Category: ERPNext

A services firm with 60 retainer clients was billing from a spreadsheet maintained by one person. When she took two weeks of leave, invoices went out eleven days late and three clients called to ask if the company was still operating.

That kind of fragility is what pushes a business toward structured systems. Here’s the sequence that worked during this ERPNext Consulting engagement, in the order we actually did it.

Steps that took this firm from spreadsheet to predictable billing

1. List every contract and its billing rhythm — monthly, quarterly, milestone, or hourly with a cap. Sixty clients, four patterns. Anything that didn’t fit went into a “needs a decision” pile.
2. Agree what triggers an invoice. For fixed retainers it’s the calendar date. For hourly work it’s the approved timesheet. Writing this down killed most of the disputes.
3. Set up the service items once, correctly — description, rate, tax treatment — so nobody retypes a contract name with a typo.
4. Automate the draft, not the send. Invoices generate as drafts two days before the due date. A human reviews the list and releases it.
5. Build one aging view that the account manager sees, not just finance. Chasing a late client is easier for the person who has the relationship.

Why we kept a human on the send button

Full automation sounds appealing until a client’s scope changes mid-month or a project pauses for two weeks. Two days of draft review catches roughly six exceptions a month at this firm — the kind that, if billed wrong, take three emails and a credit note to fix.

The measurable win came on the collection side. Average days to payment dropped from 47 to 29 over one quarter. That’s not a software feature. It’s the result of the person who owns the relationship seeing the overdue balance on the same screen they use every day.

What to check before you start

Count how many of your recurring invoices are edited by hand each month. If the answer is more than one in five, your contracts aren’t described clearly enough yet, and automating will just make wrong invoices go out faster.

An ERPNext Consulting engagement here isn’t about billing software. It’s about deciding what a contract actually promises, one client at a time, until the pattern is obvious.

A retail owner with nine stores told me his plan: get every location onto identical processes, one price list, one reorder rule, one way of doing everything. He was proud of it. Two of his best store managers resigned within four months.

This is the most common misconception I run into during ERPNext Consulting reviews — that consistency and central control mean the same thing. They don’t, and confusing them costs good people.

What should be identical, and what shouldn’t

Some things genuinely need to be the same everywhere. Financial reporting, tax handling, customer returns policy, and the master list of what you sell. If a store invents its own item codes, you lose the ability to see anything across the business.

Other things need room to move. Reorder points, display layout, opening hours, and who staffs Saturday all depend on local traffic and local customers. Forcing a store in a suburban strip mall to follow rules designed for a downtown location is how you end up with dead stock in one place and empty shelves in another.

The split that worked

We reorganised control into two buckets and told everyone which was which.

• Central, locked: item master, price list floor, credit limits, supplier terms, chart of accounts, tax rules.
• Local, editable with reason codes: reorder points, stock transfer requests, staffing rota, local promotions that stay within the price floor.
• Reported monthly to the owner: any local change to reorder points, with the reason, so patterns show up before they become habits.

The key detail is that local changes were allowed, but never silent. Every edit recorded who made it and why. That one feature turned nine arguments into nine short conversations.

The mistake that usually comes next

Owners who go through this often swing the other way and let every store do its own thing. That fails too, and it fails slower, which is worse. You find out eighteen months later when nobody can tell you total stock by item.

Start with a single page: list your twenty most important processes down the left, mark each one central or local, and get your store managers to argue with it for an hour. What survives that hour is your operating model. Everything else is a preference pretending to be a policy.

“How long from ‘can you make this’ to a price in the customer’s inbox?” I asked the sales manager of a 55-person metal fabricator. He laughed. “Four days. If it’s simple.”

His counterpart in the estimating office didn’t laugh. She said the real number was closer to six, because two of the four days were spent walking the shop floor looking for a similar job from last year.

The real cost nobody had written down

This company quoted about 210 jobs a month. Of those, roughly 60 were won. The other 150 took up the same estimating time. At an average of five hours per quote, that’s most of two full-time salaries producing nothing but politely-worded emails.

The bigger leak was elsewhere. Customers who needed a fast answer stopped asking, because they’d learned the answer would arrive after their own deadline.

