Bustlers / Case studies
TransLine
We started with fuel. Now we calculate how much each truck earns


- Client
- TransLine — a refrigerated carrier and freight forwarder: 18 company-owned rigs, 5,000+ partner trucks, more than 100 staff
- Together
- Since 2012. In that time the company has gone through three versions of its website
- What we did
- Connected to
- 1C, Bitrix24, Max, GLONASS sensors, reports from five fuel chains
In short
It all started with fuel. Then came repair invoices, expenses, and the economics of every truck. Now TransLine can see how much is left from its trips, and together we’re getting a system ready for other carriers.
The company already had paid services, Bitrix24, and 1C. But fuel had to be pulled together separately, invoices were hunted down in email and chats, and the finance specialist collected expenses in Excel. We started with one specific problem: matching fill-ups to trucks.
Now the finance specialist is a co-author of the system: she tests it, checks the calculations, and comes up with features together with us. The owner, the manager, and the mechanic each brought their own requests. The new website also takes its data from the system — from freight rates to the temperature in the cargo area.
A trip, an invoice, the economics, and the website in 26 seconds
Real screens from the system and the website. Names, truck numbers, companies, and amounts have been changed.
What changed in the work and in the money
Invoices get paid faster, revenue is counted more accurately, and clients going quiet are spotted earlier.
15–30 min
from the mechanic’s invoice to payment
The mechanic used to wait at the repair shop for up to 5 hours
17.3%
how much fleet revenue was overstated
Amounts with and without VAT were added together. Now there is one basis
70 of 414
clients have started shipping less, and you can see them right away
As of September 17, clients over the past 12 months
$850 a month
what the paid monitoring that was switched off used to cost
The trucks are now tracked by the company’s own system
Data comes from many places.The money is counted together.
Where the data comes from
TransLine’s system
Ties trips and expenses to trucks, checks the amounts, calculates the economics
A check at the door. A file whose total doesn’t add up won’t load until a person has looked at it
Who uses it
Everything used to live separately. An invoice could arrive by email, in a chat, or in Bitrix24. The finance specialist copied amounts into Excel, and the fuel chains’ reports were combined on their own. To work out what a truck earned, you first had to collect its expenses from several places.
Now every dollar has a truck. Invoices follow a single route, files are checked before upload, and expenses are tied to trucks. From any calculation you can get back to the original invoice. If data for a month is missing, the system shows the gap.
From fill-ups to a product of its own
Together since 2012. We started the system in 2026, and each new task grew out of the one before.
-
March
Started with fuel
Combined fuel card reports into one summary by truck.
-
April
Added repairs
Learned to read invoices and track maintenance. Started the new website.
-
May–June
Worked out the fleet’s economics
Tires and overhead costs first, then the result for every truck and kilometer.
-
July–August
Linked invoices to trucks
Parts analytics, invoices through Max, and the system’s own GPS track of vehicles.
-
September
Switched off paid monitoring
Prices from the company’s own trips reached the website, and a screen for clients going quiet appeared.
-
Octobernow
Preparing a product
The new website moved to the main domain. We’re getting the system ready for other carriers.
01 · Website
Prices based on its own trips
A shipper needs a price for a trip by this evening. They enter the route on the website and immediately see a rate range calculated from the shipments the company has already done.
A site on 1C-Bitrix with libraries from 2013; on a phone you couldn’t zoom the page. The calculator pulled its price from a page on someone else’s website.
A new site with a dark design and separate pages for marketplaces, retail chains, and manufacturers. The calculator works from the company’s own trips, and when there isn’t enough data, it offers to price by the hour.
- 1A route picked from a list of 541 cities.
- 2Price as a range: the model is recalculated every day from the last 180 days of orders, and the website pulls fresh prices once an hour.
- 3A “Based on our own trips” label shows what the calculation rests on.
- 4“The cold line”: the temperature in the cargo area on a real trip.
- 5A dedicated page for marketplaces, retail chains, and manufacturers.





