How to check the numbers in a performance case study
Check CPC, conversion rate and lead growth before publishing a marketing case study. Includes worked arithmetic, evidence boundaries and a verification worksheet.
1. Turn the headline into separate claims
A case study becomes easier to assess when each number answers a specific question: what changed, from which baseline, during which period and at what cost? I start with a claim register. Each row contains one statement, its source, the metric definition and any limitation. A screenshot with an upward arrow is useful only after the reader knows what was compared.
The historical Eighteeth & FGM Dental portfolio case reports 20% lower CPC and CPL, 25% more inquiries and a 30% improvement in inquiry conversion. These are resume-sourced figures for the named projects; the underlying periods and baselines require verification against advertising records. The example below explains how to audit them. It is not independent confirmation, a North American campaign claim or evidence that every number covers the same campaign.
2. Reconcile one simple funnel first
For a click-based funnel with one lead event, CPC equals spend divided by clicks, conversion rate equals leads divided by clicks and CPL equals spend divided by leads. Therefore CPL equals CPC divided by conversion rate, expressed as a decimal. These identities require compatible inputs. A platform's reported conversion rate may use eligible interactions rather than clicks alone, so check the actual column definition before using it.
The table is an illustrative USD example, not client data or a currency conversion of the historical case. At constant spend, increasing leads from 100 to 125 reduces CPL from $100 to $80. Those figures describe the same change in two ways. Reducing CPC from $5 to $4 while maintaining a 5% lead conversion rate also produces 125 leads.
| Illustrative metric | Before | After |
|---|---|---|
| Spend | USD 10,000 | USD 10,000 |
| Clicks | 2,000 | 2,500 |
| CPC | USD 5 | USD 4 |
| Leads | 100 | 125 |
| Click-to-lead rate | 5% | 5% |
| CPL | USD 100 | USD 80 |
3. Investigate numbers with different scopes
Now add a relative 30% increase in conversion rate to that same example. The rate becomes 6.5%. With 2,500 clicks, the model produces 162.5 expected leads, or a 62.5% increase from 100. A model can contain fractional expected outcomes; an actual lead count will be an integer. The arithmetic shows why 25% more leads, 20% lower CPC and 30% higher conversion cannot simply be stacked into that single model.
There may be a valid explanation. Conversion rate could refer to one landing page, CPC to one channel and lead growth to the whole account. The reporting windows or target events may also differ. Do not choose an explanation without evidence. Mark the relationship as unresolved and request the relevant exports. Also clarify whether a reported increase means a relative percentage or percentage points: moving from 5% to 6.5% differs greatly from moving from 5% to 35%.
4. Keep the timeline separate from causality
After checking arithmetic, build a timeline of changes to ads, creative, landing pages, lead handling, pricing and product availability. Give each entry a date and source. If several changes happened together, a before-and-after comparison cannot isolate their individual contribution. Saying conversion increased after a page update describes an observation. Saying the page caused the full increase needs a stronger basis.
A controlled experiment can compare a change against an original setup using a predefined outcome. Google Ads documents custom experiments, but the existence of an experiment feature does not validate a particular test. Allocation, comparable conditions, duration and available data still matter. In a historical case without a suitable test, qualify the conclusion beside the relevant result rather than hiding that limitation in a distant footnote.
5. Check lead quality and the cost boundary
More inquiries do not by themselves establish business value. Request valid-contact counts, qualified leads and sales, allowing time for the sales cycle. In professional dental equipment, a distributor inquiry, a student inquiry and a clinic purchase request may have different commercial meaning. Agree on the qualification rule before comparing periods. Otherwise the definition can change quietly to support the preferred conclusion.
The spend row also needs a boundary. Does it include media only, creative production, agency fees or the full acquisition cost? Do not compare media spend in one period with total marketing cost in another. Retain original currencies for historical evidence. When a value is unavailable, record it as missing instead of estimating a plausible number that nobody can later reproduce. The worksheet should expose these gaps before the case becomes public.
6. Publish a result another person can follow
Move verified claims into the article with their period, baseline, event definition and comparison method. If confidentiality prevents publishing absolute values, an index or percentage may still be useful, but maintain the underlying evidence privately. Confidentiality explains why a figure is withheld; it does not turn an assumption into proof.
The report-review tool on this site accepts a pasted report and suggests questions about missing information. It does not connect to ad accounts or independently verify a case study. Provide a redacted version, review the questions and attach the answers to your evidence register. A useful final check is whether another analyst can reproduce the calculation and explain what it does and does not establish without asking the original author to fill in the story.
Performance claim evidence register
Verification prompts, not underlying Eighteeth exports. Fill each gap before treating the claim as verified.
Download the working CSV template
Open in Excel or Google Sheets. Replace example rows and empty fields with your own data.
Frequently asked questions
Does a lower CPC always mean a lower cost per lead?
No. Conversion rate can fall enough to offset cheaper clicks. At a CPC of USD 4 and a 2% click-to-lead rate, CPL is USD 200. The calculation needs both traffic cost and the proportion of compatible clicks that produce a lead.
Can I publish a case without the original exports?
You can describe available information and identify its source honestly. Do not claim independent verification. If the baseline, period or scope is missing, keep that limitation visible and avoid combining separate percentages into a causal story.