On continuous improvement in ERP
An ERP is not something you install once and leave alone.
It is where a company stores what it has learned. This is about improving it, and about measuring what each improvement is worth — why this exists.
Writing
36 pieces-
Make-up orders from a supermarket chain: one PO in 26, and only the receipt tells
A store raises a PO and receives it the same day to pay back goods already delivered without one. The PO header still says "deliver tomorrow", so a make-up order is easily delivered twice and the goods behind it never invoiced. Measured on 1,258 POs, with a matching threshold chosen from the data.
Read it → -
One screen for a delivery crew's day: 43% of notes posted before the truck delivers, and returns nobody can record
The warehouse posts the delivery note as the truck leaves; the driver's screen read it as "done" and locked the returns field. Over 14 days, early-posted notes had a signed slip 1% of the time and a returned quantity 0.4%, against 89% and 15.5%. Reopen the note for the driver, add a vehicle & fuel tab that reads photos with AI — what that saves a month.
Read it → -
Exceptions that arrive with a reason: 82 slips to check, but only 68 need someone to think
When invoices to supermarket chains raise themselves, the remaining slips reach the accountant with a reason in words. Read by what has to be done, the reasons fall into three kinds: judgement, wait, one-time fix. Measured on 291 slips, and the price of the minutes spent searching.
Read it → -
Trying Zalo-in-ERP with no sign-up: in the real inbox after 2 seconds — and nobody stayed 30
A trial page puts the visitor straight into the real ERP inbox, a sample message just arriving, no form, no password. 43 visits from ads: 35 reached the inbox within 2 seconds, 5 stayed 10 seconds, none stayed 30. What a demo costs, and how much of it a trial can replace.
Read it → -
Looking up TT 38/2026 norms with one question: one code gives every material, labour and machine row — and why to ask ten items at a time
The assistant finds the code in a 10,208-code dictionary, then reads the annex page itself to pick the right column. Two neighbouring codes in one table differ by a median 14.7% in labour; one AI question is worth 24 minutes of work. How many minutes must a lookup save to pay, and what does a wrong column cost.
Read it → -
New orders carry the customer's own price list: 1,061 of 1,140 orders once carried the wrong list, after the fix 0
The order screen pre-filled the default price list, so 93% of one store group's orders carried another customer's list; the quick sales screen also took prices from other lists and wrote them over the customer's list. Now orders always carry the customer's list, and a missing price stays empty. How far each item was off, and what it is worth each month.
Read it → -
A signed invoice posts itself to the customer's Zalo group: the first PDF arrived within minutes, the ledger corrected before sending
Once an e-invoice is signed, its PDF goes by itself to the Zalo group where the customer already talks, with the amount; before sending, the invoice in the ledger is corrected to match the signed copy. The first one has gone, and only 3 of 20 stores have a group so far. What is this worth, and what assumptions does the figure rest on?
Read it → -
The store's photo in the group chat fills in the delivery note: 2 photos matched, but only 3 of 46 stores linked
The store receives the goods after the driver has left, checks them and sends a photo of the corrected note, "10 − 1 = 9", to the Zalo group; the machine finds the delivery note, records delivered, support and returns, and queues it for invoicing. Each photo is worth a few thousand dong — the real money depends on how many stores have their group linked.
Read it → -
Try it without signing up: how soon an estimator sees a first result on their own file, against waiting for a demo
Software for estimators is usually sold through a demo: leave your number, book a slot, watch a sample project, and only then try your own file. When opening the page is enough to start — one project, upload an estimate and drawings, sign up only to keep the project or to use the AI — how many hours and how many days does finding out "does this fit us?" cost, and when does it save nothing at all?
Read it → -
One photo of the day's delivery list: 462 printed cells, 13 handwritten numbers, the machine sure of 1
A driver photographs one printed list for the whole day; the AI looks for hand marks on the grid and the machine compares handwritten numbers with delivered quantities. Two live days: 10 lists, 13 handwritten numbers, the machine certain of 1 — the value lies in putting the right cells in front of a person, and this is how to price it.