What the estimator actually needed

She didn’t need a faster calculator. She needed to see, in one screen, every part the shop had cut in the past two years with the same material, thickness, and bend count — and what it took to make each one.

• Past jobs searchable by material, thickness, and number of operations, not by customer name she’d have to remember.
• Actual hours from the floor attached to each finished job, so the next quote used real numbers instead of the estimate from the last estimate.
• A short list of standard operations with rates that only two people could change, so quoting couldn’t drift on its own.
• Quote status visible to sales — sent, viewed, chased, won, lost — with a reason recorded on every loss.

None of that is exotic. What made it work was the discipline of recording actual hours, which took six weeks of nagging before floor supervisors trusted that it wasn’t a tool for blaming them.

Six weeks later, the same question

I asked again: how long from enquiry to price in the inbox? The sales manager said 40 minutes for repeat work, and same-day for most of the rest. Quote volume went up to 280 a month without hiring.

Win rate moved from 28% to 34%, which surprised everyone until we looked at the lost-quote reasons. Half of them had been price. But a solid share were “we didn’t answer in time.” Those simply stopped happening.

If you run a job shop, pull your last 20 lost quotes and write down the reason next to each. If more than two say “too slow” or “no reply,” the fix is worth more than any price adjustment you’re considering. That’s the kind of finding an ERPNext Consulting review surfaces in the first week, before anyone touches settings.

I’ve sat in production meetings at twelve manufacturers over the past two years, from 60-person machine shops to a 400-person food plant. Eleven of them could tell me exactly what came off the floor last month. Not one could tell me what would come off the floor next Thursday if a big order moved.

That blind spot is what most ERPNext Consulting work in manufacturing really attacks, whatever the project documents call it. Here’s what those twelve had in common, and the three numbers that told us whether the fix would stick.

The three numbers that predicted success

Before touching anything, we measured three things: how many open work orders were already past their promised date, how often the schedule changed within 48 hours of a job starting, and how many jobs started without all materials on hand.

At the food plant, the first number was 31%, the schedule changed nine times a day, and four in ten jobs started short of materials. At the machine shop that later ran the smoothest, those same numbers were 8%, two changes a day, and one job in twenty. The difference wasn’t discipline. It was whether the schedule lived in one place.

Why the plan breaks at around 80 open orders a month

Below that, one experienced planner holds the whole schedule in their head and reshuffles it when a machine goes down. Above it, the plan lives in three places at once: a whiteboard, a spreadsheet, and the planner’s memory. Each of them is right about different things.

The symptom managers notice first isn’t late delivery. It’s that quoting gets slow and cautious, because nobody wants to promise a date they can’t see clearly.

What actually changed in four weeks

1. Every job got one start date and one finish date, editable by both the planner and the floor supervisor — with one named owner per change.
2. Material availability was checked at the moment a job was released, not the morning it was due to start.
3. Machine hours per week were logged for four straight weeks, so capacity stopped being an opinion.
4. Friday, 30 minutes: the coming week’s schedule reviewed with floor supervisors in the room, then frozen unless the plant manager initialed a change.

By week ten, the food plant’s late work orders sat at 11% and schedule changes were down to three a day. The planner, who had been working six days a week, went back to five. That’s usually the moment a team admits the change is real.

Write those three numbers down for your own plant this Friday. Whatever they come out to, they’re the baseline you’ll be judged against — and the only honest way to separate an ERPNext Consulting project from a software installation.

In March, the warehouse manager of a 40-person building-supplies distributor slid a stock report across the table and said, “This claims we’re holding 1,200 units of a fitting we haven’t sold since 2022.” Two people walked out to check. The shelf held 90.

That gap between what the system said and what was physically there is where the money was hiding. Over the next five months, an ERPNext Consulting engagement rebuilt how this company counted, reordered, and priced its stock. The final write-off came to $380,000.

Why the numbers drifted 34% away from the shelves

Nobody was stealing. Three ordinary habits were doing the damage. Receiving clerks booked goods against the purchase order the moment the truck arrived, before anyone counted the pallets. And the annual stocktake happened on a Sunday, run by people who don’t normally touch the warehouse.

Individually each habit looks small. Together they made the stock figure useless for decisions, so managers stopped trusting it and ordered by gut. Gut ordering at this company meant carrying roughly nine weeks of slow-moving items instead of four.

