LONTA MARKETING GROWTH REVIEW
Version: October 10, 2026
Calculator: https://lonta.io/tools/marketing-growth/
Worked guide: https://lonta.io/blog/hvac-marketing-growth-control-group/

This worksheet checks a claim. Four totals do not prove causation or estimate uncertainty.
Copy and adapt this original worksheet, including for commercial use. Attribution is welcome,
not required. Third-party source material keeps its original rights.

1. DEFINE THE QUESTION BEFORE SELECTING A COMPARISON
Change being reviewed:
Implementation date and change-log reference:
Changed group and fixed membership rule:
Proposed comparison group and fixed membership rule:
Why would the comparison reflect outside changes in the changed group?
Could campaigns, customers, technicians or decisions cross between groups?

2. DEFINE THE RECORDS
Before dates (same for both groups):
After dates (same for both groups; same length as before):
Sold-job definition and cancellation/refund handling:
Time allowed for each inquiry to reach that outcome:
Record source and missing/unresolved records:
Earlier periods available to inspect both groups' levels and trends:

3. KEEP THE FOUR COUNTS SEPARATE
T0 = changed-group jobs before:
T1 = changed-group jobs after:
C0 = comparison jobs before:
C1 = comparison jobs after:

Observed change = T1 - T0.
Observed percentage = 100 * (T1 - T0) / T0; undefined when T0 is zero.

Same absolute change assumption:
Expected changed-group jobs without the change = T0 + (C1 - C0).
Difference = T1 - expected.
A negative expected count makes this scenario unusable for job counts.

Same percentage change assumption:
Expected changed-group jobs without the change = T0 * (C1 / C0).
Difference = T1 - expected.
Requires positive before counts in both groups.
This is a different assumption, not an automatic fix for unequal groups.

4. REVIEW COMPETING EXPLANATIONS
Ad spend and campaign changes in each group:
Actual weather over both periods:
Technician hours and dispatch capacity:
Phone coverage and booking process:
Service area, job mix, price and customer history:
Changes to measurement, CRM definitions or follow-up:
Other changes affecting only one group:

5. WRITE THE CONCLUSION
Which assumption is defensible, and why?
If the scenarios disagree, what additional evidence is needed?
What is observed, what is conditional, and what remains unknown?
Do not write 'statistically significant', 'proven incremental jobs' or a confidence interval
from these four totals. Do not choose the comparison after seeing which result is flattering.

INVENTED EXAMPLES — NOT CLIENT DATA OR AN INDUSTRY BENCHMARK
100 -> 140 versus 100 -> 130:
Observed +40 jobs (+40% of before). Both scenarios expect 130 and leave +10 jobs.
10 / 130 is about 7.69%; +10 jobs is not a 10% lift over that counterfactual.

100 -> 140 versus 400 -> 520:
Same absolute change expects 220 and gives -80 jobs.
Same percentage change expects 130 and gives +10 jobs.
Different signs are a reason to inspect the comparison and assumptions.

METHOD SOURCES (READ OCTOBER 10, 2026)
https://blogs.worldbank.org/en/impactevaluations/often-unspoken-assumptions-behind-difference-difference-estimator-practice
https://blogs.worldbank.org/en/impactevaluations/revisiting-difference-differences-parallel-trends-assumption-part-i-pre-trend
https://support.google.com/google-ads/answer/16104408?hl=en

Lonta AI / Lonta: https://lonta.io/about/