Read it → -
A G8 file from the work items you already have: how many hours of re-typing it saves, and how many copying typos stop reaching the estimate
Quantities worked out in your own sheet, then every code and every explanation line typed again into the G8 estimating program: a few hours a month, and a few copying typos along with them. When an assistant writes the work items as one file in the import layout of G8 2020's «Công trình» sheet — explanation lines as formulas, group and section total rows — how many hours come back, how many typos go away, and how much AI cost comes off?
Read it → -
Never invoice twice: ask the accounting ledger before creating — 199 of 209 invoices were already there
When invoicing moves from the accounting package into the ERP, there is a stretch when both create invoices. If the ERP asks the accounting ledger before it creates one, how many duplicates does a cutover avoid — and what is each duplicate you never clean up worth?
Read it → -
Sharing an estimate by one link: what the mailed-Excel loop costs, and "which version is the latest?"
An estimate passes through 2–4 people as Excel files sent by mail, so someone always asks "which version is the latest?". When the whole set — drawings, quantity table, where every number comes from — sits behind one shared project, with people invited by email or sent a view-only link that needs no account, how many hours a month does the review loop get back?
Read it → -
Invoicing at the original PO price: the delivery note says 21,000, the chain's PO says 30,000
A retail chain pays the price on its own PO, not the price on the supplier's delivery note. When a machine reads the PO price, invoices at it when the gap is plausible and corrects the price list for next time, what is that worth a month — and which lines still need a person?
Read it → -
Searching a scanned drawing set: 171 pages recognised in 338 seconds, and the paging hours that come back
In a drawing set, a scanned page yields nothing to Ctrl+F; looking up a mark means paging. Once every scanned page is recognised into text — text pages and scans counted separately, an interrupted run resuming where it stopped — how many hours a month does an estimator get back, and does the largest set even finish?
Read it → -
Ask the project, get the sheet and the row: how many minutes must a question save to be worth paying for?
Each project gets an assistant that works only on that project's drawings and estimate, and every answer links to the exact drawing page or estimate row. Nine questions measured on the live system cost 5,107–78,852 dong each. This piece works out the break-even point, and why quick look-ups are still better done by hand.
Read it → -
Invoicing what the supermarket chain actually received: 72% automatic, and why the tallest bar waits for a person
Invoicing from the delivery note would be wrong on two invoices out of three, because the chain pays on what it received. When the delivered quantity, the signed handover note and the chain's own receipt agree, the invoice raises itself at the received quantity; the rest reach a person with a reason. What it is worth.
Read it → -
A robot fetches the supermarket's goods receipts: 4,239 loaded by hand in 60 days, and one chain never in the system
An accountant downloads receipts from the supplier portal on the mornings someone has time: a receipt waits 45.5 hours on average, and the second chain has none in the system at all. When a robot fetches them twice a day and matches them to delivery notes, where is the value and how large is it?
Read it → -
Checking an estimate's arithmetic in one pass: 827 formulas, and a person reads only the flagged rows
Re-checking every formula of a 263-item estimate by hand takes about 4.6 hours, so nobody checks all of it. When the machine recomputes everything in one pass and flags only the rows that are wrong or suspicious, how many hours does an estimating engineer get back, and how much wrong money stops reaching a bid or a payment claim?
Read it → -
One construction project, from drawings to payment, on one screen: 288 acceptance records accepted, worth zero
On a real project, all 288 acceptance records say "accepted" while the accepted value is zero: 798 quantity links are still the machine's suggestions. Each screen on its own looks right; the break sits between the phases. How much waiting money and rework goes away when the whole project sits on one screen?
Read it → -
A site diary generated from one plan: 30 pages say "Sunny", the month had 18 rainy days
On a real construction-management workbook, 132 of 330 schedule rows were copies of the price annex and every diary page said "Sunny". When the diary, the acceptance records and the payment quantities come from one plan joined to the quantities, how many hours a month does an engineer get back?
Read it → -
The signed delivery note, checked the same day: the store signed for 9.5 kg, the note says 10
We read 113 photos of signed delivery notes: more than half the notes we could match had a line the store received short, and about half of those were never corrected. The AI reads the paper, the machine compares it with what was entered, a gap is flagged the same day — what knowing that early is worth.
