a year, where a disciplined analyst with a good template costs less than the integration work
- Goods receipts are booked in batches days after arrival, or purchase orders carry no confirmed dates, so on-time delivery cannot be derived from the ERP at all
- Supplier numbers are not mapped across plants and legal entities, so nobody has a group view of who a supplier is; that master-data work comes first
A question for the next management meeting
Before the next review cycle begins, can this company produce one agreed delivery and quality figure per supplier without asking three departments to export anything?
Implementation approach
A scope without ambiguity, before anything is signed.
We deliver
- One quarter of your source data profiled: how supplier numbers map across plants, how receipts are booked, what the quality system exports, where on-time delivery is not computable
- The metric definitions and weighting model, agreed with procurement, quality and the plants, in a list your team owns and versions
- Data-quality gates that hold a score back rather than publish a wrong one, with a report of what was held
- The scorecard store with its evidence lines, the release step in Teams, the Power BI report with row-level security, the pack export and the action loop in Planner
- A pilot on one category, then rollout with hypercare and a runbook for the category managers
We need from you
- Twelve months of purchase-order, goods-receipt and invoice-block history from SAP and one quarter of complaint records
- A named owner for the metric model in procurement and one supplier quality engineer for the definitions
- Read-only service accounts for SAP and the quality system, and a Power BI workspace
- The decision on supplier access: a guest view under row-level security, or the PDF pack in a shared folder
Stages
Discovery
Source-data profile, supplier number mapping, the definitions each plant uses today
Definition
Metric rules, weights, bands and exclusion policy, agreed with their owners
Build
Extract robots, scoring, scorecard store, Power BI report, pack export, Teams and Planner
Parallel quarter
One category scored both ways, differences explained before a supplier sees a score
Rollout
Remaining categories and plants, supplier access, hypercare and tuning
Departmental. Effort is driven by the number of source systems, how consistently supplier numbers and goods receipts are maintained, and how much disagreement about definitions has to be settled first.
The delivery figure in your last supplier pack came from a workbook on somebody's laptop.
Send us one quarter of purchase-order and goods-receipt history for twenty suppliers with your complaint counts. We come back with three scored drafts and the definitional gaps we found.
Score three of your suppliersThe neighbouring process usually has the same problem
Your suppliers calculate your annual price increase. Nobody on your side checks their arithmetic.
View solution Supply chainOrder confirmations and delivery dates chased automaticallyBuyers spend their mornings chasing suppliers for dates, and planners plan on dates that are already wrong.
View solution Operations & qualityNon-conformances and CAPA handled in TeamsContainment, root cause, supplier 8D and effectiveness checks live in one spreadsheet and three mailboxes.
View solutionIndustries we deliver this in most oftenManufacturing & industryRetail & e‑commerce