Define the count first

Video transcript

1. 0:00–0:06

   Choose the decision • Define the goal event and report population. • Decision · continue this flow? • Confirmed order • Order-system evidence • Send-date cohort • Outcome-date cohort • Choose the population, timezone and observation window

   A decision about continuing a flow highlights before a confirmed-order event connects to its order-system evidence. Two population choices remain separate: message send date and outcome date. A send-date cohort may collect later outcomes within an attribution window, while an order-date cohort answers another question. Save timezone, population, event definition and observation window. Keep campaigns, automated flows and transactional mail identifiable; different audiences and purposes prevent a higher rate alone from proving better design.

2. 0:06–0:12

   Name the denominator • Unique clickers and total clicks are different counts. • C17 • C18 • 3 click records → 2 distinct recipients • 49 unique clickers • 980 delivered • 5% recorded rate • Separate invented examples · recorded clicks, with stated bot filtering

   In a separate invented mini-example, three native click tokens C17, C17 and C18 move into two recipient positions. The repeated C17 merges and disappears as a duplicate, leaving C17 and C18. The article figure then independently shows 49 unique recorded clickers divided by 980 delivered recipients equals 5%. Record the links and bot-filter settings; recorded clicks do not prove human intent. The article campaign attempts 1000 sends, has 980 accepted and 20 finally unaccepted messages, assuming one message per recipient. These are invented calculations, not benchmarks or results. Opens do not establish reading.

3. 0:12–0:18

   Save attribution settings • Keep the model, window and report snapshot. • Invented rule · click-only · 7 days • Eligible click • Credited order in window • Save model + settings + report timestamp • Recent cohort? Label provisional until the agreed window matures

   The article exercise explicitly chooses click-only attribution over seven days, not a provider default. An eligible-click record connects to a credited order only under that saved event, window and model. Save settings and report timestamp. Recent cohorts remain provisional until the agreed observation window matures; compare cohorts at the same stage. Changing a window or definition changes the reporting rule rather than demonstrating improved customer behaviour. Campaign tags alone do not align different reporting scopes.

4. 0:18–0:24

   Reconcile refunds • The example nets SAR 2,520 in attributed merchandise revenue. • SAR 2,940 • SAR 420 refunds • SAR 2,520 net • 20 original purchasers / orders • 17 retained orders after 3 full refunds • Invented merchandise revenue · after discounts; excludes tax + shipping · not profit

   The reviewed ledger shows SAR 2940 of merchandise revenue after discounts, excluding tax and shipping, less SAR 420 from three fully refunded orders, giving SAR 2520 net attributed merchandise revenue at a later agreed refund snapshot. Twenty original distinct purchasers and their twenty orders stay separate from seventeen retained orders. The example does not assume every dashboard treats refunds identically. Costs are absent, so it does not calculate profit or return on investment. Every number is invented for this exercise, not a client or TecnoStack outcome.

5. 0:24–0:30

   Keep the causal question open • Attributed revenue does not prove additional sales. • Attributed ≠ incremental • Allocation under saved rules • What happens without the email? • Separate comparison + uncertainty

   The reviewed figure states attributed revenue is not incremental impact. Native panels separate allocation under saved rules from the unanswered question of what would happen without email. Only the latter connects to a separate suitable comparison design and explicit uncertainty. Match event, dates, timezone, reporting scope and attribution assumptions when reconciling tools, using the order system for order facts. Do not force dashboards to agree by changing definitions. Record one proposed change, observable effect and review date; do not promise a fixed return.