How MADE BY HUMANS made 7.7 million likes measurable – and automated their monthly reporting

The client

MADE BY HUMANS does something that seems small at first glance but has a huge impact: the agency writes comments on behalf of its clients, not on their own channels, but under viral posts from others. Right where hundreds of thousands of people are reading along.

‍

A well-placed comment under a video with a million views brings a brand more visibility than a post of its own that only existing followers see. It doesn't feel like an ad, but like a person saying something witty. When it works, it collects likes itself, climbs to the top, and ends up as the most-liked comment above thousands of others.

The team around founder Thomas Neander manages brands ranging from grocery retailers and audio manufacturers to furniture stores. The copywriters write daily across both platforms, and the best comments rack up five-digit like counts.

‍

The business model stands or falls on proving this impact. A client paying a four-figure monthly fee wants to see what they're getting out of it.

‍

The problem: Finding your own work

This was where the difficulty lay, and it’s a peculiar one: a comment under someone else's post doesn't belong to the agency. There is no dashboard where you can check how many likes your comments have collected. Instagram and TikTok don't show the comment author any statistics. You have to go back to every single post, search the comment section, find your own comment, and note down the number.

‍

That’s exactly how it went. Every month, someone from the team would open several hundred links, scroll through comment sections, search for their own post, and note down likes and replies in a Google Sheet. Brand by brand, line by line. These sheets were then used to create the monthly report for the client.

Three things made this painful:

‍

It didn't scale. With every new brand, the manual labor grew linearly. With a growing client base and several hundred comments per brand each month, it was foreseeable when the point would be reached where reporting would eat up more time than the actual work.

‍

The numbers were instantly outdated. A comment continues to collect likes, even weeks after it was written. What was in the sheet was the value from the day it was checked. A week later, it was no longer accurate, and no one had the time to start all over again.

‍

It was prone to errors. If you scan through three hundred comment sections, you're bound to miss some. And if you type numbers in by hand, you're bound to make a typo. Neither was noticed because there was no second source to cross-check against.

On top of that, there was a structural disadvantage: because the data collection was so time-consuming, it only happened once a month. Between reports, nobody knew how things were going—not the agency, and not the client. A comment that went viral would go unnoticed for three weeks. The same went for a month that started off poorly.

‍

The solution: A comment reporting system that fills itself

bakedwith built a system that handles the entire process, from capturing the link to the finished client report. It runs daily, without anyone needing to trigger it.

‍

Step 1: Collect links where they happen

The copywriter writes their comment and shares the link to the post in a Telegram group, one for each brand. That’s all they have to do. A bot monitors the group, assigns the link to the correct brand based on the group, and saves it.

It sounds simple, but it’s the crucial point: the data is captured exactly where the work is already happening. No one has to open a second tool or update anything at the end of the day. A link that isn't captured doesn't exist for the reporting. So, the barrier had to be zero.

‍

Step 2: Check how the comment is doing every day

Once a night, an automated process runs for all active brands. For every post, the system retrieves the comments and searches for the brand's comment, identified by the account it was written from.

If it finds it, it records the likes, replies, whether the creator responded, and the comment's ranking among all others. If it doesn't find it, it performs a deeper search.

What’s important here is what the system does not store: no third-party comments, no third-party profiles, and no data from uninvolved users. Only the brand's own comment is saved. Everything else is read during the search and then discarded.

‍

Step 3: History instead of a snapshot

Each run creates a new snapshot instead of overwriting the old one. This creates something that didn't exist before: a history. You can see not only that a comment has 17,000 likes today, but also that it had 3,000 two weeks ago.

This answers questions that couldn't be asked in a spreadsheet: How long does a comment actually keep gaining likes? Which posts have long-term impact, and which ones are dead after three days? How does a brand develop over the months?

‍

Step 4: The report the client opens themselves

This data powers a dashboard where every client logs in to see their own numbers exclusively. No more PDFs manually compiled and sent out every month—just a page that’s always up to date.

It shows likes over a selected period, monthly trends, the breakdown between Instagram and TikTok, the top-performing comment, and placement distribution: how often a brand’s comment ranked first among all comments on a post. Across all brands, that’s now several hundred comments.

