Partner Payouts

Your contracts are not complicated.Paying them by hand is.

A payout scheme is a set of conditions ending in a rate. DataTwin takes the scheme as it was agreed, turns it into rows, and computes every partner’s payout from your own volume and activity data. Tiers, slabs, guarantees, incentives and deductions all resolve in one pass. We start by recomputing periods you have already paid, read-only, so you see what the spreadsheet got wrong before you change anything.

How Discover works
  • A new scheme is a new sheet, not a new build
  • Slabs behave as your contract defines them
  • Read-only until you turn computation on

One scheme, two forms

A scheme is a set of conditions ending in a rate

The scheme, as agreed → the same scheme, as rows

Live process

The scheme, as it was agreed

  • CustomerBusiness model
  • TerritoryStation, cluster, block
  • CategoryAnd offering
  • Resource classThe asset used
  • Partner classDirect or vendor
  • Rate formFixed, variable, tiered
  • PeriodMonth, day, slot, season
  • RatePer unit or per day
  • Slab rateBy band
  • GuaranteeA floor, not a rate

DataTwin

Flattened to rows

The same scheme, as rows

  • Cluster 1 — Category A · DirectVariable · 15
  • Cluster 1 — Category A · DirectTier, slab 1 · 10
  • Cluster 1 — Category A · DirectTier, slab 2 · 11
  • Cluster 1 — Category A · VendorFixed · 16
  • Cluster 2 — Category B · DirectGuarantee · floor

Why this matters A new scheme is a new sheet, not a new build. Six commercially different schemes are six sheets on one engine, which is why the second scheme does not wait behind an IT queue — read-only to start, nothing changes in your payout run or your ledger.

Every stage runs on the same engine that powers the DARP Framework and FSCP: acquire, process, govern, report.

The seven axes

Every payout scheme varies along the same seven axes

Different industries use different nouns. The structure underneath does not change, which is why one engine covers a fleet operator, a distributor and an insurer without a rewrite for each.

Who

Which partner, on what terms

Customer, business model, and the partner class: direct, vendor, franchisee, agent.

Where

Where does geography bite

Station, cluster, block, region, or nowhere at all when a scheme runs nationally.

What

Which offering it applies to

Category, product, or a bundle of products treated as one line in the scheme.

With what

Which resource did the work

Vehicle class, machine, seat, licence. The asset changes the rate.

When

Over what window

Month, day, shift or slot, and named periods such as a festival or campaign.

How much

What is the rate applied to

Units, volume, days present, value shipped, or a band over any of those.

How it pays

What shape does the money take

Fixed, variable, tiered or slabbed, minimum guarantee, one-off incentive.

The leaf

And at the end of every path, a rate

Seven answers select one row. That row carries the rate. Nothing about the scheme lives in code.

The payout

From the contract to the payment

1

Part 1

The scheme becomes a sheet, and the sheet is checked before it pays

A contract is negotiated as a tree and signed as a document. Neither form can pay anybody. It becomes payable when every path through it is a row, and the rows are tested against each other before the first run rather than after the first dispute.

Captured

From the contract, not retyped

Rates, conditions and effective dates read off the signed agreement and turned into rows, with the source document kept against them.

  • Annexures, rate cards and amendments read in one pipeline
  • Every row traceable back to the clause it came from
  • Effective dates carried, so a rate change does not overwrite history

Tested

Before it goes live

The sheet is checked for the faults that quietly double or halve a payout: gaps, overlaps, duplicates and conditions that can never be met.

  • Duplicate rows, and rows that differ only in the rate
  • Bands that leave a gap, or overlap so two rates both apply
  • Conditions no partner can satisfy, flagged rather than paid

Versioned

Rates change, history should not

Each version is dated. A rate backdated to January recomputes the periods it touches, instead of silently changing what was paid.

  • Retrospective changes recompute the periods they touch
  • The difference lands as arrears or a recovery, with the reason
  • What was paid under the old rate stays visible
2

Part 2

The measure has to be right before the rate matters

A correct rate applied to a wrong count pays the wrong amount, and nobody notices because the rate looks right. The volume, the days and the value all come from your own systems and are reconciled before anything is multiplied.

Counted

From your systems, not a summary

Units, trips, shipments, disbursals, policies or hours taken from the transaction record itself, at the grain the scheme is written at.

