Chatbots in messengers as a sales and service channel for the brand
When a chatbot brings losses, not profits
The bot starts operating at a loss not due to a poor platform, but due to incorrect task setting. The company invests budget into developing complex scenarios, but forgets that a bot can close 60-80% of inquiries without an operator only with proper setup — the rest inevitably requires a human, and if escalations are not provided, the customer gets stuck in a loop of 'select a menu item' and leaves for a competitor.
- The bot is implemented as a full replacement for the support service, not as a first filter for inquiries — complex and emotional cases remain without an outlet to a person.
- The scenario is written according to the company's logic, not according to customer requests: a menu with eight levels instead of a simple text response.
- No CRM integration — the bot collects data that goes nowhere, and managers lose contact with an already warm lead.
- The bot is launched for a product with a high trust threshold (real estate, expensive services, medicine) without the ability to confirm information by a human.
- They are making bets on the 'friendliness' of dialogue at the expense of accuracy: a study published in Nature (with participation of the Oxford Internet Institute) showed that warm, conversational bot responses contain 7,43 percentage points more errors and confirm users' false beliefs in 40% of cases — this is a direct path to complaints and reputational losses.
Why people don't trust chatbots
Distrust of the bot is not an irrational fear, but a rational reaction to real experience. If we break down client objections, we will get three specific barriers, not the abstract 'people are conservative.'
First of all, 72% of consumers still prefer live communication, especially when the issue goes beyond the typical. Secondly, 66% are not ready to pay directly inside the chat assistant — money through an unfamiliar interface is perceived as a risk. Thirdly, 79% are concerned about data confidentiality which they enter in communication with the bot.
These figures point to a critical moment before payment: a significant portion of potential buyers stops at this step. Most want to see real customer reviews right in the dialogue, and many are not ready to pay without brand confirmation on the official website. Hence the practical conclusion: a bot that leads to one-click payment without intermediate proofs loses one-third of the audience at the final step. The solution is not to remove payment from the bot, but to embed reviews, a link to the product card on the website, and an explicit transition to a live operator upon request into the scenario.
Architecture: What Sets a Converting Bot Apart from an Alienating One
The difference between a bot that sells and a bot that irritates rarely lies in the technology — it is almost always in the scenario. A good bot conducts a dialogue like a sales consultant who knows when to remain silent and hand over the conversation to a human. A bad one tries to solve everything by itself, because it's cheaper for the business, but more expensive for customer relations.
| Parameter | Converting Scenario | Alienating Scenario |
|---|---|---|
| Entry into the Dialogue | One clear question, buttons with clear formulations | General greeting + menu of 6-8 items |
| Escalation to a Human | Available at any step with a single word or button | Hidden in a submenu or completely absent |
| Payment | Link to the website/CRM with order confirmation | Payment within the chat without brand verification |
| Tone of communication | Precise, short, without false 'humanity' | Excessively friendly, with a risk of actual errors |
| Client data | Immediately transferred to the CRM manager | Remain in the conversation, get lost when the operator changes |
Case: how retail reduced the load and increased open rate through the bot
The network «Daughters-Sons» transferred part of voice inquiries into chat-bot mode and achieved a 25-29% reduction in operator load — meanwhile, during peak periods (sales, start of the academic year), the service was able to serve 50% more clients without expanding the support staff. An interesting figure for the channel: the open rate of messages in the Telegram bot was 10-11 times higher than in email newsletters of the same network.
The reason is not in Telegram's magic, but in the channel's design: a bot's message enters a personal chat that a person checks constantly, not a folder «Promotions» that is opened once a week. The same applies to sales: conversion to an order through a messenger bot for ZARINA was 4 times higher than through email, and integrating the bot with a loyalty program added +30% revenue per customer. The conclusion from this case is simple — a bot wins where it solves a narrow task (reminder, notification, fast order), not where it replaces a full consultation.
Metrics and KPI: How to calculate ROI fairly
The report «bot replied to 10 000 messages» says nothing about its efficiency. The key indicators are those related to money and customer retention, not the bot's activity itself.
