Separate new customers from active customers
A new customer makes their first qualifying purchase during the reporting period. An active customer meets a chosen activity definition during that period. A returning customer has a prior qualifying purchase. These groups answer different questions, so they should not share an ambiguous “customers” label in your reporting process.
The built-in MetricBoard counter sums new customers. Supply mutually exclusive period counts from your source system so that one person is not counted as new twice. The tool receives aggregate figures, not individual identities, and cannot deduplicate customer records. A total active-customer figure belongs in a separately labelled custom snapshot.
Match the conversion numerator and denominator
MetricBoard calculates converted visits divided by eligible visits, multiplied by 100. A converted visit is a visit in which the selected goal happened at least once. Counting every order as a separate conversion can exceed the number of visits when one visit produces several orders, so that is a different metric.
Define the goal, the eligible population, the reporting timezone and any excluded traffic before exporting your counts. Internal visits and automated traffic can distort the denominator. A rate based on qualified enquiries is not directly comparable with a rate based on all website visits, even if both are called conversion.
Aggregate counts, not percentages
Suppose one illustrative period has 10 converted visits from 100 eligible visits, and another has 20 from 1,000. Their rates are 10% and 2%. Averaging those percentages produces 6%, but the combined rate is 30 divided by 1,100, approximately 2.73%.
The combined calculation gives each eligible visit equal weight. MetricBoard follows this approach by summing the numerator and denominator before calculating the rate. Import the underlying counts rather than precomputed percentages. If there are no eligible visits, the rate is unavailable; displaying 0% would incorrectly imply that people had an opportunity to convert and none did.
Distinguish percentage points from growth
A move from 2% conversion to 3% conversion is a one-percentage-point increase. Relative to the original 2%, it is also a 50% increase. Both statements are mathematically valid, but they express different things. State the starting and ending rates to prevent an impressive relative change from hiding a small absolute movement.
Small denominators produce unstable rates. One additional conversion among a handful of visits can create a large swing without establishing a reliable trend. Review counts alongside percentages and avoid describing every monthly change as evidence that a campaign or design caused the result.
Respect timing and channel differences
Customers may research in one period and buy in another. A same-month visit-to-purchase comparison can therefore misrepresent a long decision cycle. Cohort analysis follows a defined group over time and is better suited to that question, but it requires more detailed source data than this aggregate dashboard stores.
Channel mix also matters. An increase in early-stage traffic may lower the overall conversion rate while creating more eventual customers. Compare like-for-like audiences in your source reports before changing your acquisition plan. MetricBoard has no built-in channel segmentation or attribution model; use separate prepared datasets when you need a focused view.
Make the review useful and privacy-conscious
Bring three questions to a customer review: are we reaching the intended audience, are eligible visitors completing the intended action, and are those actions producing valuable customer relationships? Pair the acquisition view with retention or repeat purchasing where reliable figures exist. More new customers is not always better if acquisition costs or service demands rise faster.
Keep names, email addresses and customer identifiers out of dashboard imports. Aggregate period totals are sufficient for these calculations. Store the source records in your authorised business systems, retain only the exports you need and remember that a shared browser profile can expose locally saved dashboard data.