  • Counted at the grain the scheme uses, not rolled up first
  • Cancellations, returns and reversals removed from the count
  • The same unit never counted under two territories

Qualified

What actually counts

Not every unit qualifies. Exclusions in the contract are applied to the measure before any rate is chosen.

  • Excluded categories, customers and channels removed
  • Attendance and activity thresholds evaluated per partner
  • Disputed and cancelled units held out until they settle

Reconciled

Against the source of truth

The measure is agreed against the operating system before payout, so nobody is arguing about the count afterwards.

  • Volume agreed against the operating system, not the payout sheet
  • Differences listed by partner and period, with the records behind them
  • The agreed count frozen once the period closes
3

Part 3

Then the computation, in the shape your contract actually uses

Most payout tools support one or two rate forms well and bend the rest to fit. Real contracts mix them inside a single period, and the mixing is the part that goes wrong.

Fixed and variable

The straightforward half

A rate per day, per unit or per period, applied to the qualified measure. This part is rarely where money is lost, and it is where most tools stop.

  • Per unit, per day, per trip, per policy or per period
  • Multiple rates in one period, each with its own window
  • Currency and rounding handled at the rate, not at the total

Tiers and slabs

As your contract defines them

A band pays either on the whole volume or only on the units inside it. We do not impose a convention, because contracts genuinely differ.

  • Whole-volume or marginal, set per scheme rather than globally
  • Retrospective tiers that reprice earlier volume when a band is hit
  • The band that was applied shown next to the one below it

Guarantees and incentives

Floors, cliffs and one-offs

A guarantee is a floor evaluated over a period, not a rate. An incentive often has a cliff: one unit short and the whole component is zero.

  • Guarantee compared against the computed total, then topped up
  • Cliffs shown as a distance to the threshold, before the period ends
  • Seasonal and campaign incentives applied only inside their window

A contract question

Marginal or whole-slab is a contract question, not a product setting.

Ask whether volume of 70 against bands of 1 to 60 and 61 to 100 pays the higher rate on all 70 or on the last 10. Both are common, both are correct somewhere, and the difference is the single most frequent cause of a payout dispute. We read it from your contract and show which reading was used.

4

Part 4

One partner can qualify under four schemes in the same month

Base payout, a seasonal incentive, an attendance bonus and a minimum guarantee can all apply to the same partner in the same period. Whether they stack, whether the best one wins, or whether the guarantee absorbs the rest is a commercial decision. It should be recorded, not improvised.

Precedence

Decided once, applied every time

The order in which schemes resolve is configured with you and then applied identically to every partner and every period.

  • Stack, best-of, or absorb, set per pair of schemes
  • The same rule applied to every partner, with no local exceptions
  • A change of rule recomputes rather than applying from now on

No double counting

One unit, one payment

The same volume paid under two overlapping schemes is one of the largest and quietest leaks in any payout run.

  • Every unit tagged with the scheme that paid for it
  • Units claimed by two schemes surfaced before payment
  • Overlapping territories and categories resolved explicitly

The workings, kept

Not just the answer

For every partner and period, the computation is retained: which rows applied, in what order, against what measure.

  • Every payout traceable to the rows that produced it
  • The rejected alternatives kept alongside the applied one
  • A dispute answered from the record rather than rebuilt
5

Part 5

What comes off, before anything goes out

The gross payout is not what the partner receives. Advances, damages, shortages, asset recoveries and withholding all sit between the two, and each one is a place where a partner is either short-changed or overpaid without anyone intending it.

Recoveries

Netted, with a reason

Advances, equipment, damages, shortages and penalties netted against the payout, each carrying the event that caused it.

  • Advances recovered on the schedule that was agreed
  • Damages and shortages linked to the incident record
  • Nothing deducted without a reference the partner can check

Withholding

At the right rate, per class

Tax withheld according to the partner class and the applicable rate, with certificates produced rather than requested later.

  • Rate driven by partner class and registration status
  • Certificates generated with the payout, not at year end
  • Thresholds tracked across the year rather than per payment

The payment file

Ready to run

The bank file and the accounting entry come from one computation, so the books and the bank cannot disagree.

  • One computation behind the payment file and the journal
  • Entity, cost centre and account determined from the scheme
  • Failed and returned payments traced back to the payout line
6

The partner portal

The partner stops calling to ask how the number was reached

Most of a payout team’s month is not computing payouts. It is explaining them. A partner who can see their own volume, their own band and the distance to the next one asks a different and much shorter question.

