Data to AI™ / Gemini/TikTok
Diff lifetime video snapshots, plot retention curves and fix follower logic in your code, with TikTok data read from Dataddo inside the terminal or VS Code.
Better answers, fewer tokens.
Gemini CLI
How did our TikTok followers change week by week in September, and when did we lose most?
The account gained 6,210 net followers over four weeks. Losses peaked in the week of 8 Sep at 1,150, and 22-28 Sep was the strongest week with 2,500 net.
| Week | Daily New Followers | Daily Lost Followers | Daily Total Followers |
|---|---|---|---|
| 1-7 Sep Profile Daily Insights | 1,840 | 420 | 1,420 |
| 8-14 Sep Profile Daily Insights | 2,310 | 1,150 | 1,160 |
| 15-21 Sep Profile Daily Insights | 1,520 | 390 | 1,130 |
| 22-28 Sep Profile Daily Insights | 2,960 | 460 | 2,500 |
Daily Total Followers is the daily net change in followers, not the follower count, so it is summed per week.
TikTok Organic extracts non-paid insights from a TikTok business account: daily profile activity, follower audience breakdowns, and per-video reach, watch time, retention, traffic sources and comments.
Compare the two latest dataddo_extraction_timestamp snapshots of Video Lifetime Insights and write the Video Views and Shares gained per video to reports/tiktok-delta.csv.
Video numbers are lifetime totals, so a data engineer appends each run and lets Gemini CLI compute the difference. The content team gets a weekly file of what each video gained, not a pile of cumulative exports.
Load Second and Percentage from Video Lifetime View Retention for our last 15 videos and plot every curve on one chart in notebooks/retention.ipynb.
In VS Code, Gemini Code Assist writes the notebook cell with the real field names. The creative lead sees at a glance which openings lose viewers early and which hold them to the end.
Our loader in src/tiktok/followers.ts treats Daily Total Followers as a follower count. Query a week of Profile Daily Insights and correct it to sum the daily net change.
A developer catches a misread metric before it reaches the dashboard. Gemini describes the field from the semantic layer, so the fix follows what the data means rather than what the name suggests.
Data to AI™ is priced per seat, like the AI assistants your team already uses, with ~25% off on annual billing. Need tailored contracts or volume pricing? Talk to us about Enterprise.
Serve governed business data to Claude, ChatGPT and other AI tools over MCP.
$10.49/month
per seat
3 seats minimum
Flexibly deploy Dataddo in any cloud or hybrid environment.
Custom
We have a payment model that works for you
| Without Dataddo |
With Dataddo
|
Outcome for you | |
|---|---|---|---|
| Change between exports | Lifetime exports are saved in dated folders and someone subtracts last week's totals from this week's by hand. | Appended runs carry dataddo_extraction_timestamp, so a script written with Gemini CLI subtracts one snapshot from the next. | Weekly gains per video |
| Reading follower fields | Daily Total Followers is charted as an audience size, and the dashboard shows a tiny, flat account. | Gemini reads field descriptions from the semantic layer before writing code, so the daily net change is summed, not plotted as a count. | Follower charts that add up |
| New brand profiles | A pipeline fails on an empty audience file and nobody knows whether the export or the account is the problem. | Profile Lifetime Audience datasets describe the 100 follower minimum, so Gemini can explain an empty result and the script can handle it. | Empty data explained |
| Token consumption | Gemini spends tokens exploring raw TikTok columns, sampling rows and retrying until it understands the data. | Field definitions and relations are supplied up front, so the model skips the exploration and goes straight to the answer. | Fewer tokens per answer |
| Answer quality | Prone to wrong joins, invented fields and numbers that are hard to check. | Answers grounded in defined fields. In the Dataddo benchmark, 78.5% of questions were answered correctly, against 65.5% with plain CSV files. | Better answers |
The TikTok datasets marketing teams use most, with their real field names. Gemini queries them by name through the Dataddo semantic layer.
| Dataset | Metrics and attributes |
Ask Gemini
|
|---|---|---|
| Profile Daily Insights |
|
Export Date, Video Views and Profile Views for the last 60 days to a CSV for a notebook chart. |
| Video Lifetime Insights |
|
Which videos had more Shares than New Followers, with their Caption, as a markdown table? |
| Video Lifetime View Retention |
|
Write Percentage by Second for our ten newest Video ID values to a CSV so I can plot retention curves. |
| Video Lifetime Traffic Sources |
|
Return the Percentage per Impression Source for every Video ID as JSON for our content dashboard. |
| Video Lifetime Audience by Type |
|
Export Video ID, Type and Percentage to a CSV so a script can flag videos with a low share of followers. |
| Profile Lifetime Audience by Country |
|
List every Country above two percent per Username as a markdown table for the market review. |
| Video Comments Lifetime Insights |
|
Which comments ask about price or availability in their Text, and how many Likes did each get? |
We asked 43 pre-registered questions of three real datasets - Google Search Console, HubSpot deals and Google Ads - and delivered the same rows to the same model in four ways. Only the data layer changed.
of questions answered correctly - the best of the four delivery methods tested
fewer confidently wrong answers - 14.0% of runs against 24.0% and 26.3%
more correct answers on messy CRM data - better on 8 questions, worse on none
Same model, prompts and rows in every condition; 759 graded runs, ground truth frozen before any run. On Google Search Console alone, documented CSV files did as well as Data to AI™ - the advantage comes from messier data such as HubSpot deals.
Read the benchmark →Create a TikTok Organic source in Dataddo with Profile Daily Insights and Video Lifetime Insights and attach it to a Google Gemini destination or AI Model. Then add dataddo to mcpServers in ~/.gemini/settings.json using httpUrl https://headless.dataddo.com/mcp-data, restart Gemini CLI and sign in with /mcp auth dataddo.
Only Profile Daily Insights reads a date range. Video datasets are lifetime snapshots, so append them with dataddo_extraction_timestamp and ask Gemini to compare snapshots for day-to-day or week-to-week change.
No. It is the daily net change in followers, so sum it over a period to get growth. Daily New Followers and Daily Lost Followers sit next to it in Profile Daily Insights.
Profile audience breakdowns by age, gender, city and country exist only for accounts with at least 100 followers. Below that the datasets stay empty, whichever client asks.
Gemini CLI asks you to confirm each tool call, and you can choose Always allow per tool, for example for query and describe_entity, while still confirming others such as refresh.
Yes, with a separate entry. Gemini Code Assist in IntelliJ reads its own mcp.json, so add the Dataddo server there. Agent mode in VS Code uses the same settings.json as Gemini CLI.
No. The Dataddo tools only read TikTok data already extracted into your flows through the semantic layer. Nothing is written back to TikTok, and Dataddo never accepts SQL from the model.