Internally, the team sees more: a ranking of all copywriters, showing who wrote which comments and how they performed.

‍

Step 5: Closing the gaps technology can't

Some comments can't be tracked by tools. Maybe they were deleted, the post is no longer public, or it went so viral that the brand's comment is buried too deep. For these cases, there’s a workspace view: a list of all posts, filtered by "open," with fields to fill in the blanks. Open the post, check the likes, and enter them.

Manually entered values flow into the report just like the automated ones and are clearly marked as such. This accounts for about two out of every hundred comments.

‍

Step 6: Bringing the past along

A report that starts from scratch won't impress anyone. That’s why the old Google Sheets were imported: the entire history dating back to October 2024, month by month, including the values that were counted by hand back then.

From day one, the client doesn't see an empty history, but the entire journey of our collaboration so far.

‍

The results: From manual work to a running system

Millions of likes have been captured and verified, across tens of thousands of comments. Plus tens of thousands of replies to those comments and thousands of instances where the creator of the post responded themselves. This is the strongest proof that a comment has resonated with its community.

‍

The system finds almost all comments on its own. Only about two out of every hundred need to be added manually. The rest runs automatically.

‍

The report is always up to date. Instead of updating once a month, the numbers refresh daily. The client can check in at any time, and the agency notices immediately when a comment goes viral.

‍

Adding new brands takes almost no effort. Create the brand, assign the Telegram group, add the handles, and it’s in the system from the next run onwards. The effort per brand is no longer linear; it’s practically zero.

‍

The history is new. The imported history combined with daily snapshots provides something that was never possible in a spreadsheet: the answer to whether things are improving. For all brands, the data series goes back to October 2024.

‍

The copywriter view creates comparability. Because every comment is assigned to its author, the team can see which approaches work. Not based on gut feeling, but on likes.

‍

And perhaps the most important effect is hard to put into a number: reporting is no longer a chore. It’s not a month-end deadline, not an afternoon someone has to block off. It just happens.

‍

What we’ve learned

Data entry needs to happen where the work is done. The Telegram bot is technically the simplest part of the entire system, yet it’s practically the most important. If we had built our own form, half the links would never have been recorded. The copywriter shares a link, and that’s something they’re doing anyway.

‍

Append-only beats update. Writing a new snapshot for every run instead of overwriting the old one costs storage space but gives you the complete history. Storage is cheap. You can never get back lost history.

‍

A date is not just a date. The day a comment was written, the day its link was recorded, and the day the system checked it are three different things. If you mix them up, you end up pushing work into the wrong month and breaking your report. Keeping these separate sounds like a minor detail, but it’s the foundation of everything.

‍

Manual work must be allowed to stay. No automated system catches one hundred percent. A report that can’t accommodate the remaining two percent is incomplete, and incompleteness costs you trust. The ability to add entries manually isn’t an admission of failure; it’s part of the solution.

‍

History is a feature. Importing the old sheets was the most unspectacular part of the project, yet it had the biggest impact on adoption. A dashboard that starts from zero feels like a prototype. One with years of history feels like a tool.

‍

Only save what you really need. It would have been technically easier to keep all the retrieved comments. Choosing to save only our own and discarding the rest was a conscious decision. It makes the question of data privacy one that can be answered in a single sentence.

‍

The client should be able to check for themselves. A monthly PDF is a chore someone has to stay on top of. A login is a resource that’s always available. It might seem like a small difference, but it changes the whole dynamic: clients don't have to ask for their numbers anymore—they already have them.

‍

Are you still gathering your results manually?

Are you putting in the work, only for the impact to happen outside your own systems—on third-party platforms, in external channels, or under someone else's posts? And does analyzing that work eat up days every month that should be spent on the actual work itself?

‍

We build you a reporting system that fills itself: data collection right where you’re already working, daily updates without any extra effort, client access instead of monthly PDFs, and the flexibility to step in manually whenever technology hits its limits.

‍

Let’s have a no-strings-attached chat about your specific process.

‍

‍