Figure
We test prices against real rates
74–76% of estimates land within ±20% of the actual rate. We won’t release a model to the website if it scores below 70%.
74–76%
Finding
A real trip’s temperature sells better than promises
The refrigerated freight page shows a Moscow → Novosibirsk trip from the system: not a single hour warmer than −20 °C, 36,777 readings. The shipper sees how the company keeps the cold.
02 · Fuel
Five chains, one summary
End of the month. The accountant has five reports from five fuel chains, each in its own format. The question is simple — how much did each truck fill up — but the answer gets assembled by hand.
There was no fuel tracking either in the paid order platform or in the monitoring service. The chains’ reports were combined by hand.
Files are uploaded as they are: the system recognizes the chain, checks the total, and builds the summary by truck.
- 1Reports from five chains in one summary: liters, price per liter, total.
- 2You can see what the difference from the cheapest chain cost.
- 3Fill-ups not linked to any truck show up immediately.
- 4A mismatch with the total of more than 0.5% stops the upload until someone checks it.
- 5For a driver’s card, the system suggests the truck based on the driver’s trips and GPS track.

Finding
One chain’s report lost 35% of its total on upload
The July cross-check showed it. Fuel costs looked lower than they really were.
35%
Finding
More was bought than went into the tank
We matched the chain’s reports against the fuel level sensors. On two trucks the numbers didn’t match — that’s how we found fuel being siphoned off.
03 · Repair invoices
Paid in half an hour, not half a day
The mechanic is sitting at the repair shop: the invoice has been sent, but they won’t release the truck until it’s paid. This wait used to stretch to 5 hours.
The invoice went from the mechanic to the manager, then to the finance specialist in Excel, then to the accountants — by email and in a chat. Everyone wrote the work categories their own way.
The mechanic sends the invoice through Max, the manager approves it or returns it with a comment, and the accountant processes it. From sending to payment takes 15–30 minutes, depending on how fast the manager replies.
- 1The mechanic confirms the truck from a large plate number.
- 2Mileage is checked against previous records: if it doesn’t match, the form asks for the date of the work.
- 3Four steps: file, truck, mileage, type of work. The mechanic doesn’t type in the total or the invoice number.
- 4An urgent invoice goes straight to the accounting chat.
- 5If the line items don’t add up to the total, the invoice is flagged “Check totals.”




Figure
From idea to working system — 20 days
The mockup and description appeared on July 16. Since August 5, the mechanic and the manager have been working this way.
20 days
Why
To compare tractors by repair costs
The line items show what each truck’s repairs cost and which breakdowns keep recurring. From that you can tell whether the driver takes care of the truck and whether the tractor itself was a good buy.
Our know-how
AI reads the invoice. Code checks the totals.
A supplier sends a 10-page PDF: an invoice, a work completion certificate, and a tax invoice. Recognition used to merge the documents — which produced duplicates and an inflated total. Now the AI finds the invoice pages itself, and if it finds nothing, it rereads the whole file.
The accountant enters the total, number, and date. The code checks the sum of the line items it read against that total. The scan is always alongside — a person can correct any reading.
- 01Mechanic sends the PDF
- 02Manager approves
- 03Accountant processes
- 04AI sorts the line items
- 05Code checks the total
- 06Analytics by truck
peopleAIcode
- 15 categories
- the AI sorts invoice line items into
- 119 groups
- of parts, so prices and breakdowns can be compared
- 2%
- allowed gap from the total. More than that and the invoice goes to a cross-check


The AI reads only the invoice, so the certificate and the tax invoice don’t double the total
We give the AI only what a person can check. Code calculates, people decide.
04 · Expenses
$69,000 back in the books
The finance specialist used to spend most of the month on a spreadsheet built from emails and chats: find an invoice, copy the amount, check it. Now expenses are uploaded as files, and she checks the system’s calculations and comes up with ideas with us for what to add.
Leasing advances with no date didn’t make it into expenses: $69,000 across 8 contracts. Trucks looked more profitable than they were.
Road tolls, 1C entries, payroll sheets, and leasing schedules joined fuel and repairs. Advances are now spread across months.
- 1Platon road tolls, 1C entries, payroll sheets, and leasing — each in its own format.
- 2The file goes in as it is, and the rows and total are visible before upload.
- 3Every upload is logged: the period, how many records, and for what amount.
- 4A wrong upload can be rolled back entirely with one button.
- 5A leasing advance with no date is split across the months of the schedule.

