How to use a holdout in a lead reactivation campaign

01

Decide whether a comparison is appropriate

Choose a group with a similar relationship and reason for inactivity. Before splitting it, deduplicate records and check the permitted channel, purpose, suppression status and rules in each recipient's jurisdiction. An old inquiry does not automatically authorize marketing. Hold uncertain records for review and exclude them from both groups until resolved.

Do not withhold a promised response, required service communication or an agreed follow-up to run an experiment. A holdout is for a discretionary campaign. Decide in advance how to handle people who contact you on their own. Give both groups the same ordinary service and record any outside campaigns that reach them.

Define the unit as a distinct eligible person or account. Keep related contacts together if a message to one person can affect another's purchase decision. Choose a start date, outcome window and a confirmed sale definition before sending. Measure complaints and opt-outs alongside sales.

02

Assign groups and calculate

Randomly assign eligible units to outreach and holdout using a reproducible method, then save the assignment and date. Do not move slow responders into the holdout after seeing results. Record attempts and failures, but calculate the main comparison on everyone originally assigned to each group, including people never reached. This preserves the original comparison.

Hypothetical example: 200 eligible accounts are assigned equally. After 90 days, 12 of the 100 assigned to outreach and 7 of the 100 assigned to holdout have a confirmed sale. The observed difference is 12% minus 7%, or 5 percentage points, equivalent to five additional sales per 100 assigned accounts in this sample. This is an estimate, not proof that exactly five sales were caused by outreach or that a later campaign will perform similarly.

Separately show attempts, replies, qualification, opt-outs and costs. An attempted-contact rate can help explain execution, but comparing only respondents with the holdout selects a different group and can distort the result.

03

Read the result with care

Small groups can produce a wide range of plausible differences. Before launching, decide the smallest change worth acting on and ask an analyst to assess whether the available audience can support a useful comparison. If the group is too small, report the descriptive outcomes and avoid a confident causal claim.

At the cutoff, use the same outcome window and sale definition for both groups. Check assignment errors, missing sales and differences in other outreach. Compare the estimated additional contribution with campaign expense and staff effort. Revenue alone omits delivery costs. Document the result and decide whether a later, appropriately sized campaign is worthwhile.