Case study · ERP dashboard and agent

ERP Dashboard + AI Purchasing Agent:One Live View of the Business

See the whole business

The month at a glance.

1 / 6

We built an ERP dashboard for an Australian mobile phone parts wholesaler: one live view of sales, customers, products and the real P&L, and a purchasing agent that suggests what to order.

70,969monthly sales rows matched to the ERP
20ERP feeds, refreshed hourly

See the whole business

The month at a glance.

1 / 6

01The problem

The ERP had the data. Not the answers.

Every answer started as an export.
Buying meant five reports and two layers of Excel, every week.
Nothing checked the sheets against the ERP, so a wrong total looked as good as a right one.

rows in one sales report: 38,434

02Trusted data

We connected every ERP feed, then checked every row.

We brought 20 ERP feeds into one model that refreshes every hour. Then we compared every row with the ERP's own reports.

customer, SKU and month rows. None missing, none extra.: 70,969
ledger values, matched to the cent.: 1,614

The data sources panel. 20 sources, from articlesPricing, customers, profitAndLoss and purchaseOrders to salesReports, salesValue, stock, debtorInvoices and skuAttributes, each with the time it last synced.

03What we built

We built two tools on one set of numbers.

The dashboard and the buying plan read the same model.

The ERP dashboard: Company, customers, products and the real P&L.
The purchasing agent: Suggested quantities the team reviews, then exports.

04What the team sees

Now the team can answer nine questions in one place.

1/9Company overview

Are we on target this month?

This month against its target, and where it lands at today's pace.

The current month attainment card from the company overview. The ring reads 15% of KPI with 23% of the month elapsed and a projection of 65%. Actual $89.1K, KPI target $597K, projected $386K. — Open full size

2/9Customer detail

Which customers are falling behind?

Every customer against its own target, with its trend and its pace. One filter shows only the ones at risk.

The customer detail table for the last complete month, 1,032 customers that filter to 122 active, 42 over KPI, 4 on track and 54 at risk. Each row shows the customer's KPI, sales for the period, attainment against pace, the previous period and the change, a trend line and a pace status such as Ahead of Pace or Well Behind. — Open full size

3/9Rep performance

How is each rep tracking?

Each rep against target, with how many of their customers are over, on track or at risk.

Rep performance cards for the last complete month. Each card shows the rep's sales against KPI, customers and active customers, units, gross profit and margin, KPI attainment such as 105%, 79% or 159%, and how many of their customers are over, on track, at risk or without a sale. — Open full size

4/9Customer health

Which accounts need a call?

Accounts likely to leave, ranked by value at risk. Each one comes with a next step.

The action queue of accounts with a churn score of 55 or more, ranked by value at risk. Each card shows the spend tier, the lifecycle stage, the trend, the last order and its usual cadence, the 12 month revenue and a recommended Rescue Plan. — Open full size

5/9Products

What actually sells?

Units and revenue by category and series, month by month.

The products view for the last calendar year. Tiles read 120,480 units, 273 customers buying, 1,264 unique SKUs, Displays as the top category at 69% of units and S Series as the top series. Below, monthly units against revenue and the category split. — Open full size

6/9Financial highlights

What did the period actually make?

Revenue, cost of sales, margins and expenses for any period.

The financial highlights for the last six months to date. Revenue $3.6M, cost of sales $2.9M, gross profit $716K at a 20.0% margin, net profit of minus $201K, operating expenses $869K, 47,670 units sold and 205 active customers, with $638K of receivables owing and the receivables aging bar. — Open full size

7/9P&L statement

Which ledger lines made that number?

The P&L, line by line, straight from the ledger.

The P&L statement for the last six months with Revenue expanded into its ledger accounts, such as 1100 Default Sales Account, 1104 Returns Non Warranty and 1150 Repair Revenue, down to 1250 Freight Recovered, each with its monthly value and month on month growth. — Open full size

8/9Ratio analysis

Are the margins holding?

Gross, operating and net margin, the cost ratios and growth, month by month, from the posted P&L. The month in progress shows what has posted so far.

Ratio analysis by month from May to October, with October showing what has posted so far. Rows for gross profit margin, operating margin and net profit margin, the cost of sales and operating expense ratios, and revenue and gross profit growth. September reads a 23.1% gross profit margin, a 76.9% cost of sales ratio and a 23.3% operating expense ratio. — Open full size

9/9Debtors

Who owes us, and how late is it?

Every open invoice, aged from current to over 90 days. What is overdue and the oldest debt sit up front. Related-party balances are kept out of the total.

Receivables aging at the ledger date. Net owing $638K after $27.6K of credits, $237K overdue across 29 accounts, $429K current and the oldest debt 318 days overdue. The aging bar runs from current, $429K, through 1 to 30 days, $116K, 31 to 60 days, $95.7K, and 61 to 90 days, $6.8K, to over 90 days, $17.9K. Related-party balances, $1.5M across 3 accounts, sit apart and are excluded from receivables, above chips that filter the debtors by aging bucket. — Open full size

05The purchasing agent

The agent turns stock and sales into a buy list.

It reads stock, open orders and sales for every SKU, applies the team's buying rules and suggests what to order.

A quantity for every SKU, or the reason there isn't one.
Anything it can't work out goes to a person.
The reviewed list exports as a file.

The AI Purchasing Agent. Six status cards read urgent units, sold out, optional top-up, suggested units, needs review and not for reorder, above a table of SKUs with stock, inbound, urgent, top-up and suggested quantities and a review quantity for the team to fill in. — Open full size

06The buying rules

When a rule changes, so does the quantity.

The buying team writes the rules with spreadsheet functions like IF, MIN, MAX and ROUNDUP. They preview what changes, then make it live.

Safety stock, the standard rule

CEIL(rate × leadDays × variability)

Daily sales rate: 0.082
Lead time, days: 14
Variability (XYZ): 1.5

0.082 × 14 × 1.5 = 1.72

2 units

safety stock for this product

Purchasing Logic and Policy on the Recommendation logic step. The Safety stock formula reads CEIL of rate times leadDays times variability, validated as 2 units for one product, above colour coded blocks for data source fields, parameters, and functions and operators, with a worked example of the product's inputs on the right.

07What changed

What the team can do now.

01
Read the whole business on one screen
Sales, targets, customers, products and the P&L come from one model. Nobody rebuilds a report to ask a new question.
02
Trust the number in the room
Every row ties back to the ERP. A meeting starts with the answer, not with whose spreadsheet is right.
03
See risk before it costs a month
Accounts slipping, lapsing or behind pace are ranked by value at risk, each with a next step.
04
Read the month from the real ledger
Margins and cost ratios come from posted accounts. The month in progress shows what has posted so far, or a projection.

Related resources

Case Study

Reporting & Dashboards

CEO Profit Dashboard

The founder used to build four reports by hand every morning. Now one dashboard answers the only questions that matter. Are we ahead of last month, and what should we change? It reads sixty-nine months of the client's own data, reconciled to the cent, and it replaced a rented tool whose numbers never added up.