Finding
Advances on 8 leasing contracts were missing from expenses
Trucks looked more profitable than they were. Now each advance is split across months and counted in every truck’s economics.
$69K
Decision
A fine goes where it came from
A traffic police fine from the 1C entries can be assigned to a person, a truck, or the whole fleet. The system matches last names from 1C with staff and drivers on its own.
05 · Fleet economics
Profit without padding
The owner opens August. Without a data completeness check, he would have seen the best month of the year: an 85% margin and +$190,000. But fuel and salaries haven’t been entered yet.
Revenue with and without VAT was added together. Mileage was estimated, and loaded trips only. Incomplete expenses turned an ordinary month into a record one.
For every truck: revenue and expenses on one basis, excluding VAT, and mileage from the system’s own GPS track. If data is missing, the screen says so.
- 1A banner names which data is missing for the month.
- 2All amounts are put on one basis — excluding VAT.
- 3Mileage includes empty runs.
- 4For every kilometer you can see income, cost, and what’s left.
- 5The “Without leasing” view shows what the truck itself earns.

Finding
Revenue from the company’s own fleet was overstated by $160,000
From May to July, amounts with and without VAT were added together. We put them on one basis.
17.3%
Decision
We compared service prices and changed where trucks are repaired
Repairs in Kazan turned out to be noticeably more expensive, so now the trucks are repaired in neighboring Bashkortostan. There are no decisions yet about the trucks themselves: the economics screen is new, and there are few full months of data.
06 · Clients
Who started shipping less
The CEO opens the clients screen and decides who to hand to the sales team to win back. In front of him are those who started shipping less often or disappeared altogether.
There was no permanent screen: the list was built with a one-off script, joining two databases by hand.
The system knows each client’s usual rhythm and shows who has started shipping less often. The sales team has a new metric — winning back clients who are going quiet.
- 1The gap is compared with the client’s usual rhythm: “normally every 6 days.”
- 2Trips by month: you can see who’s shipping less often.
- 3A signal in each row: fewer trips, cheaper shipments, no shipments.
- 4Trips and revenue compared with the previous 30 days.
- 5Growth is flagged too, and an exit to another logistics provider is told apart from leaving the company.

Figure
As of September 17: out of 414 clients over the year
70 are shipping less, 64 have stopped, 38 are growing, and 21 are new. 114 of 295 clients use more than one logistics provider.
70 of 414
Decision
We measure silence against each client’s own rhythm
A simple average gave 156 false alarms on daily clients during development. A week without an order means one thing for a daily client and another for a client who ships once a month.
07 · Monitoring
Paid monitoring switched off
Before dropping the paid service, the owner sets conditions: mileage known for 90% of trips, a gap of no more than 2%, and a month of parallel running. Then he waits for the cross-check.
Monitoring for 36 vehicles cost about $850 a month. To drop it, they had to prove that the company’s own track measures just as well.
The custom system now receives truck positions, mileage, fuel levels, and temperature in the cargo area. After the cross-check, the paid monitoring was switched off.
- 1Kilometers come from the system’s own GPS track, not from a route estimate.
- 2Every trip shows actual and planned mileage.
- 3On the map, the actual route is drawn over the plan.
- 4A truck’s page shows trips, expenses, maintenance, fuel, and tires.
- 5You can see right away how much each truck leaves per kilometer.

Decision
The conditions for switching off were set in advance
On the left is the owner’s bar, on the right is the cross-check result: 145 trips out of 147.
Figure
What the service we switched off cost per year
About $24 a month per vehicle, 36 vehicles.
$10K
Really cool result, nobody else has anything like it. We’ll refine the system and sell it to colleagues
DmitryOwner of TransLineWe work it out together with the people who do the work
- 01
Together
The staff design the system with us
The finance specialist tests, checks the formulas, and suggests features. The owner, the manager, and the mechanic decide what should be on the screens.
- 02
Mockup
We show the screen first
The owner signs off on the mockup before any programming starts: what he’ll see and what action he’ll be able to take.
- 03
Copy
We check against real data
We test the calculations on a copy of the company’s data. That’s how we found missed expenses, other companies’ tractors in the orders, and amounts that didn’t match.
- 04
3 in a day
We fix things based on feedback
The finance specialist wrote about problems with the September fuel chain reports. Three fixes went out the same day.
carriers wrote on the first day that they’d buy the system “at any price”
Five wanted to buy the system on day one
Before there were any plans to sell it, Dmitry showed the system in an industry chat. On the very first day, five people wrote that they’d buy it “at any price.” Now we’re preparing a product for carriers.
Industry software is built by people from outside freight — they don’t know what a manager needs to calculate. Two kinds of expertise came together here:
Tired of spreadsheets?
Tell us which numbers you’re missing — no spec needed.
Or directly:
hi@bustlers.pro · +7 923 710-17-55
Request received
We have contact!
We’ll read it today and reach out to set up a call.
We’ll reply to:


