A supermarket can lose money in two opposite ways. Order too many tomatoes and watch them soften on the shelf. Order too few and watch the customer leave. The uncomfortable part is that both mistakes can follow perfectly reasonable judgments. Rain arrives. A promotion works. A familiar buying pattern disappears. Somewhere between yesterday’s sales and tomorrow’s delivery, a person has to choose a number. Ganit has built a business around making that choice more informed - and making sure the answer gets used.
- Ganit combines data engineering, AI and business consulting.
- Its tools address demand, delivery routes, promotions, customer sentiment and store locations.
- More Retail is a named customer; published cases connect forecasting to actual ordering.
- The recurring theme is adoption: analysis needs someone willing and able to act.
The founders’ useful obsession
Ganit began in 2017 with three founders: Shivaprasad KT, Ashok Harwani and Hariharan R. They were batchmates before they became business partners. Their leadership page traces the reunion to a shared enthusiasm for mathematics. Today, Shivaprasad is CEO, Harwani heads solutions and Hariharan leads strategic partnerships.
The company’s central preoccupation is less abstract than its founders’ favourite subject. An enterprise can accumulate data, employ analysts and buy software while leaving the resulting analysis largely untouched. Ganit’s response is to work backwards from consumption: the point at which somebody uses an insight. Shivaprasad has discussed this as Consumption Quotient, or CQ. For a buyer, the useful question is immediate: which decision will this project change?
“For us, consumption comes before technique”
Ganit’s stated approach
A tomato has a deadline
More Retail offers a concrete answer. In a joint account published by AWS in March 2021, the retailer and Ganit describe replacing forecasting and ordering bottlenecks with an integrated system. The reported results included forecast accuracy rising from 24% to 76%, fresh-produce waste falling by up to 30%, and gross profit increasing 25%. These were outcomes from that deployment, rather than promises for every customer.
+52 percentage points · Published March 2021 · Case-specific results
The account is revealing about the effort involved: more than 200 experiments, learning from unsuccessful models, and a planned adoption process for category owners accustomed to their own judgment. The team had to establish trust and keep stock records timely. A forecast could improve while the purchasing routine stayed stubbornly familiar. Changing the routine was part of the assignment.
Ganit’s separate, anonymous produce case adds another practical detail. Business teams and store managers helped formulate 82 hypotheses, including assumptions about rainfall and seasonal availability. Its solution joined forecasting, an order calculation and a scheduled data workflow. That sequence matters. Expected demand and the quantity to purchase are different numbers once existing stock, waste and display requirements enter the conversation.
Five names, five decisions
Ganit packages several recurring problems into products. Augury forecasts demand at different planning intervals. Beeline helps schedule and optimize delivery routes, accommodating such realities as truck capacity and delivery windows. Lift examines promotional campaigns and their sales effects. Slant gathers and analyzes customer sentiment. NSO uses geographic analysis to help retail chains evaluate new store locations.
Think of them as five questions: How much will sell? Which route should the truck take? Did the promotion help? What are customers saying? Where should the next shop go? The names matter less than the specificity. A transport team and a marketing team need different interfaces, constraints and definitions of success, even when both depend on data science.
Behind those tools sits the less visible engineering: moving data, modernizing platforms, setting governance and connecting systems. Ganit describes work on AWS, Snowflake and Databricks. Its Adobe practice reaches into commerce, content management, analytics and marketing automation, where analysis meets the customer-facing experience.
The economics live in the workflow
The commercial offering combines consulting, implementation and productized services for enterprises. Ganit describes customers across retail, consumer goods, financial services, manufacturing, pharmaceuticals and healthcare. Its international reach makes it part of the specialist analytics consulting market, alongside buyers’ options to build internally or use narrower software products.
In July 2022, Ganit announced an undisclosed pre-Series A investment from entrepreneurs including Addverb’s Sangeet Kumar, Kissht’s Krishnan Vishwanathan and EatClub founders Anshul Gupta and Amit Raj. The stated expansion plans included forecasting, promotion planning, routing and customer feedback tools. That list is a clue to the business: repeatable operational problems provide room for products, while each customer’s systems and working habits create the implementation work.
Its published work makes the cost argument tangible. In a June 2023 AWS apparel case, Ganit reported roughly 15% lower procurement cost per invoice for fast-fashion products. That figure describes the customer’s purchasing outcome. It should not be confused with the price of installing the system or the cost of running it.
An Indonesian promotion project shows the implementation logic. Ganit says the optimizer initially ran alongside existing processes, letting teams compare performance and become familiar with it before broader use. The company reported returns improving by up to 10%. The approach gives a buyer something sensible to copy: test the proposed decision against the current one before handing over the workflow.
The people have to come with it
Ganit’s careers page emphasizes peer learning, exposure to clients early on, and freedom paired with responsibility. Those are company-described ambitions, but they fit the work. Someone must understand a business problem, translate it into a system and explain the result to the person whose day changes.
There are practical limits. A retailer with unreliable stock counts cannot expect a more sophisticated forecast to fix every order. A routing tool needs credible delivery constraints. A promotion model needs a baseline worth comparing against. These are lessons from how the projects work: the surrounding process gives the model something useful to do.
For a reader considering Ganit, start with one repeated, costly decision. Name its owner. Record the current outcome. Ask how a proposed solution enters the working day, and how improvements will be measured. Tomorrow’s tomatoes make a fairly demanding test. They do not wait for the presentation.