Draw the funnel in separate stages
Write the path a person could take: see a platform post, visit the profile, open the website, use a free resource, view the paid offer, open checkout and complete an order. Each stage answers a different question. Platform views describe a platform-defined exposure event. Measured website sessions describe recorded browser activity. A resource download event indicates that a link was used, subject to your implementation; it does not automatically mean a unique person opened the file. A checkout click indicates interest. A verified payment-provider order is evidence of a sale. Your scorecard needs separate labels and sources for these stages. Leave a missing stage unavailable rather than filling it with another count. This makes the first diagnosis possible: weak exposure, an unclear profile destination, a weak offer or a checkout problem require different work.
Give each link a consistent source
Tag links you control with a small naming convention. A profile link can use utm_source=instagram, utm_medium=organic_social and utm_campaign=profile. Use a different source value for TikTok, Facebook, YouTube and X. A particular content placement can add a stable utm_content value such as caption-check-01 when the platform and workflow support that link. Keep names lowercase and use the same spelling over time. Do not put a person's email address, name or account identifier into a campaign tag. Open the exact tagged URL on mobile before using it publicly. Check that it reaches the right page, the resource works and the measurement implementation records only the intended fields. A label is a clue about an inbound route, not proof that the tagged post caused every later purchase. Copies, shared links and return visits can complicate attribution.
Write the metric definition beside the number
Record the platform, post URL, publication time, observation time and elapsed hours. Copy the metric's actual label and meaning from the authorized account analytics. Different platforms can count a view differently, and a platform may change its definitions. Do not add unlike counts into a total called “people reached.” For website data, document whether the count represents events, sessions or distinct visitors and how it is deduplicated. If measurement depends on consent, call the result measured sessions rather than all visitors. Ad blockers, failed requests and browser navigation can also remove events. Keep a blank cell for not yet measured and use zero only after checking and observing zero. Store a provider export or report reference where practical. This small amount of bookkeeping prevents a confident-looking dashboard from comparing numbers that do not mean the same thing.
Calculate rates only from compatible counts
A rate needs a defined numerator and denominator drawn from a compatible window and population. For an illustrative example, suppose a properly deduplicated report recorded 80 measured sessions to the offer page and 8 of those sessions reached checkout during the same window. The observed checkout-session rate would be 8 divided by 80, or 10 percent. These are teaching numbers, not Signal Studio Lab results. They do not tell you the paid conversion rate because no verified order count was supplied. If your tracking only records checkout click events, repeated clicks may inflate the numerator; label the event count instead of calling it unique checkout sessions. Small samples also move sharply when one person acts. Compare related posts at similar ages and record changes in audience, offer, price and timing. A directional clue can justify another test without supporting a precise growth forecast.
Take money from the payment record
Reconcile paid orders through the payment provider, using its confirmed order state rather than the browser's thank-you page alone. The provider record should supply the order ID, amount, currency and payment status. Record refunds and adjustments separately. Keep taxes collected, provider fees, ad spend and other operating costs separate from gross sales. Never combine USD and CAD amounts without naming the conversion source and date. In another purely illustrative example, three confirmed US$29 orders total US$87 in gross sales before refunds, fees, taxes and costs; they do not establish US$87 of profit. A dashboard without the payment connection should display sales as unavailable. Test delivery as well as payment before promoting a product. A paid order creates an obligation to supply what was actually described on the offer page.
Use the review to choose a controlled next action
Start the weekly review with one question, such as whether people who reach the toolkit page can understand its contents and next step. Inspect the relevant stage before changing everything. If measured visits exist but offer engagement is weak, review the preview, contents, price and mobile usability. If checkout opens but verified orders do not follow, investigate the actual checkout and delivery flow before buying more traffic. For a creative comparison, record one variable, the comparison, observation window and decision rule before publishing. Organic posts usually reach different audiences at different times, so call a small difference directional rather than causal. Keep an internal revenue target as a planning constraint, separate from observed revenue. The useful output is a grounded decision: fix a broken link, clarify an offer, repeat a lesson or pause an unproductive spend. More dashboard numbers cannot replace that decision.