Blog · Wallet share and penetration
Evaluate targeted account-growth outreach with a declared treatment and comparison population, a fixed observation window and reconciled outcomes. Separate outreach activity from incremental contribution.
A share-of-wallet growth experiment tests the follow-up action, not just the accuracy of the account list. A useful gap report may identify customers to discuss, but subsequent orders do not automatically prove that the report caused growth. Seasonality, contract renewals, price changes and normal customer activity can produce the same movement.
The customer-gap conversation guide owns how to discuss an opportunity. This guide defines how the business will evaluate the intervention afterward.
Specify the intervention in terms another account manager can repeat: for example, a category review invitation and a documented follow-up offer during six weeks. Distinguish it from ordinary service calls that continue in both groups.
Choose a primary outcome such as net invoiced category revenue per assigned account over the following quarter. Record credit-note treatment, baseline period and the outcome cutoff. Orders, invoices and collected cash are different measures; choose one primary basis and show the others separately if useful.
Add safeguards such as margin, returns and service burden so increased sales do not conceal deteriorating economics. A wallet estimate itself should not be the outcome if its method is refreshed during the experiment.
Freeze eligibility using product fit, account status, decision timing and evidence quality. Remove duplicate sites belonging to one buying decision. Where sellers serve related entities or share information, consider assigning the whole group together rather than pretending each site is independent.
Balance or group accounts by important pre-existing characteristics such as size, industry, prior category revenue and owner. NIST describes randomized block designs for controlling known nuisance factors and random assignment for treatment allocation. The account-growth application here is an original illustration.
If randomization is impractical, record the matching rule and acknowledge that unobserved differences may remain. A comparison group chosen after seeing the results is not an independent test.
Track assignment, attempted contact, actual discussion, qualified opportunity, quote and order separately. These stages diagnose execution; they are not interchangeable denominators for the primary result.
For a synthetic twenty-account treatment group, 18 attempts, 14 discussions, six qualified opportunities, four quotes and three orders describe the funnel. The main economic result still includes all twenty assigned accounts. Excluding unanswered accounts would make reachability disappear from the evaluation.
Keep the date and owner of every action. Record contamination when comparison accounts receive the same intervention, and preserve the original assignment rather than moving successful accounts into the treatment group afterward.
The following revenue amounts are invented, with twenty accounts in each group and equal observation windows.
| Group | Baseline revenue | Follow-up revenue | Change |
|---|---|---|---|
| Assigned treatment | $200,000 | $220,000 | +$20,000 |
| Assigned comparison | $200,000 | $210,000 | +$10,000 |
The treatment group's average increase is $1,000 per account. The comparison group's is $500. The difference in changes is $500 per account, or $10,000 over the twenty treatment accounts.
That is an estimated comparison effect under the design assumptions, not proof of causation from this small illustrative table. At an assumed 30% incremental contribution margin it corresponds to $3,000 contribution before intervention costs. Subtracting $2,000 of additional activity cost leaves $1,000 under those assumptions.
Retain account-level observations and examine unusually large orders. A single outcome can dominate a small group. Obtain an appropriate uncertainty analysis for the design before presenting a precise effect as established, particularly with clustered assignments or unequal group sizes.
Check baseline balance, price changes, acquisition events, account transfers, returns and the timing of normal renewals. Report planned exclusions and any new deviations separately. Do not remove unfavorable accounts simply because their circumstances make the result less attractive.
Conclude with reach, economic result, uncertainty, execution exceptions and the cost of repeating the action. The result may justify a larger test, a revised offer or stopping the intervention.
Use the business-case guide to distinguish scenario justification from observed return. Bring the eligible population and desired review outputs to Covirage; analysis and follow-up evaluation depend on available records and agreed scope, not a guaranteed growth outcome.
That comparison can be biased because reps may contact accounts they already expect to buy. Prefer a declared assignment process; when assignment is not random, document the selection and treat causal conclusions cautiously.
Keep them in the primary assigned-group result. Failure to reach an account is part of the intervention's effectiveness. A contacted-only analysis can be shown as a secondary operational view, with its different denominator labeled.