Read it → -
The balance right after posting: the books say one number, the bank another
Put the book balance next to the bank's at the moment of posting, and on the first day a gap of more than a hundred million showed up. How long a gap sits unseen decides the hours spent hunting it and the chance of getting money back — reconciling at every posting against once a month.
Read it → -
When to let the machine post and when to make a person press: the auto-post gate in money
Once the draft exists, auto-posting only saves the minutes of the person pressing — while one wrong posting takes back the money of more than a hundred correct automatic ones. The gate's eight conditions, why the auto-post rate fell to zero once bank notices arrived by themselves, and the break-even threshold.
Read it → -
An accounting assistant in the group chat: from a transfer screenshot to a posted voucher
Staff send screenshots as they always did, the assistant drafts two lines, a person with the right replies “post”. Keying hours alone do not pay for running it; what turns the sum positive is books that learn of money sooner and fewer keying errors.
Read it → -
Sellable stock is not stock on hand: the ROI of an "actual stock" row that answers the right question
On the dispatch grid, "actual stock" stops adding every warehouse and no longer counts the salvage warehouse, with an optional row for stock elsewhere: the money comes from selling salvage before it spoils and from no longer promising stock that cannot be sold — how to measure each, and how far to trust it.
Read it → -
ROI of a daily stock count screen: the counted number is the stock as of when?
The storekeeper counts on a phone, the screen posts only the lines that differ, a blank cell means not counted, and every count carries a time: why a count written as an absolute number erases goods that move during the wait, how to measure three small money streams, and what the table cannot count.
Read it → -
ROI of a weekly sales forecast screen in the ERP: when the forecast sits next to what shipped
Move the weekly forecast spreadsheet into an ERP screen that puts the forecast beside actual shipments and drafts the week's purchase orders: the money comes from three streams, the largest is forecast error — not yet measured — and the piece shows how much ROI is left without it.
Read it → -
Customer returns recorded on the dispatch grid: the ROI of a number that finishes its journey
Type the return into the very cell that was delivered, so every kilo a supermarket sends back is received at cost the moment the delivery note is submitted: how far the stock ledger drifts without it, how to measure the two labour streams, and why the largest effect — books that are right — is left out of the total.
Read it → -
ROI of a component catalogue read from the IFC model: when the whole company no longer has to open Tekla
Read the IFC model into a component catalogue in the ERP, so estimating, purchasing, the shop and the site stop queuing behind the design software: the money comes from four sources, each can be measured in your own company, and the ROI still holds when you drop the source you are least sure of.
Read it → -
ROI of reading AutoCAD drawings into a quotation: where the money comes from, and how to measure it
Let a machine read AutoCAD drawings into a list of members and a quotation: the money comes from three sources, each one can be measured in your own company, and the ROI still holds when the assumptions are lowered.
Read it → -
Chicken or egg: standard ERP first, or deep customisation first?
The answer is usually stated as a principle, and a principle is hard to argue with and equally hard to believe. This piece turns it into money, and names the three numbers it stands on — set all three to zero and the paths cost the same.
Read it → -
The ROI of replacing a dispatch spreadsheet with a grid inside the ERP
The outbound plan almost always starts as a shared sheet, and it genuinely works until a few hundred orders a day. The cure is not forcing people back into ERP forms — it is putting the spreadsheet inside the system.
Read it → -
The ROI of keyboard-first data entry: 1.8 seconds a cell, times a day
The most expensive office motion in an ERP is a hand leaving the keyboard. Each instance is too small to complain about — until it is multiplied by eight thousand cells a day.
Read it → -
The ROI of customer-specific quotation and contract templates
Every partner has its own print standard, so every document detours through Word. Mapping templates to customers sounds like a small interface tweak — measured, it pays for itself inside a month at mid size.
Read it → -
Measuring the ROI of a single source of truth
Fragmented data across departments is usually treated as a technical inconvenience. Seen through a finance lens, it is a silent tax eroding cash every day.
Read it →
Editorial standard
Every piece here has to say what was measured, how, and against which baseline. A piece without a measurement is an opinion piece. Numbers taken from a client system are always anonymised, and no article ends in a "book a demo" button — read the full rules.