- The share of inquiries closed by the bot without operator involvement — a benchmark of 60-80% with proper scenario settings; if the indicator is below 40%, the bot works as decoration.
- Conversion to lead with contact details — WhatsApp Business showed 100% conversion from inquiry to lead with proper CRM integration (case AVILON).
- The cost of a lead through the VK channel — after the script was finalized, leads cost 40 rubles and arrived daily for free, without additional advertising budget.
- ROI of database reactivation — with G.M., the bot integration with the loyalty program yielded a ROI of 3488% on reactivation campaigns, which is significantly higher than email channel metrics.
- Increase in purchase frequency and revenue per customer — 'Smile of the Rainbow' achieved +40% in purchase frequency and +25% in revenue per customer precisely due to the bot and loyalty program combination.
The ROI formula for a chat-bot is calculated the same way as for any channel: (revenue attributed to the bot minus development and support costs) divided by costs, multiplied by 100%. The mistake in most calculations is not subtracting the cost of maintaining the scripts and not separating 'bot revenue' from the revenue that the client would have brought without the bot.
Which products are sold through the bot, and which are not
A strategic mistake is to botize everything indiscriminately, guided by the trend of the channel, rather than the nature of the product. The product category determines whether the bot will be a helper or a barrier.
| Suitable for a bot | Does not sell well through a bot |
|---|---|
| Repeat purchase products (cosmetics, household chemicals, children's goods) | Expensive real estate and premium services |
| Tickets, bookings, simple services with fixed pricing | Complex B2B-sales with a long negotiation cycle |
| Loyalty programs and reminders about remaining points | Medical consultations requiring doctor's responsibility |
| Flash sales and promotions with limited time | Products with high emotional engagement (wedding services, exclusives) |
| Notifications about order status, delivery, return | Products where the decision is made only after a live consultation |
General rule: the higher the cost of error for the client and the more unique the decision they make, the less automation should be in the dialogue and more should be quick access to a human. A bot that honestly says 'this issue is better resolved with a manager' and immediately transfers the contact converts better than a bot that tries to close the deal on its own where the client needs assurance, not a script.
When a chatbot brings losses, not profits
The bot starts operating at a loss not due to a poor platform, but due to incorrect task setting. The company invests budget into developing complex scenarios, but forgets that a bot can close 60-80% of inquiries without an operator only with proper setup — the rest inevitably requires a human, and if escalations are not provided, the customer gets stuck in a loop of 'select a menu item' and leaves for a competitor.
- The bot is implemented as a full replacement for the support service, not as a first filter for inquiries — complex and emotional cases remain without an outlet to a person.
- The scenario is written according to the company's logic, not according to customer requests: a menu with eight levels instead of a simple text response.
- No CRM integration — the bot collects data that goes nowhere, and managers lose contact with an already warm lead.
- The bot is launched for a product with a high trust threshold (real estate, expensive services, medicine) without the ability to confirm information by a human.
- They are making bets on the 'friendliness' of dialogue at the expense of accuracy: a study published in Nature (with participation of the Oxford Internet Institute) showed that warm, conversational bot responses contain 7,43 percentage points more errors and confirm users' false beliefs in 40% of cases — this is a direct path to complaints and reputational losses.
Why people don't trust chatbots
Distrust of the bot is not an irrational fear, but a rational reaction to real experience. If we break down client objections, we will get three specific barriers, not the abstract 'people are conservative.'
First of all, 72% of consumers still prefer live communication, especially when the issue goes beyond the typical. Secondly, 66% are not ready to pay directly inside the chat assistant — money through an unfamiliar interface is perceived as a risk. Thirdly, 79% are concerned about data confidentiality which they enter in communication with the bot.
These figures point to a critical moment before payment: a significant portion of potential buyers stops at this step. Most want to see real customer reviews right in the dialogue, and many are not ready to pay without brand confirmation on the official website. Hence the practical conclusion: a bot that leads to one-click payment without intermediate proofs loses one-third of the audience at the final step. The solution is not to remove payment from the bot, but to embed reviews, a link to the product card on the website, and an explicit transition to a live operator upon request into the scenario.