The first 30 days: three unglamorous rules

We didn’t start with settings. We started with what people do at 7:30 a.m. when a truck backs in.

• Goods landed in a waiting-to-be-checked location, not sellable stock, until a blind count was entered the same day.
• Anything returned by a customer got tagged and moved to a quarantine shelf within the hour, so it could not be promised to the next buyer.
• The stocktake moved to a normal working Thursday, split into 14 zones, with each zone counted by someone who doesn’t pick from it.

Two of those rules needed a small change in how the system handled locations — the kind of tweak an ERPNext Consulting team makes in an afternoon. The third needed a rota. The rota was harder.

What the owner saw by month five

The number that ended the argument wasn’t the write-off. It was reorder spending, down 22% while order fill rate climbed from 84% to 96%. That combination is impossible if the stock figure is fiction.

1. Weeks of stock by category, reviewed every Monday for 20 minutes instead of monthly.
2. A dead-stock list anyone can sort by last sale date, with anything past 180 days flagged for a pricing decision.
3. One number per item: what’s physically on the shelf, never a report number and a shelf count added together.

If your warehouse and your reports disagree, pick ten items tomorrow, walk out, and count them. The size of that gap tells you whether you have a stock problem or a trust problem — and the second one costs more.

Let’s start at the end. A 140-employee food processor now closes its monthly books in three working days, with the same five-person finance team, on the same systems it had five years ago.

Two years earlier, the close took 11 days, depended on a 40-tab spreadsheet that one person understood, and produced numbers that were still being corrected three weeks later. Here is the story backwards, starting with what day three looks like now, after an ERPNext Consulting project that changed almost nothing about the tools and a lot about the habits.

Day 3: the close is mostly a review

By the morning of day three, the bank accounts are matched, the inventory is valued, and the only thing left is a checklist of six review items. The controller signs off before lunch.

There are no all-nighters and no spreadsheet sent around for “final checks.” That is the visible result. The cause sits much earlier in the month.

Day 1, every day: the 15 minutes that ended the scramble

The team added a short routine at the end of every working day, and this is the piece most companies skip because it feels too small to matter.

1. Post everything received that day, including deliveries that arrived without paperwork. Nothing waits for the invoice.
2. Match yesterday’s bank transactions. One day of activity takes about four minutes; a month takes half a day.
3. Look at the exception report — short shipments, price mismatches, unapplied credits — and clear or assign each item before leaving.
4. Count the three highest-value ingredients in the production area. Just three, every day, rotating through the list.

Fifteen minutes a day replaced 60 hours of month-end panic. The finance manager told me she stopped dreading the first of the month for the first time in nine years. On its own, though, the daily routine was not enough. Two structural things were still creating the mess.

The two things that were generating the work

First, inventory was valued once a month from a 300-row spreadsheet, which meant every month ended with a three-day reconciliation. Second, revenue was recorded when cash arrived rather than when goods shipped, so sales figures were argued about every single period.

Changing those two rules cut the close from 11 days to 6. The daily routine cut it from 6 to 3. This is the part of the ERPNext Consulting work clients resist most, because it means giving up a spreadsheet people have learned to trust.

Three months after the switch, correction entries after close fell from about 200 lines a month to fewer than 20. The leadership team now sees last month’s numbers on the fourth working day instead of the fourteenth. Eight extra days of knowing, every month, is worth more than any report the finance team could have built under the old cycle.

Tomorrow, take the one task that always piles up at month end — usually bank matching — and do it daily for two weeks. Time yourself. By Friday you will have your own business case, written in minutes saved.

Owner: “Every job on the report shows a 40 percent margin. Our bank account says we are running at 6 percent. Somebody is lying to me.”

Me: “Show me yesterday’s biggest job, and let’s walk it from the first phone call to the final invoice.”

That is how the ERPNext Consulting review started at a 120-technician heating and electrical contractor. It took three hours to find the answer, and it was not fraud.

The job that looked like a winner

A commercial rooftop unit replacement, quoted at $18,400. The report showed $7,100 of gross profit, a 39 percent margin. Solid work by any standard.

