Cross-Sell Screening
When you need this method
Your installed base is run toward an expansion target: net revenue retention has to rise, so every account gets offered an additional module. What nobody measures is the cost per account. A share of your customers already consumes more support, more credit notes and more discount than their revenue carries. In exactly those accounts, expansion makes things worse rather than better, and the cohort average hides it.
Approach
- 1Measure four traits per existing customer across several periods: spending level, frequency of credit notes, reversals and downgrades, number and cost of service requests, and the share of deals closed on discount.
- 2Test the traits for persistence. A single spike does not count; a recurring pattern across several periods does.
- 3Assign support and service costs to the actual account instead of spreading them evenly, otherwise the loss disappears into the cohort average.
- 4Remove accounts showing the adverse pattern from the expansion list: no extra module, no new service category, no campaign.
- 5Instead, check whether an upsell holds, meaning a higher tier of what the account already uses, because that opens no new service paths.
- 6Where no viable upsell exists, mark the account as not to be targeted and pull support intensity back to what the contract promises.
- 7Re-measure contribution per account after every expansion step and recalibrate the screening rule against your own data.
Typical application
A typical B2B SaaS in the HR space sets itself an NRR target and has account management offer every existing customer a second module. A year later revenue is higher and contribution is lower. The per-account view shows a small group that stood out even before the expansion: many support tickets, recurring credit notes after complaints, every deal closed on discount. With the second module their ticket volume doubled, the rollout consumed consulting days, and several of these customers then renegotiated the total price downward. The team takes those accounts off the expansion list, offers them a higher tier of the module they already use, and from then on tracks contribution per account net of cost to serve.
Limits and counter-indications
The evidence comes from catalogue and fashion retail, a retail bank, and financial and IT service firms, not from software subscriptions. Transferring it to module expansion is a reasoned analogy, not a measured finding. The method presumes that cost to serve is captured per account at all; without those numbers the screen degrades into a gut call. Where marginal cost per additional seat is near zero, for instance in pure self-service, the effect is small. Stopping expansion is also a relationship decision: it prevents losses but can trigger churn, and that counter-effect is not measured in the primary source. There is counter-evidence too: a follow-up study in a consumer setting finds that discount-driven category buying depresses purchase probability and purchase amount in the short run but works positively in the long run. The static view can therefore mislead.
How to measure impact
Track contribution per account net of cost to serve and watch how it moves with each additional category. Alongside it, two portfolio figures: the share of expanding customers with a negative contribution, and that group's share of total losses from customer relationships.
Related methods
Sources
- 1.Shah, Kumar, Qu & Chen: Unprofitable Cross-Buying: Evidence from Consumer and Business Markets, Journal of Marketing 76(3), 2012, S. 78-95 (opens in a new tab) · American Marketing Association / Journal of Marketing (SAGE) · 2012 · academic and scholarly literature · evidence of effectivenessCarries the core finding: across the customer databases of five firms, 10 to 35 percent of customers who cross-buy were unprofitable, and that group accounted for 39 to 88 percent of total losses from customer relationships, with losses rising at higher levels of cross-buying. It names four persistent traits: limited spending, revenue reversals, excessive service requests, and promotion-driven buying. Limitation: the full text sits behind a paywall, so citation and abstract were verified through the lead author's publication list; the firms studied are retail, banking and services businesses, not software subscriptions, and the two-stage screening framework it sketches is documented by name only.
- 2.Shah & Kumar: The Dark Side of Cross-Selling, Harvard Business Review 90(12), Dezember 2012 (opens in a new tab) · Harvard Business Review · 2012-12 · practitioner source · supports the underlying mechanismCarries the typology behind the screen: service demanders require markedly more support once they cross-buy, revenue reversers return goods or reverse revenue, promotion maximizers only buy at steep discounts, and spending limiters operate under structurally capped budgets. Limitation: this is the practitioner version of the same study, not additional evidence, and the full text is paywalled beyond the opening section, so only the publicly readable typology is used here.
- 3.Guerreiro, Bio & Merschmann: Cost-to-serve measurement and customer profitability analysis, The International Journal of Logistics Management 19(3), 2008, S. 389-407 (opens in a new tab) · Emerald Group Publishing · 2008 · academic and scholarly literature · describes the methodCarries the measurement procedure without which the screen cannot be applied: cost to serve as the administrative, commercial and logistics cost of serving a customer, captured through activity-based costing; customer profitability as the contribution of what was delivered less those costs; plus the cumulative profitability curve showing that a share of customers erodes total profit again, and the classification of accounts by net margin against cost to serve. Limitation: the paper is a single case study in a Brazilian food company, the authors themselves rule out generalisation, and it says nothing about expansion into further categories.
- 4.Morisada, Miwa & Dahana: Dynamic Impact of Unprofitable Cross-buying on Purchase Incidence and Purchase Amount, Journal of Management Research 10(2), 2018, S. 65-84 (opens in a new tab) · Macrothink Institute · 2018-04 · academic and scholarly literature · limits the methodSets a boundary on the finding: the authors explicitly fault the primary source for treating the effect statically, and in the purchase histories of an online shopping mall they find that discount-driven category buying lowers purchase probability and purchase amount in the short run while affecting both positively in the long run. Limitation: it covers consumers, purchase behaviour rather than contribution margin, and a single dataset, and the outlet is a small open-access journal without the standing of the primary source.
Origin: Shah / Kumar