# How data quality prevents wrong business decisions > Duplicate, incomplete or stale data distorts reports and decisions. See the most common problems, what they cost, and how to keep data you can actually trust. - Author: Marlon Trettin - Published: 2026-06-18 ยท Updated: 2026-08-21 - Language: en - Canonical: https://yowpi.com/en/blog/how-data-quality-prevents-wrong-business-decisions --- Incomplete, duplicated or outdated data distorts reports and produces wrong decisions that look certain. Avoiding that runs on three fronts: standardizing how information is entered, automating validation and cross-checks, and designing internal systems that preserve the integrity of a record from entry to report. The tool is rarely the problem. ## TL;DR - A wrong report does not announce that it is wrong. Decisions about inventory, pricing and budget quietly inherit the defects of the data feeding them. - The cost is measurable. Thomas Redman, writing in [MIT Sloan Management Review](https://sloanreview.mit.edu/article/seizing-opportunity-in-data-quality/), puts bad data at 15% to 25% of revenue for most companies. - The bar is lower than managers assume: research in [Harvard Business Review](https://hbr.org/2017/09/only-3-of-companies-data-meets-basic-quality-standards) found only 3% of company data meeting basic quality standards. - Data quality is solved in the architecture: standardized entry, automatic validation, and systems that talk to each other. ## Why does bad data lead to wrong decisions? Picture a distributor where the sales director asks to replace the CRM because sales numbers never match finance. The board approves the budget, the migration starts, and three months later the reports still disagree. When somebody finally audits the database, the cause turns up: duplicated customer records, charges logged without a link to an order, and fields filled differently by each salesperson. The tool was never the problem. What entered it was broken. That scenario is hypothetical, but the pattern repeats at companies of every size. When data is stale, incomplete or wrong, the analysis stops being trustworthy. It is like trusting a GPS that shows streets which do not exist. The driver's skill is irrelevant, the destination is still wrong. In practice the distortion lands on the most sensitive parts of management. - Inventory decisions produce shortages or overstock. - Financial forecasts miss because they start from a broken history. - Strategy leans on guesswork dressed up as a report. - Service fails because orders, deadlines and contacts were recorded wrong. The perverse detail is that nobody notices at the time. A report built on bad data looks exactly like a report built on good data. ## Which quality problems show up most often? Before investing in correction, it helps to know what to look for. Six problems account for most of the damage. - **Duplicate or conflicting records.** The same customer entered three times, each version holding different data. - **Blank fields.** Required information nobody filled in because the system never demanded it. - **Stale data.** Addresses, prices and terms that changed in the real world but not in the registry. - **Inconsistencies between systems.** The ERP says one thing, the CRM says another, and each department trusts its own version. Much of this starts with [failures in the integrations between internal systems](/en/blog/5-failures-that-break-internal-system-integrations). - **Manual entry without validation.** Free typing, no mask, no check. Every record becomes a lottery. - **No fill-in standard.** "NY", "New York" and "new york" turn into three different states in the report. Each of these produces rework and a distorted picture of the operation. Added together, they produce decisions that work against the company's own strategy. ## What does deciding on bad data cost? More than it looks, and there are numbers for it. Thomas Redman estimated in [MIT Sloan Management Review](https://sloanreview.mit.edu/article/seizing-opportunity-in-data-quality/) that bad data costs 15% to 25% of revenue for most companies, counting rework and lost opportunity. He also argues that about two thirds of those costs can be identified and removed permanently, which makes this one of the rare problems where the fix compounds. The quality bar is lower than managers assume. Research by Tadhg Nagle, Thomas Redman and David Sammon, published in [Harvard Business Review](https://hbr.org/2017/09/only-3-of-companies-data-meets-basic-quality-standards), concluded that only 3% of the data at the companies analyzed met basic quality standards. For a mid-sized company those percentages have concrete names: stock bought that does not move, a campaign aimed at an audience that does not exist, a price calculated on an outdated cost. ## How do you get data you can trust? Turning scattered data into reliable information takes method, and the method fits in three pillars. - **Standardized entry.** Required fields, closed lists, input masks and clear rules from the very first record. Blocking the error at the door is cheaper than hunting it later. - **Continuous validation.** Automatic routines that cross-check databases, flag divergences and mark suspicious records before they reach a report. Audit logs help trace where and how a record changed. - **Cross-checking and enrichment.** Automations that compare information between systems and fill gaps, instead of depending on somebody remembering to check. ![Analyst pointing at charts and data dashboards across three side-by-side monitors](/blog/qualidade-dados-decisoes-certas-1.webp) Insisting on manual controls and unvalidated spreadsheets works against all three pillars at once. Free-form entry, where each person fills things in however they like, guarantees inconsistency at the source. No later analysis fully repairs that. ## What role do internal systems play here? A well designed internal system sees past the entry screen. It models the company's real flows, validates information at the moment it arrives, and holds integrity end to end, from order to invoice. A beautiful dashboard on a bad database is just well presented error. Centralizing matters too, with one caveat: pooling bad data in one place produces a large pool of bad data. Ordering the entry rules comes first, and the same reasoning decides [when to replace spreadsheets with custom software](/en/blog/when-to-replace-spreadsheets-with-custom-software). ![Team gathered in a meeting room in front of a large screen showing data flows, charts and diagrams](/blog/qualidade-dados-decisoes-certas-2.webp) When the base is solid, the decision cycle flows: the record is collected and stored correctly, checks run automatically, analysis starts from consistent reports, and action comes quickly because people trust what the number says. In the custom systems we build, that validation is designed alongside the process, because fixing data on the way out always costs more than validating it on the way in. Deciding on sound data buys something beyond safety. It lets you see problems and opportunities while they are still cheap. ## Frequently asked questions **The reports do not match. Should I replace the system?** Not yet. Before any migration, audit the current base: look for duplicates, empty fields and divergences between systems. If the problem is in the data, replacing the software only moves the mess into a newer one, with the project cost on top. Migration makes sense when the current system actively prevents the validation and integration the operation needs. **Is data quality IT's responsibility?** IT looks after the infrastructure, but the people who generate and consume the data are in the business areas. Without a clear owner per process, each department blames the other and nothing changes. The arrangement that works names owners for the critical records, agrees fill-in rules across departments, and automates validation so nothing depends on individual discipline. **Where do I start if the base is already a mess?** With the data behind the most expensive decision. If the company struggles with inventory, start with product records and stock movements. Fixing everything at once paralyzes the business. Fixing what hurts first produces a visible result that funds the next stage. In parallel, standardize entry so the new mess stops growing. ## Next step If your decisions depend on reports nobody fully trusts, the problem is rarely the team or the tool. It is the architecture that lets the record break along the way. An Operational Architecture Diagnostic maps, in 30 minutes, where information gets lost between entry and report. [Book a conversation](/en/contact).