Define the outcome and its counting rule
Before launching ads for an analytics workshop, decide whether the goal is a paid booking, an inquiry, or an email subscription. These are not interchangeable outcomes. Document what counts once, what may count repeatedly, and which actions influence automated bidding. A thank-you-page view should not quietly stand in for a purchase unless the implementation reliably represents that purchase.
Google Ads distinguishes primary conversion actions in the Conversions column from broader reporting such as All conversions, with settings that affect inclusion. [1](https://support.google.com/google-ads/answer/6270625?hl=en) Read those definitions before comparing reports. Otherwise a campaign can appear to improve simply because someone added a low-value action to the measured outcome set.
Calculate the basic funnel with explicit denominators
Use a small hypothetical example: 10,000 impressions, 400 clicks, 20 tracked inquiries, and 600 currency units spent. Click-through rate is 400 divided by 10,000, or 4%. Average cost per click is 600 divided by 400, or 1.50. Cost per tracked inquiry is 600 divided by 20, or 30. These figures illustrate arithmetic, not expected performance.
For a click-based analysis, inquiry rate is 20 divided by 400, or 5%. Google Ads defines conversion rate using eligible ad interactions, which are not always clicks, and notes that multiple counted conversions can produce rates above 100%. [2](https://support.google.com/google-ads/answer/2684489?hl=en) Label your denominator rather than assuming every platform uses the same one.
Connect inquiries to business value
Suppose only eight of those twenty hypothetical inquiries meet the workshop's qualification criteria. The cost per qualified inquiry is 600 divided by eight, or 75. If two become customers, acquisition cost for that cohort is 300 before including other acquisition expenses. Keep the cohort and follow-up window consistent so incomplete sales follow-up does not distort the comparison.
For revenue campaigns, distinguish return on ad spend from profit. Revenue divided by ad spend omits delivery costs, refunds, fees, and overhead. A campaign can report attractive revenue while producing poor contribution margin. Define whether value is booked, collected, or net of refunds, and avoid assigning invented lifetime value to justify spending that current evidence does not support.
Diagnose the stage that changed
When results worsen, locate the change before editing everything. Stable impressions but fewer clicks may suggest a relevance or creative issue. Stable clicks but fewer completed inquiries may point toward the landing page, audience mix, or broken tracking. Stable inquiries but fewer qualified leads may indicate a mismatch between the promise and the offer.
Segment cautiously by device, geography, search term, placement, or audience when those views are relevant and available. Look at sample size and time lag before acting. A handful of conversions rarely supports precise conclusions about many tiny segments. Record plausible explanations and the evidence that would distinguish them rather than treating every dashboard fluctuation as a reason to rebuild the campaign.
Account for delay and attribution differences
Some people convert days after interacting with an ad. Recent periods may therefore look worse before outcomes mature. Note the reporting date, conversion window, attribution settings, and whether the report uses interaction time or conversion time. Compare like with like whenever possible.
Reconcile platform data with analytics and CRM records using clear definitions. Consent choices, cross-device behavior, modeled measurement, and attribution rules can produce differences without a simple one-to-one match. Investigate large unexplained gaps, especially duplicate events or missing purchase values. Do not choose whichever system reports the largest number and call it the truth; describe what each system observes and how it assigns credit.
Make one budget decision with a written rationale
Create a weekly decision note containing spend, qualified outcomes, economics, data-quality concerns, and the action you will take. Change one important variable when running a practical test, and specify what would count as evidence for keeping or reversing it. Budget increases should consider lead-handling capacity and outcome quality, not only cheap clicks.
Set a review date and a maximum test loss appropriate to the business. Avoid guaranteed performance claims and premature declarations of winners. Your aim is a repeatable process that links ad delivery to real business outcomes while acknowledging uncertainty. A small, well-measured campaign can teach more than a large campaign whose conversion definition changes every week.
Sources & further reading
Last editorial review: 27 October 2026. Examples are educational, not live employer data or guaranteed outcomes.