A paint and varnish manufacturer based in Eskisehir — 312 employees, mid-sized by any standard — launched an ERP project last year with the kind of confidence that comes from a clear business case. The new system was going to cut production costs, eliminate excess stock and help the company fulfil customer orders faster. By the second month of implementation, however, the project had hit a wall that no one had anticipated: more than half of the 1,400-plus raw material items used in production either had no defined bill of materials at all, or existed in multiple conflicting versions across different departments. The software was ready. The infrastructure was ready. User training sessions had been scheduled. The only thing missing was the data itself.Working in the chemical sector forces you to confront a structural complexity that most other manufacturing verticals do not share to the same degree. A paint formulation, an adhesive compound or a cleaning product recipe is not a static document. When a formula changes, it rarely stops at swapping one ingredient. Storage requirements may shift, packaging specifications may need updating and — in a regulated product category — the legal classification can change entirely. These changes can be logged on paper, but when they are not tracked inside a single system, production, cost accounting and procurement will continue operating from different versions of the same reality. The real danger is not the discrepancy itself: it is that everyone in the organisation believes their version is the correct one.The thesis I want to defend here is a direct one: in the chemical sector, the single greatest obstacle to ERP success is not the choice of software. It is the failure to consolidate and clean bill of materials data under a single ownership structure before the project begins. This sounds straightforward, but the resistance it meets in practice is remarkable. The production manager believes his data is accurate. The R&D team is confident that the formulation archive is current. Procurement insists that the supplier-linked item list is complete. All three are right — from inside their own window. The problem is that none of these windows looks into the same room.In the Eskisehir project — documented here as an anonymous case — we ran a six-week data inventory exercise before a single line of software configuration was touched. A named owner was assigned to every raw material item actually in use in production; that person became accountable for both the technical specification and the system record. Conflicting entries were placed side by side, and production and R&D staff resolved each discrepancy together in the same room. What we found was that 312 of the 1,400 items carried more than one item code, and 89 items had no record at all. Had we gone live without this exercise, every one of those errors would have been embedded in the system. A separate sub-module in the same project — the production order component — was deployed ahead of schedule on a batch of uncleaned data, and within 14 days it required three separate manual corrections to produce accurate output. That figure alone justified the entire pre-project data effort.When you move from data to operational process management, a second structural problem surfaces. In the chemical sector, the number of companies that can describe a single production process consistently across all relevant departments is smaller than you would expect. The production shift form, the quality control checklist and the cost accounting entry often document the same process in ways that do not match. When an ERP brings these three sources together into a single report, the inconsistencies become visible — and when management cannot decide which figure to trust, the system loses credibility. Users keep their parallel spreadsheets open. Within six months, the ERP has quietly become the secondary system rather than the primary one. This pattern recurs across Turkish chemical companies of this size, and it begins almost always with the same root cause: processes were never written down in a way that every stakeholder could agree on.The remedy is simple in concept and time-consuming in practice. Before configuring any process workflow inside the software, you need to run what I call a process validation session: every department that touches a production line sits in the same room and describes the process step by step until everyone agrees on a single written version. In Eskisehir, we ran these sessions for eight separate production lines. Each session lasted an average of three hours. The written process descriptions that came out of those sessions were, on average, %57 shorter in step count than the shift forms that had been in use before — not because the process had been improved, but because redundant and duplicate steps that existed only on paper were finally identified and cut. The ERP then modelled a leaner, agreed-upon reality rather than an inflated, contested one.For any manager preparing to run an ERP project in a chemical manufacturing environment, three steps are worth applying before the vendor selection conversation even begins. First, count how many separate sources currently hold your bill of materials data. If the answer is more than one, reserve at least %62 of your project budget and timeline for data consolidation work — not configuration, not training, but data. Second, do not ask ‘who knows this process best?’ before you map your operational workflows. Ask instead ‘which departments describe this process differently?’ The map of disagreements is more valuable than any single expert opinion, because the ERP will need to serve all departments, not just the most informed one. Third, when you meet with an ERP consultant, do not ask for a feature list. Ask for a data ownership model: who enters which data, under whose approval, and with what frequency. A system that lacks a clear data ownership structure will run correctly on the day it goes live and degrade steadily for the next twelve months. Cleaning the data before the project is painful. Rebuilding it afterwards, while the business is running on a live system, is genuinely costly — and in the chemical sector, where a single formulation error can affect an entire production batch, the margin for that kind of error is essentially zero.
This article was originally published in Turkish by Gökhan MERCANOĞLU on January 20, 2007. The English edition has been reviewed and edited by the author.