Their own numbers

Live, not at month end

Volume so far, the band it currently falls in, and what has accrued. Visible during the period, when the partner can still act on it.

  • Running volume and accrued payout, updated as data arrives
  • The current band and the distance to the next one
  • Attendance and qualifying thresholds shown against the target

The statement

Workings included

Every payout line with the scheme, the rate, the measure and the deductions that produced it. The same view your team sees.

  • Gross, deductions and net, each with the rows behind them
  • Downloadable statement per period, per partner
  • Withholding certificates attached to the period they cover

Queries in one place

Not in an inbox

A partner disputes a line rather than an amount, your team answers against the record, and the exchange stays attached to the payout.

  • Disputes raised against a specific line, not a total
  • The evidence already attached when the query opens
  • Resolution recorded against the period it affects

Then it runs forward

The same sheets run forward, before the money leaves

Prevent is not a second build. The rows that recomputed your history are the rows that compute the next run. That is where a one-off recovery starts compounding.

  1. 1

    Scheme agreed

    A new scheme becomes rows and is tested for gaps and overlaps before it is live.

  2. 2

    Measure agreed

    Volume and activity reconciled to the operating system and frozen for the period.

  3. 3

    Payout computed

    Every scheme resolved in the agreed order, tiers and guarantees together.

  4. 4

    Checked and approved

    Outliers and swings against prior periods surfaced before approval, not after payment.

  5. 5

    Paid and posted

    One computation behind the bank file, the journal and the partner statement.

Under the hood

What is actually doing the work

Specialised agents with narrow, defined duties, coordinated by an orchestration agent. Deterministic rules run wherever something posts to the books. Nothing posts because a model was confident.

Reading the agreement

Scheme Agent

Contracts, annexures and rate cards read into rows, including scanned and amended documents, with each row tied to its clause.

Applying the rows

Computation Agent

Measure qualified, schemes resolved in order, tiers and guarantees applied, and the workings kept against every payout line.

Finding what looks wrong

Exception Agent

Swings against prior periods, partners near a cliff, units claimed twice and deductions without a source, surfaced before payment.

What changes

What changes, honestly stated

Not a longer list of features. A different distribution of who does what, and when a difference is found.

The scheme

How payouts run today

A signed PDF, retyped into a spreadsheet by whoever owns the run

How they run on DataTwin

Rows tied to the clause they came from, tested for gaps and overlaps before they pay

A new scheme

How payouts run today

A new tab, a new formula, and a person who now cannot take leave at month end

How they run on DataTwin

A new sheet on the same engine, live in the period it was agreed

Tiers and slabs

How payouts run today

Whichever reading the formula happened to encode

How they run on DataTwin

Marginal or whole-volume as the contract defines it, and the reading is visible

Overlapping schemes

How payouts run today

Resolved differently by different people in different months

How they run on DataTwin

One precedence rule, applied to every partner and every period

Disputes

How payouts run today

Rebuilt from scratch each time, often to a different answer

How they run on DataTwin

Answered from the retained workings, in the partner portal

Rate changes

How payouts run today

Backdated by editing the sheet, so history quietly changes

How they run on DataTwin

Versioned and dated; a backdated rate recomputes and raises arrears

Audit

How payouts run today

Sample testing of a spreadsheet nobody fully understands

How they run on DataTwin

Every payout traceable to rows, measure and approval, across the full population

Start here

Before you change anything, find out what it has already cost you

Everything above is prevention; it stops the next run going wrong. It does nothing about the periods already paid, which are sitting in your payout history right now. So we start there, read-only, before there is anything to sign.

2min

This is how long it takes to understand how much can be recovered.

Send us the schemes and the volume history. We recompute every partner and every period across the full transaction population, not a sample, and come back with what is recoverable, split by heading. Before any contract.

  • The schemes
  • The volume history
  • Every partner and period recomputed
  • What is recoverable, by heading

Overpaid

The wrong band applied. A scheme run past its end date. A condition never actually met.

Underpaid

Volume wrongly excluded. A rate agreed and never applied. Arrears nobody asked for.

Paid twice

The same units paid by two schemes. A guarantee topped up where the base already cleared.

Never netted

Advances, damages and shortages that were never deducted, and withholding taken at the wrong rate.

The same engine runs payouts for logistics and last-mile fleets, gig and field-force platforms, franchise networks, staffing and managed services, and bank and NBFC sourcing partners. The nouns change; the seven axes do not.

How Discover works