When Bangladesh's internet went down in July 2024, an e-commerce company's online revenue fell to zero. Calling past customers brought back almost half of it. What worked, what broke, and how to run data calling properly.
In July 2024, the internet in Bangladesh went down. For the e-commerce company I was working with, that meant the website, Facebook and paid ads, the channels that brought in almost every order, stopped at the same time. Online revenue went to zero.
The company still had one asset that did not depend on the internet: its customer database. With the data team's help, it started calling past customers to ask whether they needed anything. That month, data calls brought in more than ৳1 crore, close to half of what the company normally made through all of its online channels combined. The channel did not stop when the internet came back. It later grew to around ৳10 crore.
What data calling is, and what it is not
Data calling means phoning customers who have already bought from you, using your own order history, to create new repeat orders. It is proactive selling to people who already know and trust your brand.
It is not cold calling, because every number belongs to an existing customer. It is also not order confirmation: you are not checking an order that already exists, you are creating a new one.
How it started: one Excel export
The first batch was simple. The customer database was exported to Excel, formatted, and handed to the customer care agents the company already employed. The list held thousands of customers who had bought food, fashion and healthcare products.
What the numbers showed
- Calling the general customer list converted at more than 10%.
- Short, targeted lists converted at 40–50%.
- In the shutdown month, data calls recovered close to half of normal monthly revenue: more than ৳1 crore.
- No new hires were needed. Existing customer care agents made the calls.
That last point matters for any seller doing the maths. Most e-commerce companies already pay for customer care agents. Data calling puts that capacity to work on revenue instead of adding a new cost line.
What broke when calling at scale
Calling everyone the same way did not last. Customers reacted very differently:
- Some felt genuinely helped by the call and placed an order.
- Some were angry, asked never to be called again, and said they would order on their own when they needed something.
- Some were happy to be called, but only after a certain gap: a week for some, a month for others.
A spreadsheet cannot remember any of this. Each new export starts from zero, so the customer who said no last week is back on the list, and the customer who asked for a call in a month gets one tomorrow. After frustrated customers started pushing back, the team had to build calling logic on top of the raw list.
The reputation risk
Data calling only works within limits. Calling the same people too often, or calling again after a customer has said no, turns a revenue channel into a reputation problem. Customers remember the brand that would not stop calling.
The rules that make data calling work
- Treat "no" as permanent. A do-not-call request goes on a blocklist and stays there.
- Respect each customer's timing. If someone asks to be called next month, nobody calls them before then.
- Set a calling frequency. A simple rule such as "call each customer at most every N days" protects the customer and spreads your list across the month.
- Call the right list. A targeted short list converted at 40–50% against just over 10% for the general list.
- Record every outcome, so tomorrow's list is better than today's.
Where smarter targeting fits
In my experience, the biggest gains come from finding the gold customers in a large list: the ones most likely to buy again. That is where AI will add the most value. CallBridge starts with transparent, rule-based selection, so a manager can always explain why a customer was called today.
How CallBridge turns data calling into a system
We built CallBridge at EXCO to run this process properly. It does not replace your ERP. It reads your data and works your customer list.
- Data in: customer and order data arrives daily from your existing ERP by automated CSV or ETL. Phone numbers are normalised to Bangladesh format and duplicates are removed.
- Who to call: an eligibility engine applies the frequency rule your manager sets. A tighter interval means more calls, a looser one means fewer.
- Batches: the manager builds a batch sized to the daily quota and the number of active agents.
- Calling: agents work one customer at a time and record one of six outcomes: Ordered, Interested (call back), Not interested, Do not call again, No answer or Invalid number.
- Learning: outcomes feed straight back. Do-not-call removes the customer, a callback snoozes them, and an order resets their cycle.
Protecting customer data from the inside
When agents call from a spreadsheet, every phone number and address is on their screen and easy to copy. In CallBridge, the agent console is locked, watermarked and copy-protected. Agents see the customer's name, area and data-call history, but no phone number and no address. A Call button opens their dialer.
Contact details appear only when the agent confirms an order, as copy-only fields for entering the order into your ERP. Every reveal is logged against the agent.
Catching agents who fake orders
Because contact details unlock on a confirmed order, the real fraud risk is internal: an agent faking "Ordered" outcomes to collect phone numbers. CallBridge tracks confirmed orders with no amount for each agent. The manager cockpit we are building now adds next-day reconciliation: confirmed orders are matched against the next day's ERP data by phone number, and confirmations that never become real orders flag the agent.
Is data calling right for your business?
- You have a customer base with order history in an ERP or database.
- You already employ customer care or sales agents.
- Your customers buy again, as with food, fashion or healthcare products.
- You want revenue that does not depend entirely on ads and platforms.
If that sounds like your business, data calling is likely one of the cheapest revenue channels you are not running properly yet. We are onboarding a small group of sellers to the CallBridge pilot.