How to clean CRM data before a calling campaign: 8 steps
Sado AI team · · 4 min read
In short
Before a calling campaign, clean your CRM base in eight steps: keep only people you may call, remove those who said “don’t call”, bring numbers to one format, delete incomplete numbers and duplicates, fix the name column, set aside customers with open deals or complaints, split the list by purpose and test on a small group first. A clean base doesn’t annoy customers, doesn’t spend money on calls to the wrong people and keeps your results honest.
Why clean it: an example
For example, the Oltin Kalit property agency exported 4,000 contacts from its CRM. After checking: 550 duplicates, 230 incomplete or wrong numbers, 310 people who had asked not to be called, and 420 with an open deal with a manager. That left 2,490 people to call. The agency called 2,490 people instead of 4,000, and nobody got two calls. The numbers are only an example.
Unanswered numbers cost nothing, but if the wrong person picks up, the call still happens: you lose minutes and some of your reputation. And a customer called twice, or a buyer already talking to a manager, will be annoyed.
The 8 steps
- Who you may call. Keep only people who left a request, gave consent or are customers. Remove bought lists.
- “Don’t call”. Exclude contacts who refused before and are marked so in the CRM.
- Number format. Bring everything to +998 XX XXX XX XX. Delete or fix incomplete numbers (fewer than 9 digits).
- Duplicates. Each number once on the list. More: duplicate leads in the CRM.
- Name column. The operator greets customers by name, so fix entries like “Dilshod warehouse”, “test” or “Customer 1”, or leave the name empty.
- Open matters. Set aside customers currently talking to a manager, with an open complaint, or who recently bought the same product.
- One list, one purpose. Don’t mix news about a new building and a payment reminder in one list.
- Test run. Call 50 numbers first, listen to 5–10 calls, then start the rest.
Plan the calling hours too. In the portal you choose “start now” (every day 9:00–20:00) or “working hours only” (Mon–Fri, 9:00–18:00), and the extra settings change the days and the daily limit. Allow for Friday prayers and iftar time during Ramadan.
Checking column by column
| Column | What to check | Typical mistake |
|---|---|---|
| Phone | +998 and 9 digits, one format | “90 123 45”, an incomplete number |
| Name | Just the name, nothing extra | “Akmal (old customer)”, “test” |
| Last contact | Not too long ago | A lead from five years back |
| Status | Open deal, complaint, refused | A “do not call” mark ignored |
| Source | Where it came from, consent or not | A list of unknown origin |
Many of these can be checked with CRM filters without exporting to Excel. The rest you clean in Excel with filters and sorting.
Handing the clean list to an AI operator
You upload the list as an Excel file to the portal’s Auto-calls section. The portal first asks “Who are these people?” and saves the answer, shows how many numbers from the file were accepted and, before you start, the estimated cost. One auto-call holds up to 10,000 people. Step by step: calling a list from Excel.
You can skip Excel too: with Bitrix24 or amoCRM connected, you pick the CRM as the source or tie the call to a deal stage. Set a daily limit, and stop everything with one button if needed. Sado never calls again a number that said “don’t call me”. More: AI outbound calls.
Questions
Do I pay for wrong numbers?
Unanswered and unconnected numbers cost nothing: you pay only for talked minutes, 800 so‘m a minute. But if someone else picks up, that conversation counts as minutes too.
May we call an old customer base?
If they were your customers or left a request, usually yes, but choose very old records with care. Uzbekistan has a law on personal data; if in doubt, ask a lawyer.
Do we have to clean the base every time?
Before every large campaign, yes. Afterwards, bring the results back into the CRM: with wrong numbers and refusals marked, there’s much less to do next time. How to export results: call results to Excel.