Then we rebuilt the job from the actual records. Two technicians for a day and a half, but the report only counted 12 hours, because the clock started when they arrived and stopped when they left, and the yard is 50 minutes from the site. Parts came off the van, which nobody tracked. The crane subcontractor’s $1,900 invoice went to the company books, not to the job.

Eleven days later, a technician went back to replace a faulty sensor at no charge. That visit existed nowhere. Real margin on that job: 4 percent, not 39.

Four gaps that made every job look better than it was

• Drive time was invisible. Across 120 technicians, that was roughly 700 unpaid hours a month appearing as profit.
• Van stock was a black hole. Parts were pulled and consumed, but only the ones bought for a specific job got charged to it.
• Warranty callbacks were not linked to the original job, so rework looked like a fresh service call with its own revenue line.
• Subcontractor and permit invoices were recorded company-wide, never attached to the job that needed them.

The conversation that changed the pricing

Once jobs were costed honestly — the part of the ERPNext Consulting project that took the longest and mattered the most — a pattern appeared within three weeks. About 22 percent of jobs lost money, and nearly all of them were the same type: after-hours emergency calls on equipment older than 15 years.

The company had been pricing after-hours work at 1.5 times the standard rate. Real cost, including drive time and the callback rate on old equipment, was closer to 2.4 times. They moved to 2.2 times for anything over 12 years old and stopped advertising same-night service for that equipment class.

Company-wide margin went from 7 percent to 15 percent over nine months, on flat revenue. Nothing was sold differently, and nobody was let go. The only thing that changed was what the company knew about each job.

This week, take your ten largest jobs from last month and add drive time, van parts, and any repeat visit. Recalculate the margin yourself and compare it to the report. You may not like the answer, but at least it will be one you can act on.

A 20-person cross-border e-commerce team sold into the United States, Germany, and Australia from two legal entities and three warehouses. The founder kept his own spreadsheet of the “real” numbers because he did not believe the reports his own team produced.

When the person running the business keeps a private version of the books, you have a problem worth fixing. The symptom looked like an accounting issue. The cause was an operating one. The ERPNext Consulting work took about ten weeks, end to end.

Before: four versions of the same month

Closing the month took 14 days. During those 14 days, nobody could say whether the business had made money.

• Currency conversion happened in a spreadsheet, using a rate someone looked up on whatever day they got around to it.
• Stock moving from the UK warehouse to the US warehouse was recorded as a sale in one entity and a purchase in the other, inflating revenue by about $90,000 a month.
• Inventory was valued in three local currencies and added together, which is like adding 40 kilometers to 40 miles.
• Marketing spend, freight, and duties sat in one pooled account, so nobody could tell which market was profitable.

After: one number, three markets, five-day close

The changes were about setting rules, not about features. Each rule took an afternoon to agree and a week to make stick.

1. One exchange rate source, one rate per month, locked on the last working day. Everyone uses the same rate, including the freight cost on Australian orders.
2. Transfers between the two entities are recorded as transfers, not as sales. Revenue now means money from customers, full stop.
3. All inventory is valued in one base currency. Local currency figures are shown for reference, never added together.
4. Every cost gets attached to the market that caused it — freight, duties, marketplace fees, returns. A product’s margin now includes what it actually costs to get it to the customer.

What the honest numbers revealed

The Australian market, which everyone assumed was the growth engine, had been losing money for four quarters once freight, import duties, and returns were counted. Margins there were negative 6 percent, not the 22 percent the old reports showed.

They raised Australian prices by 8 percent, dropped two heavy low-margin products from that market, and stopped paying for express shipping on returns. Within two quarters, Australia contributed 11 percent of gross profit instead of dragging it down.

The founder deleted his private spreadsheet in week six. That was the real signal that the numbers had become trustworthy, and it is the point of an ERPNext Consulting engagement done properly.

Pick the market you are least sure about, take one product, and add freight, duties, and returns to its margin by hand. If the number surprises you, you have found exactly where to look next.

Most purchasing problems don’t look like problems. They look like speed. Someone needs a part, they get it, the job moves on, and the invoice gets approved because the job was urgent and the relationship is good.

