In mid-October a lab equipment distributor held its monthly review. The sales assistant exported the enquiry log from the website: 63 entries. Finance pulled the customer list from the ERP: 41 of the same companies matched. The mini program member list showed 27. Three numbers on one table, and nobody could say how many customers the company actually had. The general manager asked, “So how many customers do we have?” The room went quiet. Nobody was being lazy. Three systems were each keeping their own books, with nothing connecting them. Integration sounds like a technical word, but in business terms it means one thing: enter it once and let the other systems follow. Without it, people copy data by hand.

One customer, three half-records

The website enquiry form, the ERP customer file and the mall member list are usually kept by three different people. The marketing specialist sees a new enquiry and forwards it by chat. The sales rep creates a customer in the ERP and shortens the company name along the way. The same company ordered once through the mini program and registered with another phone number. Each record is right on its own. Put together, nobody can tell whether this is one customer or three.

The direct cost is duplicate work and sales collisions. Two reps quote the same company at different prices, and the buyer asks why the company cannot keep its own story straight. A service agent cannot see what a repeat customer bought three weeks ago and has to ask again. No customer walks away over one failed lookup, but the next purchase call goes somewhere else.

The warehouse hears about orders late

In retail and wholesale there is a familiar scene: an order arrives on the mall or in the mini program, and the warehouse only learns about it through a screenshot in a group chat, a spreadsheet, or a phone call. The store owner is still asking at eight in the evening whether the afternoon order has shipped, and the warehouse keeper scrolls back through messages to confirm.

When order details travel by hand, some of them fall out. One missing order means a customer calls the next morning, calls again, then asks for a refund and leaves a short review. On a promotion day with hundreds of orders, manual copying goes wrong easily, and the cost of wrong, missing or duplicated shipments often exceeds that day's margin. Warehouse and service staff blame each other, while the real gap sits in the middle where nobody owns the handover.

Staff working as human middleware

Some roles exist mainly to move numbers from one system into another: mall orders typed into the ERP one by one, website enquiries copied into a customer sheet, both sides exported and checked at month end. At two hundred orders a day and forty seconds each, that is over two hours of typing, before interruptions, rework and corrections.

The cost comes in two layers. One is salary and overtime, which adds up over a year. The other is turnover: people in these roles rarely stay long, handovers are rushed, and whoever arrives next does not know which fields to fill, so the error rate climbs. The business pays for a person and receives a stretch of repetitive work that did not need to exist.

A three-thousand-yuan gap that took three days to trace

At month end, finance compares three sets of numbers: mall receipts, ERP receivables and bank statements. They do not agree. The gap is usually small, a few thousand yuan, but it has to be explained. Finance works through it line by line, sales staff are pulled in to remember which order was adjusted, and three days disappear. Normal reimbursements and payments wait behind them.

The gap usually comes from something missing or something counted twice: an order paid in the mall that never became a receivable in the ERP, a refund processed at the payment provider but not cleared on either side. After a few rounds, the owner stops trusting the numbers and re-checks every report by hand. Unreliable data is harder to work with than no data, because it makes people hesitate before deciding.

Why a favour-based integration ends up half finished

To save trouble, some companies ask their old website vendor or a freelancer to “just connect the two systems.” The quote is low and data does move, but in use it turns out to be one-way only, or the fields do not line up: one side calls it customer name, the other calls it cust_name; one phone number carries an area code, the other does not; one date uses dashes, the other slashes. Half of what arrives is wrong, and the rest still needs a person to fix it.

Then there is history. The old system holds four records for the same customer under slightly different names, and importing them straight into the new system multiplies the duplication by four. An incomplete integration looks connected and still needs daily manual patching: the money is spent, the work remains, and any future change has to be made inside somebody else's code.

Unstable integrations cost money off the books

An integration that works most of the time is easy to ignore. A request times out at peak, the customer sees a confirmed order that never reaches the back office; stock sync lags by a few minutes and the same item oversells during a promotion, leaving inventory at a negative number; loyalty points post twice and service handles a dozen complaints in a day.

None of this appears on a monthly report. It shows up as service hours, refund rates, review scores and weaker repeat purchase, and by the time the owner notices, it has been running for months. That is the nature of plumbing: invisible while it works, and felt across every department when it does not.

Three business signals that justify the budget

You do not need a technical view to decide whether integration belongs in the budget. Three business measures are enough. Is the same data copied by hand more than twice a day? Does that data touch orders, stock, payments or member benefits? Is there already a role whose main job is re-typing numbers? If any one is true, the process is worth investing in.

Once the decision is made, confirm three things early: whether the existing systems allow external access at all, which side is the source of truth for each field, and who watches the connection and handles exceptions after launch. Written into the contract, these three points prevent most of the arguments that follow.

Having several systems is not the problem. Systems that never speak to each other are. Integration is not another IT project; it is a way to move staff out of repetitive copying and bring three scattered records back into one. Xi'an Zunyun Technology has handled a fair number of these connection projects between websites, ERPs, malls and official accounts, and they tend to be among the easier ones to justify on return. List the systems you run today, and we can look together at which links are worth connecting first.