Outcome Contracts

Five ways to buy. All of them published.

Every tier has a floor so we can staff you properly, a cap so your CFO can budget, and a baseline locked at signature so nobody argues about attribution six months in. All figures exclude GST.

Tier 0 · Growth Audit — the front door
₹75,000 ≈ $790 · 2 hours · 100% credited

Two hours with your leadership, and a written baseline at the end of it. Credited in full against any engagement signed within 30 days.

Value Baseline Memo — your current-state numbers, agreed and signedIncluded
Three costed use cases, ranked by paybackIncluded
Fixed price, paid upfront, no scoping call required

Extended version — two weeks, five stakeholder interviews, prioritised roadmap: ₹2,50,000 (≈$2,630). We price the audit because free diagnostics select for people who will never buy.

Book the audit
Tier 1 · Sprint + Kicker — default first engagement
₹8L – ₹15L build · then ₹1.5L – ₹3L / month

A fixed build, a monthly run fee, and an upside kicker when the agreed KPI lands. No downside for you.

Build & deploy sprint (6–8 weeks)₹8L – ₹15L
Monthly run fee₹1.5L – ₹3L
Outcome kicker, paid quarterly10–15% of base
Kicker cap50% of base
Tier 2 · Baseline + Bonus — fee genuinely at risk
₹1.89L – ₹3.15L / month

For clients with clean data and a number somebody owns. We take a lower base and put part of it at risk against the target.

Base fee (70% of the fixed equivalent)₹2,10,000
At risk if KPI < 50% of target−10% of base
Bonus, 100–120% of target+25% of base
Bonus, above 120% of target+50% of base
Tier 3 · Pure per-outcome
Per unit · floor ₹1.5L / month · cap ₹6L / month

Only where the unit is countable and your system already emits it. You pay for the outcome, not the attempt.

Resolved support ticket₹95 – ₹190
Qualified meeting held₹15,000 – ₹30,000
Qualified lead₹4,500 – ₹15,000
Recovered rupee (dues, disputes)20–25%

Benchmark: a human-handled ticket costs ₹500–₹1,100. An in-house SDR meeting costs roughly ₹78,000–₹1,09,000 fully loaded.

Tier 4 · Gain-share — ₹1 Cr+ opportunities only
₹3L / month floor + 20–30% of verified value · cap ₹1.5 Cr / yr

The top of the ladder, and the one we are most careful about. A non-refundable floor covers our cost base; above that we take a tiered share of value your finance team has certified.

Of the first ₹1 Cr of verified value20%
From ₹1 Cr to ₹3 Cr25%
Above ₹3 Cr (to the annual cap)30%
Maximum measurement window per cohort2 quarters
The machinery

What makes an outcome contract survive contact with reality.

Baseline

Trailing twelve months of your data, normalised for seasonality, volume and mix. Locked at signature, changed only by formal change control. Exogenous factors named in a schedule so neither side can argue them later.

Attribution

A randomised 10% holdout wherever you can give us one. Where you can’t, a pre-agreed proxy metric decided upfront. Never last-touch.

Verification

Quarterly certification by a joint steering committee, with your finance team’s sign-off required before we raise an invoice. Shared dashboard, buyer audit rights, fifteen-day dispute window.

Clawback

If the verified value reverses in the following quarter, we repay pro-rata — capped at 100% of that quarter’s outcome fee. We have not met anyone else who will write this down.

All fees are professional services fees under SAC 9983, quoted exclusive of GST, and invoiced against a defined quarterly certification date. Not tax advice; your finance team should review the contract.

Choosing

Where most clients start.

Where you areStart hereWhy
Never bought AI services; no clean baselineTier 0 → Tier 1Build the baseline first, then take upside only. No downside while you learn how we work.
You have data and a number somebody ownsTier 2Lower base, real fee at risk. The cheapest way to buy accountability.
A countable unit your system already emitsTier 3Tickets, meetings, recoveries. Nothing to argue about — the system counts it.
₹1 Cr+ P&L line, CFO-sponsored, holdout availableTier 4The largest upside for both of us, and the only tier where we underwrite the whole number.
Comparison one

A chat window is where your team thinks. This is where your business runs.

Generic AI is not worthless — your team uses it every day and it is making them faster. But it is ungoverned, unmeasured and un-auditable. All three are fixable, and the fix is what you are paying for.

A Digital Theory agent system
Generic AI chat in a browser
Reaches your systems
Reads and writes to SAP Business One, Odoo, Salesforce, your warehouse and your ticketing system.
No access to ERP, CRM, warehouse or ticketing. Every answer is ungrounded in what is true today.
Memory
Persistent, governed, org-level memory — the Growth Graph.
Session-scoped, per-user, non-transferable. Nothing the organisation learns is retained.
Accuracy on business tasks
Grounded retrieval cuts hallucination 30–70%; under 2% on grounded summarisation.
Legal-query hallucination measured at 58–88% (Stanford RegLab).
Evaluation
Offline evals pre-deploy, online LLM-as-judge in production, every failure converted into a regression test.
None. No ground truth, no regression tests, no quality gate.
Observability
OpenTelemetry gen_ai.* spans — model calls, tokens, agent steps, tool executions.
No traces, no token accounting, no audit trail.
Governance
Policy, role-based access, audit log, human-in-the-loop and a model allow-list from day one.
52% of organisations have no formal policy on external AI tools.
Data exposure
Data stays inside the governed boundary — DLP, redaction and retention controls.
27% of employees have entered confidential company data into public AI tools.
DPDP readiness
Purpose-limited, consent-manager compatible, auditable — ahead of May 2027 enforcement, penalties to ₹250 Cr.
Consumer accounts sit outside consent management, retention limits and audit obligations.
Payback
Agreed in the blueprint before we build, then measured against the locked baseline every quarter.
Unmeasured — that is rather the point. There is no instrumentation to measure it with.

Sources: MIT NANDA The GenAI Divide (2025, preprint) · Chroma Context Rot · Stanford RegLab · Gartner (June 2025) · IBM · Cyberhaven · India DPDP compliance timeline.

Comparison two

A typical agency versus us.

Eighty-five percent of agencies prefer retainers. There is nothing wrong with a retainer — but you should know what you are buying and who carries the risk.

Digital Theory
A typical agency or consultancy
What you buy
A growth system that runs inside your business.
Hours, headcount, decks and campaigns.
Pricing
Outcome contracts with a locked baseline, a floor and a cap. Minimum retainer — or none.
A monthly retainer regardless of result.
Time to first value
2-hour audit · 5-day blueprint · 30-day deployment.
6–12 months to deployment, against an 8-month expectation.
What gets measured
The number in your P&L, against a randomised holdout.
Impressions, ROAS, deliverables shipped.
Where the work lives
In your ERP, CRM and warehouse — with an eval suite you own.
In the agency’s tools and the agency’s heads.
What you own at the end
Agents, evals and a Growth Graph that appreciate.
A body of work you have to re-buy every year.
Accountability
Fees at risk, in writing, with a clawback clause.
“Market conditions changed.”
When you stop paying
The agents keep running.
Everything stops.
Start here

Two hours will tell you more than a quarter of pilots.

Bring the people who own revenue, operations and data. Leave with a written baseline and three costed agents, ranked by payback — whatever you decide next.

₹75,000 · credited in full against any engagement signed within 30 days