The quick answer
Measure a campaign as a journey: eligible audience, attempted sends, delivery, observed reads, replies and verified business outcomes. Keep unknown results visible, define attribution rules and create follow-up audiences from relevance and permission rather than from silence alone.
A campaign with a large send count can still create very little value. Some messages may be waiting, some may fail and some may reach customers who need a human answer before they can buy.
Good analytics helps the team decide what to do next. It should explain the audience, the actual message outcomes, the response workload and the evidence behind any claimed sales. Begin with those questions rather than a dashboard full of percentages.
Keep the audience and the sending journey separate
Record the selected audience before launch, then explain exclusions: duplicates, invalid contact details, missing permission, opted-out customers or records that no longer meet the segment. The audience requested by the business and the recipients eligible to send are not always the same number.
Track queued, attempted, accepted, sent, delivered, read and failed states as distinct events where the connected systems provide them. Meta’s message status notifications distinguish sending, delivery and reading; a request accepted by software is not proof that a customer received it.
Avoid adding overlapping totals together. A recipient who has read a message has also progressed through delivery, so “delivered plus read” is not a unique audience. Label whether a figure is a cumulative milestone or a current-state count.
Give every rate an explicit denominator
A percentage without a denominator is easy to misunderstand. A reply rate based on delivered messages answers a different question from a reply rate based on the original uploaded audience. Choose the definition and display it consistently.
Keep unknown results separate from failures. If a delivery update has not arrived, report that uncertainty. Likewise, the absence of an observed read should not be presented as proof that the person ignored the message.
| Metric | Example definition | What it helps you decide |
|---|---|---|
| Eligibility rate | Eligible unique recipients ÷ selected unique audience | Whether audience preparation is healthy |
| Delivery rate | Delivered unique recipients ÷ attempted unique recipients | Whether messages are reaching the audience |
| Observed read rate | Observed readers ÷ delivered recipients | How much reading evidence is available |
| Reply rate | Attributed unique repliers ÷ delivered recipients | Whether the conversation is starting |
| Matched outcome rate | Verified matched outcomes ÷ delivered recipients | How the journey relates to a business result |
Explain which replies belong to the campaign
A customer may receive a campaign and later write about an unrelated order. Counting every later incoming message as a campaign reply can make results look much better than they are.
Prefer explicit evidence where available: a response to the specific message, a campaign-specific button or a tracked journey reference. If the system uses a time-based association, name the attribution window and explain what happens when several campaigns overlap.
Count unique replying contacts separately from total incoming messages. One customer sending six short replies is one engaged person and six pieces of conversation work. Both figures can be useful, but they answer different questions.
Connect outcomes without overstating causation
Choose the outcome before the campaign: a consultation booked, a requested quotation, a completed order or a support issue resolved. Use the relevant booking, deal or store record to verify it. A button click is an action, not automatically a sale.
State the attribution rules, date range and matching method. If the same purchase could be associated with two campaigns, avoid presenting it as two independent sales. Where possible, show matched revenue alongside refunds or cancellations rather than only the largest gross figure.
A sale after a message does not prove the message caused the purchase. To estimate incremental effect, use a carefully chosen comparison or holdout group and an appropriate analysis. For everyday reporting, honest matched outcomes are more useful than a confident but unsupported uplift claim.
Build follow-up audiences around the next useful action
A customer who asks about delivery needs an answer, not another promotional broadcast. Someone who completes the purchase should leave the recovery audience. Someone who asks to stop should be excluded before another scheduled send.
Use results to prepare a proposed audience, then recheck current contact preferences, recent messages, purchases and active conversations before launch. Campaign data is a useful clue about relevance; it is not a new permission to contact the person.
| Observed result | Useful next action | Exclude or review |
|---|---|---|
| Asked a product question | Assign a relevant agent response | Do not repeat the same generic offer |
| Booked an appointment | Move to the booking journey | Exclude from acquisition reminders |
| Delivery failure | Review the reason and data quality | Do not retry every failure blindly |
| No observed reply | Review timing, relevance and permission | Do not assume lack of interest or consent |
| Opted out | Stop further outreach | Exclude from every relevant future audience |
Review costs, quality and workload together
Separate Meta messaging fees, Wamigos platform usage and optional AI or connected-service costs. A campaign can have strong response numbers but require more staff time than the team can handle. Plan ownership and handover before sending to a large audience.
For a Wamigos campaign review, ask to see the recipient outcomes, attribution method, wallet entries and available export together. Confirm the reporting and audience controls enabled in your workspace. A simulated volume test proves software behaviour under that test; it does not guarantee real-world delivery to every contact.
After the review, change one meaningful thing: audience, timing, content or next step. Keep a short note about the hypothesis and compare similar campaigns. Continuous learning comes from clear decisions, not from collecting more charts than anyone can use.
Common questions
Is a successful send request the same as delivery?
No. Track the subsequent delivery outcome. Software acceptance and customer receipt are different milestones.
Can we see everyone who has read a campaign?
Use only the read information actually available from the connected platform. An absent read event should remain unknown rather than being labelled definitely unread.
Should every later customer message count as a campaign reply?
No. Define an attribution method and time window. Prefer explicit message or journey references and handle overlapping campaigns consistently.
Can we retarget everyone who did not reply?
First review relevance, current preferences, recent purchases and active conversations. No reply is not fresh messaging permission and does not automatically justify another message.
Does campaign revenue prove return on investment?
Matched revenue is useful evidence, but account for costs, refunds, attribution limits and purchases that might have happened anyway. Label the calculation clearly.
Sources & further reading
Primary documentation reviewed 16 November 2026. Provider rules, pricing and eligibility can change. Check the linked source before making account or billing changes.






