AI solution · Analytics · Decision models
Same inputs. Same answer. Every time.
We build analytics on deterministic models, so the numbers your team decides on can be reproduced, traced to their source and defended in the room.
Free · 30 minutes · with a founder
The track record behind 10dem
Engagements delivered by our founding team and partner practice over 10+ years of consulting.
clients served
geographies
sectors across India
AI tools tested on live work
The problem
01 / 06
Three teams, three numbers, one decision.
Marketing, finance and product each arrive with a different figure for the same metric, and the meeting turns into an argument about whose spreadsheet is right. Asking a chatbot adds a fourth number that changes when you ask again. We replace all of that with one defined metric, one model and one answer that anyone can check.
What deterministic means
02 / 06
Four properties every model we ship has.
Reproducible
Same inputs, same answer, today and next quarter. A number that changes when you ask twice is not a number.
Traceable
Every figure links back to its formula and the source rows behind it. Anyone can follow the chain.
Tested
Model logic is checked on every run, so a broken input fails loudly instead of quietly shifting a forecast.
Explainable
The logic is written down in terms an operator can read. Language models help narrate a result. They do not compute it.
What we build
03 / 06
Start from the decision. Build the model it needs.
We do not start with a dashboard. We start with a decision your team makes again and again, define the metric behind it, and build the smallest model that answers it reliably.
mahalanobis.siModels we build most
- Demand and revenue forecasting
- Pricing and discount rules
- Unit economics and contribution margin
- Marketing attribution and budget allocation
- Cohort retention and lifetime value
- Inventory and capacity planning
Proof
04 / 06
Decisions we have already put on a model.
Food manufacturing
25%
Improvement in forecast accuracy with the data-driven demand model
Demand planning that runs ahead of demand
A food & baking company
Read the caseSaaS & finance
1-hour close
Down from 2–3 days, with ETL automation
Month-end close: 2–3 days to 1 hour
4 SaaS companies, $2M combined MRR
Read the caseRetail
45,000
SKUs classified for inventory control
45,000 SKUs classified. Inventory stopped eating cash.
A premium multi-outlet retail business
Read the caseReal engagements delivered by our founding team and partner practice. Clients are described by profile, not named.
How it runs
05 / 06
Three steps. You can stop after any of them.
01
Free audit
30 minutes with a founder. Bring one metric or one recurring decision and we show you where it breaks.
02
Fixed-scope plan
The model, its inputs, the tests it must pass and the milestones, with the price agreed before any work starts.
03
Build, test, hand over
We ship the model with its tests and documentation, and hand over code your team can rerun without us.
FAQ
06 / 06
Frequently asked
- What is a deterministic model?
- A model where the same inputs always produce the same output, through logic you can read and test: defined metrics, explicit formulas, rules and established statistical methods. Nothing is sampled or improvised at the moment you ask, so the result can be reproduced and audited.
- Why not just ask an AI model for the answer?
- A language model generates plausible text, and it can give different figures for the same question on different days. That is acceptable for a draft and unacceptable for a price change, a budget or a forecast. We use language models to explain a result in plain words, and deterministic logic to produce it.
- What kinds of decisions do you model?
- The recurring ones that move money: demand and revenue forecasts, pricing and discounting, unit economics, channel attribution and budget allocation, cohort retention and lifetime value, and inventory or capacity planning. We start from the decision and work back to the model it needs.
- Do we need a data warehouse first?
- No. We start with the data you have and one decision worth getting right. If that work shows a warehouse or a cleaner pipeline is needed, we scope it as its own milestone instead of making it a precondition.
- What do we own at the end?
- Everything: the metric definitions, the model code, the tests, the documentation and the dashboards. Your team can rerun any number and see how it was computed, without calling us.
Have a number nobody fully trusts?
Book a free 30-minute audit. Bring one metric or one recurring decision and we will show you where it breaks and what a reliable model for it looks like.
Book an analytics audit →