Best time for door-to-door sales: a field test

Choose the question before the shift

Start with a narrow decision: does one of our two working windows produce more recorded conversations per completed first attempt?

That measures contact, not sales success. Qualified leads, appointments and sales need their own definitions and review. A higher conversation rate does not establish that more people bought.

Choose both windows within your existing campaign rules and rep-safety arrangements. This guide does not set permitted doorstep hours. Follow householder instructions and your organisation's exclusions throughout.

Agree what counts as a conversation before anyone starts. Use the same definition for every rep and window. The door-knocking KPI guide explains how to keep reporting definitions consistent.

Door-knocking KPI guide

Plan comparable work

For this comparison, a first attempt means your campaign's first recorded attempt at an eligible address, not a returning visit by a different rep. Give each address to one window only. Check the history first and keep callbacks, previous refusals and excluded addresses out of the first-attempt batches. Do not make an extra visit just to generate test data.

Define a batch as a named set of eligible addresses in a patch. Where practical, split comparable work across both windows for each rep. Record how you assigned the batches. If one window gets experienced reps on familiar streets and the other gets a new starter in a different area, the pooled rate will not tell you much about timing.

Across later comparable sessions, balance which reps and batches go into each window where feasible. Record differences you cannot control, such as weather, access problems, the offer or a changed opening script. Never describe ordinary field records as a controlled or randomised experiment unless you actually used that design.

NIST's guidance on experimental design explains why background factors matter and how grouping comparable work can help a comparison. The worksheet below is a practical adaptation. NIST does not establish any best hours or sales benchmark for UK canvassing.

• NIST's guidance on experimental design

Copy this comparison worksheet

Fill this in before the work, then complete the result rows afterwards. You can copy these fields into your own working sheet.

Decision to review: ____________________

Campaign and patch: ____________________

Eligible first-attempt addresses and exclusions: ____________________

Conversation definition: ____________________

Earlier permitted window, local dates and times: ____________________

Later permitted window, local dates and times: ____________________

Rep/batch assignment method and known differences: ____________________

Callbacks held separately, with agreed times preserved: ____________________

Use one result row for each rep, batch and window:

Rep and batch ID: ____________________

Date and actual time window: ____________________

Planned eligible first attempts: ____________________

Completed distinct first attempts, with matching address/attempt references: ____________________

Conversations linked to those completed attempts: ____________________

Missing or unclear records: ____________________

Weather, access, offer, script or other differences: ____________________

Conversation rate, with raw counts: ____________________

Keep the address-level record in your normal controlled system. A comparison summary needs references that let the manager check the counts, rather than another copy of household details.

Finish the review with:

What the results support and what remains unknown: ____________________

Next scheduling action and reason: ____________________

Manager responsible and next review date: ____________________

Reason to pause, stop or repeat the comparison: ____________________

Read the counts before changing the schedule

The following home-improvement example is entirely fictional. The figures show the arithmetic, not customer results, a target or a recommended sample size. Each row uses different eligible households. Both reps work in both windows, but patch and session differences remain uncontrolled.

Rep A, earlier window: 40 completed first attempts, 12 conversations, 30%.

Rep A, later window: 40 completed first attempts, 16 conversations, 40%.

Rep B, earlier window: 40 completed first attempts, 14 conversations, 35%.

Rep B, later window: 40 completed first attempts, 18 conversations, 45%.

Conversation rate is conversations divided by completed first attempts, multiplied by 100. If there are no completed first attempts, the rate is unavailable, not 0%. Pool the raw counts rather than averaging percentages from batches of different sizes.

Earlier window: (12 + 14) / (40 + 40) x 100 = 32.5%.

Later window: (16 + 18) / (40 + 40) x 100 = 42.5%.

Difference: 42.5% - 32.5% = 10 percentage points.

Both reps also have a 10 percentage point difference between windows. That is a reason to investigate further. It does not show that the later window caused the difference, establish statistical significance or justify moving every shift later.

Check missing records before treating either rate as complete. If attempted addresses have unclear outcomes, show that uncertainty alongside the counts. A zero is a recorded result; a blank needs checking.

Write a scheduling decision you can revisit

A manager's note for the fictional example could read:

Keep both permitted windows for the next two comparable sessions. The later window had more recorded conversations for each rep, but we have not ruled out differences in the batches or session conditions. Before assigning new eligible addresses, check the history and balance rep/batch assignments where practical. Keep agreed callbacks separate. The canvassing manager will review the counts and missing records after those sessions, then record the next decision.

Two sessions is an example review point, not a universal evidence threshold. Set your own review date before collecting more results. If the work is too different or the records cannot be reconciled, retain the current schedule and fix the comparison first. You do not need to force a winner.

Use the canvassing patch-review guide when the next decision is where to allocate the team. This worksheet addresses when to compare first-contact work.

Canvassing patch-review guide