Where Bad Data Gets In | CRM Data Quality

Most companies we meet have cleaned their customer data at least once. Somebody spent a month merging duplicates, filling in industries and reassigning records from people who left. Three months later it looked the same as before. The cleanup was fine. The problem is that nobody looked at where the records come in. There are three doors into Zoho CRM, and each one has a control that decides what happens to a bad record at the door. Block it, flag it, or merge it. If that decision was never made, the door is open, and the cleanup is a job you do forever. Everything on screen is a mock demo of Zoho CRM with sample data. Six minutes, three doors. Here is the shape of it. Records enter a sales system three ways. A person fills in a form on your website. Someone imports a spreadsheet, usually after a trade show or a list purchase. And another system writes records through a connection: the accounting system, the help desk, the marketing tool. Each door has its own volume and its own kind of mistake. The form produces one record at a time and gets typos and fake entries. The import produces a thousand at once and brings in whatever was in the spreadsheet. The connection runs all night and repeats one mistake until somebody notices. What they share is that the record is only cheap to fix at the door. Once it is inside, it has activities, notes and a place in somebody's pipeline, and fixing it means deciding what to do with all of that too. Door one is the form on the website. In Zoho CRM the form is built inside the system, and two settings decide what it lets through. The first is which fields are required. A form that asks only for an email address will produce a record with no company, no industry and no phone number, and that record will be routed to nobody and scored as nothing. Required fields are a decision about what a lead is worth having. The second is the duplicate check. The form can be told to look for an existing record with the same email address before creating a new one, and to update the record it finds instead. That one switch is the difference between a returning visitor who fills in the form twice and two leads owned by two people who both call. The decision at this door is: what is required, and what happens on a match. Our rule is that a match updates and never creates. Door two is the import. This is where most of the damage is done, because it is done in bulk and by someone in a hurry. The import wizard in Zoho CRM asks two questions that people click past. How should the columns in the file map to fields in the system, and what should happen when a row matches a record that already exists. Skip the row, overwrite the record, or update only the empty fields. The mapping question is where an industry column becomes free text in a notes field, and where a state column spelled two ways becomes two states. The duplicate question is where a purchased list overwrites three years of notes with a vendor's guess at the company name. The decision at this door is: who is allowed to import, and which answer they give to the duplicate question. On most teams the honest answer is that one person imports, and the setting is update the empty fields only. That keeps the new information and protects the old. Door three is the connection from another system, and it is the door people forget, because nobody is standing at it. When the accounting system creates a customer, or the help desk creates a contact, it writes into Zoho CRM through the applications' own connections, at night, without a person looking. If the accounting system spells a company one way and the sales team spells it another, the connection creates a second record every time, and it does that every night until somebody notices the duplicates and blames the sales team. The control at this door is a validation rule. Zoho CRM can check a record against a rule before it is saved, whatever wrote it, and refuse the save with a message that says why. A rule that says a company must have a country, or an email must look like an email, applies to the connection the same as it applies to a person. The decision at this door is: which rules apply to records nobody typed, and who reads the refusals. A refused save that nobody reads is the same as an open door. Here is why this is no longer a tidiness question. When you turn on an AI feature in Zoho CRM, it reads the records as they are. It does not know that two companies are the same company. It does not know that ninety leads share a status because a campaign was never retired. It does not know the owner left in March. It reads what came through the door, and it answers confidently. So the order of work is not: turn the AI on and then clean up. It is: decide what happens at each door, clean up once, and then turn the AI on, because from that day the doors hold. Three doors. Three decisions. Written down, they fit on one page, and the page is the first thing we ask for in a readiness review. If your data was cleaned once and it is dirty again, do not book another cleanup. Walk the three doors first. Find the form settings, find out who imports and what they click, and find out which other systems write to your sales records at night. We do that walk as the first hour of a readiness review, and we write the three decisions down before anything else is touched. Discovery is no-risk: you pay for it only if you go ahead. The link is below.