Architecture: What Sets a Converting Bot Apart from an Alienating One
The difference between a bot that sells and a bot that irritates rarely lies in the technology — it is almost always in the scenario. A good bot conducts a dialogue like a sales consultant who knows when to remain silent and hand over the conversation to a human. A bad one tries to solve everything by itself, because it's cheaper for the business, but more expensive for customer relations.
| Parameter | Converting Scenario | Alienating Scenario |
|---|---|---|
| Entry into the Dialogue | One clear question, buttons with clear formulations | General greeting + menu of 6-8 items |
| Escalation to a Human | Available at any step with a single word or button | Hidden in a submenu or completely absent |
| Payment | Link to the website/CRM with order confirmation | Payment within the chat without brand verification |
| Tone of communication | Precise, short, without false 'humanity' | Excessively friendly, with a risk of actual errors |
| Client data | Immediately transferred to the CRM manager | Remain in the conversation, get lost when the operator changes |
Case: how retail reduced the load and increased open rate through the bot
The network «Daughters-Sons» transferred part of voice inquiries into chat-bot mode and achieved a 25-29% reduction in operator load — meanwhile, during peak periods (sales, start of the academic year), the service was able to serve 50% more clients without expanding the support staff. An interesting figure for the channel: the open rate of messages in the Telegram bot was 10-11 times higher than in email newsletters of the same network.
The reason is not in Telegram's magic, but in the channel's design: a bot's message enters a personal chat that a person checks constantly, not a folder «Promotions» that is opened once a week. The same applies to sales: conversion to an order through a messenger bot for ZARINA was 4 times higher than through email, and integrating the bot with a loyalty program added +30% revenue per customer. The conclusion from this case is simple — a bot wins where it solves a narrow task (reminder, notification, fast order), not where it replaces a full consultation.
Metrics and KPI: How to calculate ROI fairly
The report «bot replied to 10 000 messages» says nothing about its efficiency. The key indicators are those related to money and customer retention, not the bot's activity itself.
- The share of inquiries closed by the bot without operator involvement — a benchmark of 60-80% with proper scenario settings; if the indicator is below 40%, the bot works as decoration.
- Conversion to lead with contact details — WhatsApp Business showed 100% conversion from inquiry to lead with proper CRM integration (case AVILON).
- The cost of a lead through the VK channel — after the script was finalized, leads cost 40 rubles and arrived daily for free, without additional advertising budget.
- ROI of database reactivation — with G.M., the bot integration with the loyalty program yielded a ROI of 3488% on reactivation campaigns, which is significantly higher than email channel metrics.
- Increase in purchase frequency and revenue per customer — 'Smile of the Rainbow' achieved +40% in purchase frequency and +25% in revenue per customer precisely due to the bot and loyalty program combination.
The ROI formula for a chat-bot is calculated the same way as for any channel: (revenue attributed to the bot minus development and support costs) divided by costs, multiplied by 100%. The mistake in most calculations is not subtracting the cost of maintaining the scripts and not separating 'bot revenue' from the revenue that the client would have brought without the bot.
Which products are sold through the bot, and which are not
A strategic mistake is to botize everything indiscriminately, guided by the trend of the channel, rather than the nature of the product. The product category determines whether the bot will be a helper or a barrier.
| Suitable for a bot | Does not sell well through a bot |
|---|---|
| Repeat purchase products (cosmetics, household chemicals, children's goods) | Expensive real estate and premium services |
| Tickets, bookings, simple services with fixed pricing | Complex B2B-sales with a long negotiation cycle |
| Loyalty programs and reminders about remaining points | Medical consultations requiring doctor's responsibility |
| Flash sales and promotions with limited time | Products with high emotional engagement (wedding services, exclusives) |
| Notifications about order status, delivery, return | Products where the decision is made only after a live consultation |
General rule: the higher the cost of error for the client and the more unique the decision they make, the less automation should be in the dialogue and more should be quick access to a human. A bot that honestly says 'this issue is better resolved with a manager' and immediately transfers the contact converts better than a bot that tries to close the deal on its own where the client needs assurance, not a script.
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
