Incorta Analyst Studio: From Spreadsheet Chaos to Clean Data

Analyst Studio is our new visual workflow builder in Incorta: and I think it's going to change how a lot of teams build their analytics. On a recent episode of The Weekly Cup of Joe, I walked through a live demo.

AI writes a lot of our code these days, and it does a great job. But I keep noticing that the people who own the numbers end up out of the loop. The logic lives somewhere in a script nobody has read.

I think there's still a strong place for logic you can actually see: a simple flow, almost like a DAG diagram, that people can vet, verify, and test against. That's the idea behind Analyst Studio.

If you're an existing Incorta customer, you may know our Data Studio - but don't just think of Analyst Studio as just an improvement - it's a replatformed experience, with a lot of lessons learned went into it.

What is Analyst Studio?

At its core, Analyst Studio is a visual builder for analytical and operational workflows, and you don't have to be a code expert to use it. On the left you get a library of drag-and-drop recipes, each one an analytical step: join, filter, select, calculate, aggregate. You drop them on the canvas and string them together, in one process or in parallel branches.

Every step is observable. Click anything on the canvas and the panels change with it: you see the data at that point, what the join is doing, which records a filter keeps and which it drops. We sample the data across the whole process, so nothing is hidden.

One flow can also feed many outputs, which hand-written code rarely does well. Take the classic example of joining transactions to loyalty rewards. From that one join you might want three answers:

• Which loyalty members aren't spending?
• How much are loyalty members spending?
• Who's spending without a loyalty account, and how do we sign them up?

Same logic, three outputs, and no copying the join three times.

The demo: diagnosing monthly revenue at Summit Trail

For the demo I used a fabricated dataset for a retailer I call Summit Trail. Think of it as a point-of-sale system tracking transactions across all of my stores. My core question was simple: how is revenue performing every month? And when there's a dramatic change, what can I attribute it to? Not enough labor on shift, less traffic, deeper discounts, or people shifting what they buy?

Here's how the flow gets there:

1. Four inputs. Sales lines (every item on a receipt, with list price, discount, and extended amount), receipts, daily labor, and store details.
2. Join sales to receipts on receipt ID to pick up store information. The join panel lets me choose which columns carry forward, rename them, and change data types, with no SQL.
3. Filter to completed receipts. We had trainees practicing on the registers, plus some voided receipts. Both get filtered out, so I'm looking at real revenue only.
4. Select and rename. Extended amount isn't the friendliest name, so it becomes net amount; extended cost becomes net cost and quantity becomes net units.
5. Line-level calculations. Year-month, a period index, fiscal quarter, discount and return dollars, and a flag for loyalty receipts.
6. Aggregate by store and period: net amount, cost, units sold, receipt totals, and loyalty receipts for each store in each period.

One note: I did the sales-to-receipts blend inside the flow for illustration. You don't have to. You could build that blend as an analyzer table first and bring it into Analyst Studio instead. Totally fair game.

From "revenue dropped" to "here's why"

Next I bring in labor. I have tables that tell me, for each store and date, the hours planned versus worked and the headcount on shift. I pre-aggregate those to the same store-and-period grain before the join, which keeps it clean.

Then I add store data. There's a lot in there (cost centers, states, cities) and I drop most of it, but square footage stays. That lets me ask what revenue per square foot looks like, or how efficient each store really is.

The key step is the prior period index. I join the dataset to itself so each store's current month sits beside the month before. From there it's easy to calculate the change in traffic, labor hours, average price, and basket size, and split the revenue change into its parts:

• Traffic effect: did fewer people come into the store?
• Basket effect: did shoppers put fewer items in their baskets?
• Price effect: were the items they bought cheaper on average?

In my example all three were down, but traffic did the most damage. I don't want to check those columns by eye every month, so one formula names the primary driver of the change: traffic, price, basket size, or labor. Another column suggests what to do about it.

Fast on real volumes, and schedulable

While you build, the canvas works on a live sample of about 1,000 records, so every step stays responsive. The sales line table behind it holds about 2 million records.

When I kicked off a full run, it finished in 29 seconds across those 2 million sales lines and 608,000 receipts. The output is one flattened diagnostic table, and I opened it in an exploration to pull store, period, net revenue, net cost, primary driver, and recommended action side by side. That's really the point: Analyst Studio curates the dataset that dashboards, AI interactions, and applications consume.

And yes, I know the next question is always "Can I schedule it?" Absolutely. Create a schedule and the same diagnostic lands in your dataset every month.

AI built in: an assistant that explains, and an agent that acts

I almost forgot to show you the best part. Every flow comes with a Workflow Assistant.

I played a teammate who'd just been handed my flow with "I'm switching teams, this is yours." My first question was simply "What does this do?" The assistant walked through the inputs and each container of steps: the point-of-sale lines, the labor calculations, and the KPIs coming out. I asked it to explain how the diagnostic formula is derived, and to write a formula for calculating the primary driver. It answered in plain language, with a formula ready to plug in.

But wait, as my friend Billy Mays would say, there's more. In agent mode, I don't just ask questions; I ask it to act. I asked it to keep only stores whose revenue changed by more than 20%. With auto-apply on, it picked a spot in the flow and added a filter on the absolute value of the revenue change. It landed in a slightly odd place, honestly, but that's the point: the new step sits on the canvas like any other recipe, so I can open it and check exactly what the AI did. Turn auto-apply off if you'd rather review changes first.

Two kinds of flows, and how to get it

The flows I see tend to fall into two buckets:

• Analytical: describing and explaining your data, like the revenue diagnostic above.
• Operational: repeatable pipelines, such as consolidating financials every month into a prescriptive report.

Either way, you get a fully interactive canvas for people who want to see, feel, and audit their data at every step, without a black-box back end, while still producing richer outputs.

Availability. Analyst Studio is in preview in Incorta 8.2, with general availability planned for 8.3. It's expected to be part of the standard Incorta enterprise offering, with no separate SKU, though we're still finalizing go-to-market details. Reach out and we'll get you set up.

What's next. In the next session I'll connect Analyst Studio to applications: parameterizing these workflows so users get a front end, with a fully observable engine under the hood.


Watch more Weekly Cup of Joe here.