Why Telegram bots are the right first channel in Ethiopia

Nearly every discovery call we join ends with the same question, asked in different words: where do we meet the people who will use this? The question is usually framed as a feature comparison between an app, a website, a WhatsApp integration and a Telegram bot. That framing is wrong, because it treats distribution as a consequence of the build. Distribution is the reason to build.

The reason we usually recommend a Telegram bot first in Ethiopia is simple and unglamorous. Telegram already has a very large installed base here, and the people in it are already spending their data on it. You are not asking anyone to install a thing to reach them.

Telegram is the installed base, not an experiment

Ethiopia is the exception to the WhatsApp everywhere rule that holds across most of Africa. Google Play Store rankings as far back as 2017 put Ethiopia as the only country in the world where Telegram was more widely used than any other messaging app, and that pattern has held rather than reversed. Ethiopia also has a large online population by regional standards, around 29 million internet users, with low overall messaging app adoption, which leaves a lot of room for one app to dominate the connected population.

A caution about measurement, because it affects how you should talk to a client about this. The widely used social media trackers will give you Facebook, Messenger, Instagram and LinkedIn numbers for Ethiopia, and will quietly leave Telegram out. That says more about how these datasets are collected than about how much Telegram is used. Do not put a confident figure on a slide. Put the client on Telegram and count your own users, which is the only number that will be true about your product.

What is not ambiguous is the global position. Telegram has said publicly it is past a billion monthly active users, it is the largest messaging app not owned by Meta or Tencent, and the 18 to 34 bracket is close to half its users. That is a very good demographic match for a country where the median age is in the high teens and low twenties.

Choose the channel your users already spend their limited data on. Everything downstream is easier than you think, because the hardest problem was never the software.

Why it became popular here, and why that still pays

Telegram's dominance here was not an accident of marketing, and the reasons are still the reasons it wins.

It is cheap in bytes. Telegram's Android install has historically been a fraction of the size of its main competitors', and its idle data footprint is small. In a market where mobile data is prepaid, metered and often the largest line item in a household's monthly spending, a lighter app is not a nice detail. It is the product.

Text is the payload. This is the same argument we make in our offline first work, applied to a channel instead of an app. Message text costs a tiny fraction of what an image or a video costs, and Telegram makes text the cheap default. Build the flow so a user who never loads a single image can complete the whole task, and your completion rate will be higher than the one your design mockups predicted.

Channels, not just private chat. This is the part that gets overlooked. Telegram is not only a one-to-one messenger. Groups and broadcast channels go to 200,000 members, which makes Telegram a distribution channel in its own right, not just a conversation transport. A bot that is useful inside the groups it serves is far cheaper than paid acquisition.

Nobody has to install anything to use your bot

This is the underrated advantage. A user can complete an interaction with your bot from a t.me(currently moving to .gram tld) link in a browser or in an old version of the app, on a two year old Android handset, without an app store account, without a review queue and without a large download over a weak connection.

Set against the cost of launching a mobile app in Ethiopia, that comparison is not close. An app launch is a build, a store presence, a download that competes with the user's data budget, and a version compatibility surface you own forever. A bot is a URL and a service. For a first release, this is the difference between reaching people this quarter and reaching people after a six-month engineering cycle.

The bot API will not be your expensive part

The Telegram Bot API is free. There is no per-message charge, no seat licence and no revenue share. Hosting a small bot is a few dollars a month, and the interaction model is small: commands, inline keyboards, webhooks, a media pipeline. A competent engineer can have something in front of real users in days, not sprints.

The platform also handles the parts that quietly consume weeks on a web or app build: identity, notifications, file transfer, delivery state and the rich text rendering. You are not rebuilding those. You are writing a handler and a small state machine.

And Telegram's client is multilingual, so Amharic, Afaan Oromoo and English can coexist in one bot. For a product serving Ethiopian users, that is not a feature to add later. Language selection is a day-one decision and retrofitting it means reworking every string, every button and every prompt.

Two places Telegram will disappoint you

We would rather write this section than take the project and disappoint you later.

1. Bots are not end to end encrypted. Telegram's secret chats are, but bot chats and group chats are encrypted client to server, not end to end. If you are handling medical records, financial detail or anything with a genuine confidentiality obligation, the channel choice is part of your threat model and you should say so out loud to the client before launch.
2. There is no app store review, and no store credibility either. The flip side of shipping without review is that you have no store listing, no star rating and no obvious install icon. Users trust the link in front of them, which means the message that gets them to tap carries the trust the store would otherwise carry. Put effort into that message.

Keep the logic out of the bot

The mistake we see most often is putting the business logic inside the bot handler. It works, it demos beautifully, and it is expensive to change, because the second channel you add then has to be a rewrite rather than an adapter.

Write the actual work as an ordinary service with an HTTP API. Retrieve documents, score an answer, run the eligibility check, compute the quote. Then make the bot a thin conversational layer over that API, and make your web app a second thin layer over the same API. Adding WhatsApp or a web widget later becomes a small project rather than a new product, and it costs you days instead of months.

Measure it honestly while it is live. Daily active users, completion rate per flow, the share of sessions abandoned at a specific step, and median time to a completed task are the four numbers we put in front of a client at the end of a sprint. If a flow has a 40% drop off at one button, that is a design problem, and no amount of distribution will fix it.

The short version

Telegram has a large user base in Ethiopia, and that is the whole argument at the strategic level. Below it there are four concrete reasons it is the right first channel here: a light download and a light data footprint, a text first payload that respects metered data, a free bot API that makes the build cheap, and a URL based entry point that lets people use it without installing anything.

None of that makes it the only channel. It makes it the one that gets you in front of real users soonest, which is the only advantage that matters while you are still learning what your users actually want.

If you want the reasoning applied to a specific product, we are happy to walk through it. That is the conversation we would rather have in the first week than in the third month.