Home · Solutions · Management & planning
Solution · Management & planningFour DMS, four importer portals, nine sites, one set of numbers every morning
One morning view of every brand, site and workshop
Robots collect the daily and monthly figures from every DMS and importer portal, finance owns the definitions, and the board, each brand and each site read the same Power BI page.
Executive summary
Four DMS, four importer portals and nine sites, and the board sees last month in the third week of this one.
We build the layer between nine sites and one page.
Group, brand and site figures exist before the showrooms open, so Monday's meeting starts with the business, not with whose number is right.
the Power BI semantic model and app; the SharePoint library, definition register and reconciliation list
Business problem
Group management
A dealer group rarely chooses its systems. Each brand contract came with a DMS the importer accepts, a portal the importer requires and a reporting convention of its own; two acquisitions added sites with their own habits. The ledger consolidates money once a month; nothing consolidates cars, hours and parts, so head office does it by hand: two analysts, a workbook with nine tabs and a macro nobody dares to edit.
The people who feel it are not the analysts. The group managing director runs Monday's meeting on a number a day and a half old. Brand directors compare sites that count differently: delivered means invoiced at three sites, handed over at four, registered in the importer's portal at two. The CFO sees the consolidated month three weeks late and learns then that a service department has been below cost for a quarter. The aftersales director, whose workshops are where the group earns its money, gets the least timely figure of all.
At scale it stops working quietly. A fifth brand adds a fifth DMS and a fifth convention, a brand moved to a different remuneration model cannot be compared with the others, and when one analyst leaves, the mapping between nine sets of site codes and the board's view turns out to exist only in a macro and in one memory.
How it works today
- PersonEach site exports the day's sales, workshop and parts reports from its DMS and emails them to head office
- PersonTwo analysts paste them into the group workbook, translating site and brand codes and repairing date formats
- WaitingThe group figure exists late in the morning, once the slowest site has sent; a missing site is carried forward from yesterday
- SystemRegistrations, pipeline and target attainment are read from four importer portals and typed beside the DMS figures
- Risk of errorDelivered means invoiced at three sites, handed over at four and registered at two; the total adds them as if they were the same
- PersonAt month-end 31 reports are rebuilt from fresh exports; a site is re-cut whenever a posting is corrected after the export
- WaitingThe pack reaches the board in the third week of the following month; a question about one workshop waits for the next meeting
Why the current process costs more than it appears
Behind every exception is an hour nobody logged.
- Two analysts at €27 an hour are the visible cost; the invisible one is a site director acting for three weeks on figures that have since moved, and a board deciding on stock from a picture six weeks old.
- Definitions decide money. A brand director praises a site for deliveries that are invoices, while in most bonus schemes the importer settles on registrations; the gap surfaces at settlement.
- Aftersales is where the margin is, and it is the last figure to arrive; a workshop selling fewer hours than it pays for shows up in the annual accounts, not in the month it started.
- Corrections after the export make the pack wrong silently, and the mapping of nine code lists and four portal screens exists in a macro from 2019 and in two memories; a replacement needs a quarter to become useful.
Cost of inaction
Analyst time is the only line here that can be priced without a client ledger, and it is the smaller one. The larger cost is a service department running below cost from February to the annual accounts: a group that earns its margin in aftersales, used cars and importer bonuses is steering its three most important numbers with its slowest reports.
The rows also grow with the group: every acquisition brings a DMS and a convention, every new brand a portal, and each is absorbed by the same two people until one of them leaves.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
An illustrative Polish dealer group: four brands on nine sites with showroom and authorised workshop; about 620 employees; four DMS, four importer portals, one ledger; Microsoft 365 E3 and Power BI Pro.
14 daily reports (four brand flashes, nine workshop flashes, one group summary) and 31 monthly reports; two analysts; about 260 hours a month.
Sites export from their DMS and email; head office pastes, maps and reconciles in a workbook; portals are read by hand; the pack reaches the board in the third week of the following month.
46 minutes per report run on average, a daily figure that waits for the slowest site, and definitions that differ by site.
Robots take the figures from each DMS and portal as offered, map them through finance-owned tables and refresh one Power BI semantic model; board, brand and site open the same page; the daily brief and every difference land in Microsoft Teams.
In the modelled group the daily view exists before the showrooms open, the month is complete two working days after the last DMS closes, and about 260 analyst hours a month move from assembly to explanation. Modelled figures, not a client result.
Proposed solution
We build the layer between nine sites and one page. Unattended UiPath robots run on an Orchestrator schedule and take from each DMS whatever it offers, an export, a file or the screen a controller would read, and from each importer portal the registrations, pipeline and target attainment as the importer counts them. Every extract is dated, filed in a SharePoint library and checked against row counts and day-end totals; a source that fails is held back and reported.
