Avoid double-crediting outcomes across overlapping campaigns
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
A recipient receives an introduction, a case-study follow-up, and a reminder, then books one meeting. If each campaign claims one meeting, the reports contain three claims against a single outcome. Adding those claims does not produce a portfolio result.
The practical fix is to maintain one deduplicated outcome ledger and allocate credit from it under a published rule. Choose first-touch, last-touch, or shared credit according to the question the report must answer. Keep uncertain journeys visible rather than forcing them into a campaign.
Attribution across overlapping audiences is an accounting decision about observed journeys, not proof that a campaign caused a response. The worked example below provides a reusable reporting structure. It addresses credit assignment after recorded touches, not audience exclusions, scheduling, or integration implementation.
1. Separate outcomes from claims on outcomes
Start with two reporting layers:
- Portfolio outcomes: distinct qualifying events, counted once under an agreed definition.
- Campaign attribution: credit allocated among eligible campaigns under a named rule.
A campaign can also report how many outcomes included it somewhere in the preceding journey. Call these associated outcomes, not allocated credit. Associated counts overlap and must not be summed into a portfolio total.
Define the outcome before choosing the attribution rule. A response, a qualified response, and a meeting booking are different events. Treating them as interchangeable conversions makes reconciliation unreliable even when the credit model is consistent.
Decide whether the counting unit is an event or a person. An event-based report can count two genuinely separate bookings by one recipient. A unique-responder report counts that recipient once within the reporting period. Both reports can be useful, but they answer different questions.
Do not deduplicate solely by recipient when legitimate repeat events matter. Conversely, a status update or repeated import of the same booking should not create another outcome.
2. Write an attribution contract before calculating
An attribution contract states which records qualify and how they receive credit. Keep it beside the report so another analyst can reproduce the result.
- Outcome definition and key: what qualifies, and which stable identifier distinguishes one event from another.
- Reporting period: whether inclusion follows the outcome's occurrence date or another declared date.
- Lookback: how far before an outcome a touch remains eligible.
- Touch evidence: whether the report uses recorded delivery, click, reply, or another specified event.
- Credit unit: whether shared credit is divided among distinct campaigns or individual touches.
- Exceptions: treatment of missing timestamps, uncertain identities, and outcomes without eligible campaigns.
- Revisions: when late records can change published allocations.
Be precise about evidence. Recorded delivery is not proof of attention. A prepared campaign or approved template is not a delivered touch. For WhatsApp, keep campaign preparation, template review, and actual delivery separate: template approval does not guarantee delivery. Consult the current WhatsApp Business Messaging Policy for precise messaging requirements; attribution eligibility is not permission to send.
Choose a lookback that fits the decision cycle being examined. A short window can exclude earlier introductions; a long window can credit old campaigns with little practical connection to the outcome. Neither establishes causality. Apply the same contract to campaigns being compared and document justified exceptions.
3. Apply three rules to the same fictional journeys
Hypothetical worked example: All recipients, campaigns, dates, counts, and allocations below are fictional. The outcome is a distinct qualified meeting booking occurring during a hypothetical June reporting period. Eligible touches are recorded campaign deliveries within the preceding seven days. Delivery is an observable proxy for possible exposure, not confirmed attention.
Campaign A is an introduction, B a case study, and C a reminder. Shared credit is divided equally among distinct eligible campaigns, so repeat deliveries from one campaign do not increase its share. All timestamps use one reporting time zone.
Each row is a canonical outcome record. Preserve both the journey evidence and the resulting allocations when adapting this filled ledger.
| Recipient / outcome | Recorded journey and booking | First-touch credit | Last-touch credit | Shared credit |
|---|---|---|---|---|
| R01 / O01 | A: June 2, 09:00 B: June 4, 10:00 C: June 6, 11:00 Booking: June 6, 12:00 | A: 1 | C: 1 | A: 1/3 B: 1/3 C: 1/3 |
| R02 / O02 | A: June 3, 09:00 B: June 5, 10:00 Booking: June 5, 15:00 | A: 1 | B: 1 | A: 1/2 B: 1/2 |
| R03 / O03 | B: June 7, 09:00 C: June 8, 10:00 Booking: June 8, 16:00 | B: 1 | C: 1 | B: 1/2 C: 1/2 |
| R04 / O04 | C: June 9, 09:00 Booking: June 9, 14:00 | C: 1 | C: 1 | C: 1 |
| R05 / O05 | A: June 1, 09:00 Booking: June 12, 14:00 A falls outside the seven-day lookback. | Unassigned: 1 | Unassigned: 1 | Unassigned: 1 |
| R06 / O06 | A and B: June 13, times missing Booking: June 14, time unspecified Both deliveries confirmed before booking; their order is unknown. | Ambiguous: 1 | Ambiguous: 1 | A: 1/2 B: 1/2 |
| R07 / O07 | A: June 15, 09:00 A: June 16, 09:00 C: June 17, 10:00 Booking: June 17, 14:00 | A: 1 | C: 1 | A: 1/2 C: 1/2 |
R06 shows why ambiguity is model-specific. Both campaigns precede the booking, but their order is unknown. Equal shared credit is calculable without inventing an order; first-touch and last-touch are not. If it were unclear whether either delivery preceded the booking, shared credit would also remain unresolved.
