How LIQUIDZ swapped 40 weekly hours of spreadsheets for data ready by 7am Monday.

Context

LIQUIDZ is one of the largest beverage e-commerce operations in Brazil. Extensive catalog, multiple sales units, multiple channels: marketplace, direct retail, B2B. At that scale, this means thousands of transactions a day, each with its own SKU, cost, margin, turnover, seasonality. The data is there, in BigQuery. The problem is what gets done with it.

LIQUIDZ's PMO and data team had a rhythm familiar to any e-commerce operation that grew faster than its analytics structure: spreadsheets. Saturday morning, someone would turn on the computer, open several tabs, pull data from multiple sources, copy, paste, check, format, send. Sunday, final adjustments. Monday afternoon, the weekly report was ready. And with it, the previous week's decisions had already passed.

As the operation grows, this becomes an enormous invisible cost. It's not just the team's time consumed: it's delayed decision time. A buyer who needs to decide on restocking, marketing that needs to evaluate a campaign, a store manager who needs to adjust the mix: all waiting on the report from an overloaded team.

The challenge

Thomas, LIQUIDZ's PMO, framed the problem the right way right from the diagnosis: "I don't need a prettier BI. I need data that reaches my inbox exactly when I'm about to make the decision."

That completely changes the scope. It's not about building a dashboard someone has to open. It's about building a data operation that runs on its own, delivers on the right channel, at the right time, with the right level of detail. And that accepts natural-language questions when someone needs to dig deeper.

The solution

NexUnio built a conversational BI architecture integrated with BigQuery, with three delivery fronts:

The first is the conversational agent via WhatsApp. Anyone on the team can send a question, for example "how's SKU X's turnover this week?" or "how is brand Y performing in stores in the South?", and get an answer with a number, a chart, a comparison. The agent was trained on LIQUIDZ's vocabulary: brand names, internal categories, preferred date formats, the metrics that matter for each role. Behind the agent, a multi-agent architecture: an orchestrator identifies intent, specialized sub-agents query BigQuery, validate the result, and return it in natural language.

The second is the automated report suite. Four fixed reports run on their own: a daily report at 7am with sales and alerts; a weekly scorecard on Saturdays at 10am; a weekly marketplace report with take rate and returns; a weekly retail report with category mix.

The third is the alert layer. When some indicator moves outside expectations, such as turnover dropping, margin sliding or a top SKU out of stock, the system alerts the responsible people directly on WhatsApp, with the problem described and context for the decision.

The results

• 40 weekly hours of spreadsheet work became 2 hours of analysis. The team that used to run spreadsheets now interprets results and suggests action. Operational work became strategic work.
• The Monday report arrives ready at 7am, every day. The week's decision is made Monday morning, not Thursday afternoon.
• 100% automated since 2025. No relapse. The system ran through Christmas, Carnival, the September 7th long weekend. The team didn't have to log in.

What this unlocks

Schedule freedom here means business decisions before the problem escalates. When the report arrives Thursday afternoon, that week's problem is already this week's problem too: you'll only be able to act next week. When the report arrives Monday at 7am, last week's problem is solvable Monday afternoon. In e-commerce, where margins are tight and the cycle is fast, that's real money.

LIQUIDZ bought decision time. It also bought operational independence: the report no longer depends on someone being willing to sacrifice their Saturday.