Definitions live in one place, and finance owns it. Mapping tables in Excel Online translate nine sets of codes into the group structure; a definition register says what a delivered unit, a productive hour and a gross per unit mean. The Power BI semantic model applies them once, per brand, site and day: units ordered, delivered and registered, gross per unit, stock age, hours sold and available, parts turnover. Published as an app with audiences and row-level security through Microsoft Entra ID groups, the same page shows the board every brand, a brand director every site and a site director one site.
Two things reach people in Microsoft Teams: the daily brief, posted into the group channel and each site channel before the showrooms open, and every difference between a DMS and an importer portal, listed by vehicle and assigned to the site controller with a mention. Nothing is netted or hidden, and the robot never changes a number: a missing extract shows as a gap, not as yesterday's figure carried forward.
UiPath Orchestrator time triggers, queues and audit; unattended UiPath Robots; UiPath Integration Service connectors for Microsoft OneDrive & SharePoint and Microsoft Teams; Power BI semantic models, apps with audiences, row-level security and scheduled refresh; the Power BI app in Microsoft Teams
The extraction robots per DMS and portal, control totals, mapping tables and definition register, the semantic model and its pages, the reconciliation and its Teams routing, the daily brief, the runbook
Each DMS through its exports, files or screens, with no vendor API assumed until checked; each importer portal through unattended browser automation; the ledger through its export
How the automated process works
- AutomationOrchestrator starts the run at night and after each month-end close; every source, site and day is a queue item with its own retry and log
- SystemRobots take orders, deliveries, stock, hours clocked and sold and parts sales from each DMS, and registrations, pipeline and target attainment from each portal
- AutomationEach extract is dated and checked, codes are mapped and the semantic model refreshes; a failed check holds back that source and posts the reason to finance
- AutomationBefore the showrooms open, the daily brief is posted into the group channel and each site channel in Teams, with a link to the Power BI app
- AutomationThe DMS-to-portal comparison lists every vehicle present in one system and not the other, and assigns each to the site controller with a mention
- PersonThe controller explains or corrects the difference at source (a late registration, a demonstrator, a wrong site code); definition questions go to finance, which decides
- PersonBoard, brand and site directors read the same page; the month is complete two working days after the last DMS closes
Human-in-the-loop model
Automation handles
- Scheduled extraction, mapping and model refresh, with a dated copy of every extract and a log of every run
- Completeness checks and control totals, and holding back a source that fails them
- The comparison of DMS and portal, and the assignment of every difference to a named controller
People decide
- What a delivered unit, a productive hour and a gross per unit mean, and the targets and thresholds the brief highlights; every change is versioned
- Whether a difference is a timing effect, a data error or a real event, and what to correct at source
- Who belongs to which audience
Before and after
Systems and integrations
The stack is deliberately short: one engine, one execution layer, one place where a person decides.
Inputs
- four DMS (sales, stock, workshop, parts)
- four importer portals (registrations, pipeline, stock, targets)
- the ledger export
- targets and mapping tables in Excel Online
Automation layer
- UiPath Orchestrator
- UiPath Robots (unattended)
- UiPath Integration Service
Target systems
- the Power BI semantic model and app
- the SharePoint library, definition register and reconciliation list
Human touchpoints: the daily brief in group and site channels in Microsoft Teams; reconciliation items assigned in Teams; the Power BI app in Teams
Technologies used
scheduled unattended extraction from four DMS and four portals; queues, retries, run history, audit
Afiles extracts, reads mapping tables, writes the reconciliation list, posts the brief
Aone governed model and one page for board, brand and site
Awhere the brief is read, differences are assigned and the app is opened
Athe extract library, the definition register, the reconciliation list and the mapping tables
Athe sources, read as each offers them; no vendor integration assumed
CIllustrative economic model
Numbers you can check against your own data.
Forty-five recurring reports are the whole model: 14 daily reports across 22 working days plus 31 monthly reports make 339 report runs a month, and 46 minutes is the average across a ten-minute daily flash and a site scorecard that takes most of a day. €27 an hour is a fully loaded cost for an analyst post in a Polish dealer group. Nothing was measured at a client; the rows show capacity released, not posts removed.
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
- Group, brand and site figures exist before the showrooms open, so Monday's meeting starts with the business, not with whose number is right
- The month is complete two working days after the last DMS closes, and a service department below cost is visible in the month it happens
- Site director and board read the same page with the same definitions, so a comparison between sites is a comparison, not a dispute about counting
- Every DMS-to-portal difference is listed by vehicle and owned by a controller, so late registrations and demonstrators stop distorting the bonus position
- About 260 analyst hours a month move from pasting to explaining variances; a fifth brand means one more robot, not another analyst
The management view
- Group, brand and site are three audiences of one report; nobody maintains a second version for a director who wants a different cut
- A definition change is a decision with an owner, a date and a version, so any month can be re-read as it was defined
- Site comparisons run on one scale, which is what a remuneration negotiation with an importer or a decision about a workshop needs
Board-level KPIs
Security and governance
Control is not an add-on.