R07 illustrates another choice. Dividing by individual deliveries would give A a larger share because it appears twice. This contract deliberately divides by distinct campaigns, avoiding an automatic credit advantage from additional deliveries.
4. Reconcile campaign credit to the portfolio total
The hypothetical ledger contains seven distinct bookings from seven recipients. Event count and unique-recipient count happen to match here; they need not match in another period.
| Campaign or bucket | First-touch | Last-touch | Shared credit |
|---|---|---|---|
| A | 3 | 0 | 11/6 |
| B | 1 | 1 | 11/6 |
| C | 1 | 4 | 7/3 |
| Allocated campaign subtotal | 5 | 5 | 6 |
| Unassigned: no eligible touch | 1 | 1 | 1 |
| Ambiguous order | 1 | 1 | 0 |
| Portfolio total, including exceptions | 7 | 7 | 7 |
B receives 1/3 from O01 and 1/2 each from O02, O03, and O06: 1/3 + 1/2 + 1/2 + 1/2 = 11/6. The exact campaign fractions sum to six allocated outcomes: 11/6 + 11/6 + 7/3 = 6. One additional unassigned outcome brings the portfolio total to seven.
Retain exact fractions or sufficient calculation precision, and round only for display. Rounding each one-third contribution before aggregation can introduce apparent gaps. If displayed decimals do not sum exactly, explain the rounding rather than changing the underlying allocation.
The same ledger produces four associated outcomes for A, four for B, and four for C, including O06 for both A and B. Those twelve associations relate to only six bookings with eligible campaigns. O05 remains in the portfolio total without campaign credit.
Show portfolio outcomes, allocated credit, and exception buckets separately. If associated counts appear, label them non-additive. Their sum is not a total-outcomes measure.
5. Choose the rule for the reporting purpose
Use first-touch to identify the earliest campaign in the observed eligible journey. This supports introduction-stage reviews. Its limitation is that the first recorded touch may not be the recipient's first actual encounter. Later campaigns receive no credit even when they closely precede the response.
Use last-touch to identify the campaign immediately preceding the outcome. This supports closing-stage reviews, but can make reminders look dominant while hiding earlier communications. In the hypothetical example, C receives four last-touch credits but only one first-touch credit.
Use equal shared credit for a transparent allocation across participating campaigns. It recognises multiple eligible campaigns without estimating their individual causal contributions. Equal participation is not equal influence: a lightly noticed delivery receives the same share as a message that prompted substantive engagement.
For a recurring portfolio review, use shared credit as the primary allocation when there is no defensible basis for unequal weights. Keep first-touch and last-touch as diagnostic views. For an introduction or closing-stage review, make the corresponding single-touch rule primary while retaining reconciliation to the portfolio total.
Do not average the models merely to soften disagreement. Their differences reveal journey structure. None establishes incremental impact or proves that removing a campaign would remove its attributed outcomes.
6. Expose ambiguity instead of manufacturing precision
Maintain an exception ledger with the outcome key, issue, affected rule, owner, and next review date. Separate recoverable data gaps from limits in what was observed.
- Missing order: retain an ambiguous bucket for single-touch rules. Do not silently break ties alphabetically or by import order.
- Uncertain identity: hold allocation when a touch cannot confidently be connected to the outcome recipient. Similar-looking fields alone are not enough to merge people.
- Duplicate outcome: retain one canonical event and record which duplicate records were suppressed. Preserve genuine repeat events when the contract permits them.
- No eligible touch: keep the outcome in the portfolio total without campaign credit. This does not prove it was organic.
- Incomplete history: label the journey as partial instead of describing its earliest recorded touch as a proven first encounter.
For hypothetical O06, the exception entry would state: delivery order missing; first-touch and last-touch affected; reporting owner to verify original timestamps before the next report. Shared credit remains available because pre-booking eligibility is confirmed.
If a timestamp is later verified, record the evidence and allocation revision. If it cannot be recovered, retain the uncertainty. Explained exceptions are more useful than complete-looking credit built on guesses.
7. Make the report reproducible and testable
Keep the original outcome records, eligible-touch ledger, contract version, and resulting allocations together. This is a recommended reporting design, not a claim that a particular product supplies these controls automatically.
Before publishing, run these checks:
- Outcome uniqueness: each counted event has one canonical key.
- Temporal eligibility: every credited touch precedes its outcome and falls within the declared lookback.
- Per-outcome balance: each fully allocated outcome contributes exactly one unit across campaigns.
- Portfolio balance: campaign credit plus unassigned and ambiguous buckets equals the deduplicated outcome count.
- Exception visibility: uncertain records remain visible rather than disappearing through filtering.
Test sensitivity without silently changing the headline rule. Recalculate a sample with a shorter or longer lookback and inspect which allocations move. Large shifts show dependence on the chosen window; they do not establish which window is true.
For the next reporting cycle, copy the ledger structure and apply it to a manageable batch of outcomes. Publish the distinct outcome count, allocated credit, exception counts, reconciliation difference, and corrected duplicate count. The immediate success criterion is zero reconciliation difference with every exception explained—not an increase in attributed conversions.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications