Tutorial: launch a product on a marketplace with Modelyx
In this tutorial you use Modelyx to prepare a thermal mug launch on Wildberries and Ozon, from niche research to product cards, in one chat. You finish with your own agent that checks prices every Monday and posts a summary to Telegram.
What you'll build
- A niche analysis of thermal mugs on Wildberries and Ozon, as an Excel file and an interactive chart.
- A table of competitor pain points and a list of arguments for your card.
- A launch price recommendation based on the prices you collected.
- Five product card images in 3:4 and a launch post.
- A "Marketplace analyst" agent that checks prices every Monday at 9:00 and posts a summary to your Telegram channel or chat.
Before you start
- Sign in to Modelyx. New accounts may join through a waitlist first.
- Have a product photo ready. A sharp, well-lit photo on a plain background gives the closest cards.
- Make sure you have credits. Data collection, image generation and Cloud Sandbox runs cost credits and need a positive balance; see Credits and usage.
- For step 8, have a Telegram channel or chat where you can add a bot.
Open Home and keep the defaults: Agent mode and Auto model selection. Do steps 1 to 5 in the same chat, so the agent can reuse the data and files from earlier steps. See Modes: Agent, Chat and Plan.
Step 1. Analyze the niche on Wildberries and Ozon
Type:
I want to launch a 450 ml steel thermal mug on Wildberries and Ozon. Collect the top 100 listings for "термокружка" on each marketplace. Split them into price segments, show the top 10 brands by number of listings, and use review counts as a demand signal. Give me an Excel file and an interactive chart of listings and reviews by price segment.
What you'll see: Modelyx OS may ask one short question first, such as which region to use. The timeline then shows the data collection run and "Sandbox is starting" when the agent begins calculating.
What you get: an Excel file with a download link, saved to Resources, and a chart that opens as an artifact in the side panel.
Modelyx collects public marketplace data live and calculates in the Cloud Sandbox, so segments and medians come from the items that actually came back. See Analyze a marketplace niche before you launch and Web research and live data.
Step 2. Collect reviews on the top competitors
Type:
Take the 5 mugs with the most reviews from that table and collect their latest reviews on both marketplaces. Group the complaints into pain points with counts and one example quote each. Then turn the top 5 pain points into arguments for my product card.
What you'll see: a second collection run, then grouping and counting.
What you get: a pain points table and a short list of card arguments, such as "lid doesn't leak" or "keeps heat for hours".
Competitors' weak points tell you what to promise on your card. Use only arguments that are true for your mug. See Find product weaknesses in customer reviews.
Step 3. Decide the price
Type:
Using the prices you collected, show the median price on each marketplace and the price range of the segment with the most reviews. My cost is 420 rubles per mug. Recommend a launch price and show where it sits against competitors.
What you'll see: the agent calculates on the data already in this chat.
What you get: median prices per marketplace, your position in the range and a recommended launch price with the reasoning.
The Cloud Sandbox keeps files between messages, so follow-ups don't collect the data again. The final price is your decision. See Cloud Sandbox: code, calculations and large data.
Step 4. Create product card images
Attach your photo with + → Attachments and type:
Here is a photo of my mug. Make 5 product card images for Wildberries and Ozon in 3:4: a main card with the product name, three cards with the top arguments from the reviews, and a specs card. Text in Russian. Show me the text for every card first, and use exact text.
What you'll see: the agent proposes the text for each card and waits for you to confirm or edit it.
What you get: five 3:4 cards that keep your product from the photo and share one style.
Modelyx sets your confirmed text on each card exactly as you wrote it, which avoids typos. Still proofread each card and compare the product with the real item. See Create product card images for Wildberries and Ozon and Generate and edit images.
Step 5. Announce the launch
Type one of these:
Make a 5-slide Instagram carousel in 4:5 announcing the mug launch, in the same style as the cards. Slide 1 is a hook, slides 2-4 are benefits, slide 5 says "Now on Wildberries and Ozon". Add a caption.
Write a short Telegram post announcing the launch, with a 4:5 cover image in the same style as the cards.
What you'll see: the slide or post text for your approval, then the images.
What you get: slides or a post with a cover, ready to download.
Modelyx does not publish to Instagram, so you post the carousel yourself. See Create an Instagram carousel.
Step 6. Build a "Marketplace analyst" agent
On Home, click Create agent. Agent Builder asks: "Tell me your use case." Type:
Build an agent called "Marketplace analyst". I sell a 450 ml steel thermal mug on Wildberries and Ozon. Each time, collect the top 50 listings for "термокружка" on both marketplaces, report min, median and max price per marketplace, show where my price sits, list the 5 biggest changes since the last report, and deliver an Excel file plus a 5-line summary.
What you'll see: Agent Builder fills in the name, description and instructions.
What you get: a new agent in your Agents list. Open its Chat page and run the request once to test it.
An agent keeps fixed instructions, so you don't paste the same request every week. See Build an agent.
Step 7. Schedule weekly price monitoring
Open the agent's Tasks page and create a task titled "Weekly mug prices" with this instruction:
Run the weekly price check for my thermal mug on Wildberries and Ozon and deliver the Excel file and the 5-line summary.
In Automation, choose Scheduled → Weekly, Monday, 9:00 and your timezone. Click Run now once to check the output.
What you'll see: the task on the board and its Next run time.
What you get: a report in the task's artifacts every Monday.
The schedule runs the agent without you. To change the report, leave a comment on the task and the next run follows it. See Tasks and schedules and Monitor competitor prices on Wildberries and Ozon.
Step 8. Post the weekly summary to Telegram
- In Telegram, create a bot with @BotFather and copy the bot token.
- On the agent's Channels page, choose Telegram and paste the token into Bot Token. Set DM policy so only you can message the bot.
- To post to a channel or group, add the bot there as an admin. Set Group policy to Allowlist and add the chat ID under Allowed Channels.
- Add one line to the task instruction:
After the report, post the 5-line summary to my Telegram channel "Mug launch".
What you'll see: after the next run, the summary appears in your channel.
See Channels: bring an agent to Telegram and Run a Telegram channel on autopilot.
What you learned
- Live data and the Cloud Sandbox (steps 1–3): Modelyx collects public marketplace data and calculates on it with code, in one chat that keeps your files.
- Images (steps 4–5): generated visuals keep your product from a photo, and your text appears exactly as you wrote it.
- Artifacts and files (steps 1–5): charts open as artifacts, and exported files land in Resources.
- Agents and schedules (steps 6–7): an agent turns a one-off request into repeatable work.
- Channels (step 8): the agent delivers results to Telegram.
Next steps
- Share the chart: Publish the artifact to get a public link. See Results: artifacts, pages and files.
- Turn the cards into a short video. See Make a short product video.
- Plan a month of posts for the launch. See Plan a month of content and produce the posts.
- Learn how to steer the agent. See Control what the agent does.