- Robots use a report-only account in each DMS and the group's own login in each portal; secrets come at run time from the Orchestrator credential store backed by your Azure Key Vault, never from a workflow
- Extracts, model and app stay in your Microsoft 365 tenant; the automation runs in the European Union region of UiPath Automation Cloud with its run history and audit log
- Access follows Microsoft Entra ID security groups as app audiences and row-level security, which applies to viewers, so board, brand and site are viewer audiences
- Personal data stays out of the model: the reconciliation carries vehicle identifiers, the brief carries counts and values; a definition or mapping change is approved by its owner in finance and versioned
Why now
Several manufacturers have introduced or announced agency distribution for some brands or for electric models, and dealers' associations have raised remuneration and investment concerns; a group that cannot compare a brand's income per site on one scale negotiates without a number of its own
Poland registered 597,400 new passenger cars in 2025, 8.3% more than in 2024, more than half of them not purely combustion engined (PZPM and KPMG quarterly report, 3 February 2026); brand mix, stock age and workshop content move faster than a third-week pack can show
Robots reading an export or a portal screen, Orchestrator time triggers and semantic models with row-level security in Teams are configuration rather than development, so a DMS without an API no longer blocks a group model; the assembly meanwhile costs about €7,020 a month
Relevant executive roles
Monday's meeting starts from one page every director has already seen, and a site below cost is a March conversation, not a year-end discovery
The consolidated month exists on the second working day, with definitions finance owns and a trail from every figure to its extract
Hours sold, hours available and parts turnover per site arrive the next morning on one scale for nine workshops
Deliveries, registrations and the gap between them are visible per site every morning, before the importer's settlement
Common questions and objections
Then the robot uses the screen a controller uses today, unattended and on a schedule, and the extract is dated and checked like any other. That is the normal case in dealer groups, which is why the work starts with an inventory of what each system offers.
The model does not replace the importer's count; it shows it beside the group's own, with the difference listed by vehicle and the site controller owning the explanation.
They argue today, in the meeting, with different spreadsheets. The register moves the argument to one place, gives it an owner in finance and a version, and settles it once.
When this is not the right solution
- A single-brand group on one DMS whose reporting module already produces the figures; a scheduled export and one report are cheaper than a model
- A group about to consolidate onto one DMS within months; we agree the definitions now and build the extraction after the cutover
- A management team not prepared to let finance own the definitions; automation delivers a contested number faster
A question for the next management meeting
Had the workshop at our smallest site been running below cost since February, on what date this year would this board have found out, and from which report?
Implementation approach
A scope without ambiguity, before anything is signed.
We deliver
- An inventory of every report: source, recipient, the form each system offers and the time it takes today
- A definition workshop with finance and the directors: one group definition per KPI with an owner
- Extraction robots per DMS and portal, with schedules, control totals and the dated library
- The semantic model, row-level security, the three pages, the daily brief and the reconciliation in Teams
- One month in parallel with the existing reports, then the switch with hypercare and a runbook
We need from you
- Last month's daily flashes and month-end pack, as the sites send them
- Report accounts in each DMS, portal accounts and a Power BI workspace in your tenant
- A named owner in finance for the definition register and one controller per site
Stages
Inventory and definitions
Every report, its source and the form each system offers; one group definition per KPI with an owner
Extraction
Robots per DMS and portal, schedules, control totals and the library
Model and pages
The semantic model, row-level security, the three pages and the daily brief
Parallel month and handover
One month beside the existing reports, then publication, hypercare and the runbook
Departmental. Effort depends on how many DMS and portals there are, whether each offers an export, a file or only a screen, and how far the sites' definitions differ.
March reaches your board in the third week of April, one DMS at a time.
Give us one week of daily flashes and one month-end pack, as the sites send them. We return a written read-out: which figures a robot can take from each system as they are, which need a group definition first, and a first estimate of the assembly hours.
Line up your DMS exports against one pageThe neighbouring process usually has the same problem
The board pack should not depend on which analyst merged which spreadsheet on which day.
View solution Finance & accountingManufacturer bonuses claimed and matched to the payoutImporter bonuses are a large part of your margin, and they are tracked in a workbook one person understands.
View solution Sales & marketingEvery lead with the right adviser within the hourNine channels, four brands, one receptionist deciding who calls the customer back. The Friday lead waits until Monday.
View solutionIndustries we deliver this in most oftenAutomotive retail