Discounts need a contribution justification

A discount is a transfer of economic value under an offer. Its purpose might be to acquire a new relationship, retain a customer, encourage volume, change payment timing or clear an obsolete service commitment. Those purposes require different evidence. Describing every concession as growth investment obscures which mechanism is supposed to repay it. A discount request should name the customer action being purchased and the obligation the operator accepts in return.

The first question is whether the concession changes price for the same service or changes the package itself. A lower price with a reduced service scope can preserve contribution differently from a lower price with unchanged duties. A volume commitment can make some costs more predictable, while an open-ended discount can reduce revenue without reducing work. The offer record should retain eligibility, duration, customer obligation and the consequences of failing that obligation. A percentage on a price list is not a complete commercial arrangement.

Meridian’s discount-and-acquisition case supplies a lower starting price and a higher acquisition budget together. It retains CAC, churn, expansion, service cost, operating envelope and payment assumptions. The model therefore creates more expected new customers from the larger budget while giving each active account a lower starting contribution path. It does not estimate that the discount causes greater acquisition. The two changes are an authored package to compare, not a measured response or an optimized promotion.

This distinction is commercially important. Sales teams can reasonably propose that a discount improves conversion, but a supplied CAC assumption does not verify that claim. If the acquisition budget is increased independently, the model’s greater customer count can be caused entirely by that input. Attributing the resulting count to the lower price would misread the calculation. The analyst should state the acquisition assumption and identify offer-specific evidence that could establish a real conversion effect.

Discounts can also change who buys. A promotion may attract customers whose recurring support demands differ from the reference, or it may bring forward customers who would have bought later at the original price. The model has no selection or timing mechanism for these possibilities. The decision record should identify whether they matter for the proposed action and what evidence would reveal them. Ignoring them because a contribution ratio remains positive turns a scenario into an unsupported intervention story.

The payment purpose deserves particular attention. A concession for immediate payment can improve cash timing while reducing recurring contribution. That tradeoff is different from a concession for longer credit or for an uncertain future volume promise. A real payment-linked offer needs an enforceable eligibility event and a collection record. Entering both lower price and lower receivable days would represent a supplied joint scenario; it would not establish that customers comply or that the discount is the cheapest way to obtain cash.

Meridian’s fictional record separates an acquisition concession from a collection intervention. The discount-and-acquisition fixture changes price and budget only. The collection fixture changes receivable days only. This separation preserves the mechanisms so that the committee can see which input does what. A complete negotiated offer could combine them later, but it would need a new typed fixture and evidence for the terms. Combining attractive results from separate cases without recomputation would overstate what the model has shown.

The comparison challenges a superficial growth argument. The package acquires more expected accounts under its supplied acquisition assumptions, and it can produce substantial later activity. That does not make it easier to fund. A larger early acquisition outflow and lower starting recurring contribution can deepen the modelled resource need. A committee must inspect the path and the customer-value perimeter instead of treating a larger final customer base as a sufficient answer.

The case does not prove that all discounts are undesirable. It proves that the stated package cannot be defended by the customer-count story alone. A different concession could reduce acquisition cost, preserve service contribution through a narrower package or materially improve collection. Those possibilities require their own inputs and evidence. The commercial owner should describe the alternative mechanism rather than borrowing a favourable intuition about discounts in general.

An approval record should distinguish a permitted discount from an unbounded delegation. Sales staff need to know who is eligible, which service is included, what customer commitment is required and when the concession expires. Finance needs to know how the net amount is represented in the contribution and cash ledgers. Operations needs to know whether demand can be served. Treasury needs to know whether payment terms change. Without these details, the financial comparison is disconnected from the concessions being offered.

The evidence requirement can be modest for a routine concession under an established policy. It becomes stronger when the concession changes the customer universe or creates a sustained below-reference amount. A one-time correction for a documented service failure is different from a permanent acquisition tariff. The decision record should not force them into the same category simply because both appear as a price reduction. The purpose determines the relevant response and stop event.

Cannibalisation is a useful conceptual falsifier. If customers who would accept the reference offer receive the concession without changing their behaviour, the discount can transfer contribution rather than acquire an incremental relationship. The canonical engine does not estimate this effect. A real evidence design should therefore distinguish newly induced activity from relabelled or accelerated existing demand where that distinction could change the decision. The paper makes no empirical estimate of how often this occurs.

Another falsifier is an unfulfilled customer commitment. If a concession is justified by a promised volume or prompt payment, the concession record needs a response when that promise is not met. The operator might revert future pricing, suspend further service credit or require a new approval. The actual contractual entitlement must be established from the agreement; the management protocol cannot create it. What the protocol can do is prevent a conditional benefit from being treated as unconditional throughout the financial case.

Meridian completes a concession incidence note for OFFER-B. The reference package promises standard recurring support; the proposed package lowers its amount while leaving that duty unchanged and supplying greater acquisition spending separately. The note identifies the action supposedly purchased by the concession as additional accepted activation, but marks the causal connection unverified. It distinguishes the customer benefit from the operator's acquired-account assumption. Lena retains the offer comparison; Tomas retains the canonical outputs; neither treats the larger acquired count as observed response. This completed record makes the missing relationship explicit without losing the favourable or adverse mechanisms already visible in the exact package.

A bounded concession experiment can be preferable to a long predictive debate. If it uses established service capacity, settled customer permissions and an affordable commitment envelope, it can generate useful offer-specific evidence. The appropriate approval then specifies the experiment’s action boundary and stop event. The canonical illustration does not compute that experiment’s uncertain response, so the resulting evidence should be recorded separately before it is used to update a scenario.

The concession note then compares a correction with an acquisition incentive. A correction responds to an established service failure and needs the relevant case and applicable remedy. An incentive seeks a future customer action and needs an eligibility universe, offer event and assessment of incrementality. Calling both discounts can conceal the evidence they require. In Meridian's fictional disposition, OFFER-B remains the second type and cannot borrow the evidential strength of a documented correction. If a customer would accept the reference without the incentive, the concession transfers contribution rather than purchasing the intended change. The instrument records that counterfactual question and identifies which proposed benefit remains unsupported.

This chapter leaves a deliberate limit: it does not solve the optimal discount. A valid optimization would require a response model, constraints and an objective that match the offer and customer universe. The exact canonical package is sufficient to demonstrate a tradeoff and to hold an unsupported implementation request. It is insufficient to claim that another price or acquisition combination is optimal. That boundary prevents the paper’s useful decision record from becoming a promise of automated commercial strategy.