The most expensive line in an ERP project is not the licence fee. That became clear in the very first meeting with a water pump manufacturer based in Konya. The company had 284 employees, a product range covering both industrial and agricultural pumps, more than 40 authorised service points across Turkey and an active export line. Yet sales orders, service records, production planning and finance were each moving through four disconnected file systems. Management could not close the books monthly; the plant manager resolved production priorities by scrolling through email threads; the service department maintained its spare parts stock in a spreadsheet. Purchasing an ERP system looked like the obvious answer. The real problem had started earlier: no one had concretely defined which processes actually needed to be integrated and how.ERP implementations in machinery manufacturing are structurally different from retail or service sectors. In pump production, every order carries a technical specification behind it — head pressure, flow rate, motor power, material selection. An ERP system that fails to bridge the gap between a customer order and a production order through these technical parameters can never correctly build the chain through which sales triggers manufacturing. At this Konya firm, that was precisely the situation. The sales team received orders on their own forms and forwarded them to the plant manager by fax or email. Production planning then re-entered that information into its own spreadsheet. The same data was created manually in two separate places twice over, carrying two separate error risks. During the process-mapping work carried out before the project kicked off, it turned out that fully entering a single order into the system took an average of three working days. Some 57 per cent of that time was consumed by data entry and verification — production had not yet begun.The service department presented a separate and equally telling picture. Fault reports from authorised service points arrived by telephone or fax, a service coordinator noted them in a ledger, and spare-part requests were posted as a separate accounting entry. Service history and warranty status were not linked. When a customer requested service for a pump whose warranty had expired, the service team could only find this out by contacting the sales department for confirmation. Building that bridge made it clear that sales-service integration was not merely a customer-relations question; it was fundamentally a production-record question. Being able to trace a pump’s serial number back to its production batch and the material lot that fed it was the basic prerequisite for both quality tracking and warranty management. An ERP system can create that traceability — but only if the serial number and production batch were correctly entered into the system from the very beginning.Finance integration is where resistance was strongest. The accounting department had been closing its books on its own dedicated software for nearly a decade. Moving to the ERP finance module meant the accounting team had to change habits built up over those years. In more concrete terms: accounts receivable, cash, bank movements and journal entries all had to be relearned in a new interface. A very practical complication arose at this stage. The firm had an active e-Beyanname setup, with a continuous declaration flow connected to the Revenue Administration system. Was the ERP finance module compatible with this flow, and how would declaration data be transferred without triggering duplicate entries? Projects that ask this question too late end up with the accounting team running two systems simultaneously on go-live day — inefficient and full of error risk. At this firm, we resolved the compatibility issue during an eleven-week parallel-run period, building a file-transfer bridge between the ERP finance module and the existing e-Beyanname infrastructure.User habits were the most underestimated dimension of the project. Operators on the production floor were accustomed to following production orders from paper forms. The ERP required them to enter data at a terminal. The terminals were in place, the ADSL connection was running on the office floor, and the local network had been extended to the factory floor. But acquiring new habits requires repetition, not just training. What is learned in a training room is forgotten under production pressure. For this reason we delayed the production module go-live by six weeks relative to the sales and finance modules, and dedicated those six weeks to on-site coaching on the floor. That decision changed the outcome. Production data-entry accuracy reached 73 per cent within the first two months; in comparable projects this figure tends to be far lower at the moment of go-live, and the cost of error correction compounds quickly.Looking back at the end of the project, one conclusion stands firm: ERP is a powerful tool for moving the four departments of a pump manufacturer onto a common data foundation. But that tool generates value only after the process map has been drawn, the integration points defined one by one and real investment made in changing user habits. The Konya firm reduced its order-to-production cycle from three working days to under four hours over the eleven-month project; the service team could check warranty status directly in the system instead of calling the accounts department; the monthly accounting close dropped from several weeks of effort to three days. These outcomes belong to correctly analysed processes, not to the software itself. To every machinery manufacturer considering ERP, I put one question: before you purchase the system, have you written down on paper exactly where the processes that are currently disconnected need to come together — and at which precise point? If that paper does not exist, the outcome will be disappointment regardless of the project budget.
This article was originally published in Turkish by Gökhan MERCANOĞLU on January 17, 2006. The English edition has been reviewed and edited by the author.