That is how a 200-person building services firm was running $60 million of annual spend across 2,400 suppliers and roughly 9,000 purchase transactions a year. The ERPNext Consulting review started with a simple question: how many different prices did we pay for the same item last quarter?

The answer was 11. For a standard pair of safety gloves.

The five leaks nobody writes down

• Approval by email thread. Nine people could approve, nobody knew their limit, and the average approval took 3.5 days — which is exactly how an “emergency” gets created.
• Same supplier, different price per site. Each branch negotiated its own deal, so identical cable trays ranged from $38 to $61.
• Storeroom blindness. 22 percent of orders were for items already sitting in a bin 30 feet away, because nobody could see the bin without walking to it.
• Emergency buying as a habit. 18 percent of spend went out on same-day terms at an average 24 percent premium.
• Expired contract pricing. Discounts negotiated in January were still being invoiced at list price in September, and nobody was checking.

What we changed in the first quarter

None of this needed more people. It needed three rules and one report that anyone could read.

1. Three approval tiers with hard dollar limits — under $1,000 by the site lead, $1,000 to $15,000 by the department head, above that by finance. Every limit visible in the system, no exceptions.
2. Every approval request gets a 48-hour clock, then escalates automatically. Approvals averaged 3.5 days before; they now average 9 hours.
3. One supplier record per supplier, with the negotiated price list attached. If the price on an order does not match the list, it cannot be sent without a written reason.
4. A monthly review of the top 20 suppliers by spend, checking three invoices each against the contract price. Twenty minutes of work, and it caught $180,000 of overbilling in six months.

The number that changed the conversation

Spend outside agreed contracts fell from 34 percent to 9 percent in five months. Overall purchase prices came down about 6 percent, which on $60 million is roughly $3.6 million a year — more than the company’s entire admin budget.

The finance director summed it up in one line: most of that money was not lost to bad suppliers, it was lost to good intentions and no visibility. The ERPNext Consulting work that mattered most was the boring part — writing down who could approve what.

Before your next leadership meeting, pull the last 90 days of purchases over $1,000 and check three of them against what you thought you were paying. That is your baseline, and it usually takes under an hour to gather.

A 45-person cabinet maker came to us with a clean theory. Orders were 30 percent late, so the schedule must be broken. They wanted a better plan, a better board, a better way to slot jobs into the week.

Before touching the schedule, the ERPNext Consulting team spent two days on the shop floor with a clipboard, writing down what each job actually consumed. The schedule was fine. What we found instead was that no two people built the same cabinet the same way.

Same product, different parts, every single time

This company made 14 cabinet families with maybe 200 variations. On paper, each had a defined parts-and-steps list. In practice, those lists had been written years ago and never updated after the designs changed.

So the crews improvised. One team cut shelves at 24 inches, another at 23 and three-quarters, because the hardware on hand that month was a different brand. Purchasing bought to whatever the last shortage reminded them of, not to what the next two weeks of orders needed.

• Scrap and offcuts ran at 11 percent of material cost, against a 4 percent target.
• Production staff spent an estimated 6 hours a week hunting for parts that should have been kitted on a cart.
• Three customer disputes a month came from dimensions that did not match the original quote drawing.

Rebuilding a build list people actually trust

We did not rewrite 200 documents in a back office. We rebuilt 14, one per family, with the people who make them.

1. Pick one cabinet family. Have the lead builder assemble one unit and record every part and every step as they go, including the small stuff nobody writes down.
2. Compare that record against the current drawing, the quote, and the last three jobs of that family. Every difference gets a decision, not a compromise.
3. Attach a photo of the finished assembly and a photo of the kitted cart for that family. New hires follow pictures faster than documents.
4. Lock the parts list so purchasing can plan from real demand, and update it the day a design changes — not at the end of the year.

The result that surprised everyone

Six months in, material scrap fell from 11 percent to 5 percent. On-time delivery went from 70 percent to 92 percent, and overtime dropped by nearly a third, with no change at all to how jobs were scheduled.

The general manager’s comment stuck with me after the ERPNext Consulting engagement ended: the schedule was only ever as good as the recipe behind it. Seven months earlier, his team was ready to buy a new planning tool.

Start with the product family that causes the most arguments on the floor. Build one unit with your best builder and write down everything. That single afternoon will tell you more than a month of meetings.