Attribution assigns conversion credit to the customer interactions you can observe. For email and WhatsApp, start with tracked links, a shared customer identifier and a verified purchase or lead event. Compare models using the same conversion window. Attribution describes recorded paths; a controlled test is needed to estimate additional impact.

Last reviewed: October 8, 2026. This guide is for marketing teams managing email and WhatsApp journeys, not a claim that either channel always closes the sale.

Define the decision before choosing a model

Decide whether you need to explain a purchase, compare campaigns or decide which channel deserves another test. A model that answers the first question does not automatically answer the others. There is no universal number of touches a customer needs before buying.

Multi-channel attribution compares credit across channels. Multi-touch attribution allocates it across individual recorded interactions, which may occur within one channel or several. A channel report can be built from a multi-touch model; these are related views, not competing systems.

In an email and WhatsApp journey, a buyer might click an email, ask a question on WhatsApp and later purchase directly. Only some of those steps may reach your analytics system. Start your omnichannel plan by documenting what is observable and what is missing.

Which attribution model should you use?

Google’s attribution documentation distinguishes data-driven attribution, paid and organic last click, and Google paid channels last click. These are Google Analytics reporting options, not promises about CampaignHQ features.

Last click: a clear baseline with a narrow view

Last click assigns credit to the final eligible click. Google’s paid and organic last-click model ignores direct traffic unless the path is entirely direct. A WhatsApp link may receive credit when it is the last eligible click, but that does not prove the conversation caused the purchase. MoEngage’s last-touch explainer offers another description of the approach.

Linear: equal credit for the interactions you include

A custom linear model divides credit equally among eligible recorded touches. Define those touches first: counting every delivered message rewards sending volume, while counting verified clicks answers a different question. Do not treat an email open as confirmed attention. Use your email operating rules to keep the event definitions consistent.

Time decay: more weight near the conversion

A time-decay model assigns less credit to older interactions. Choose and document the decay setting based on your sales cycle, then test how much the result changes with another setting. There is no default fraction of a sales cycle that makes the answer correct.

Position-based models: explicit business assumptions

U-shaped models emphasise the first and last eligible interaction. W-shaped models also emphasise a middle milestone, often lead or opportunity creation. Choose the weights and milestone definitions openly. They express your reporting assumptions; they do not measure the true causal value of each step.

Linear, time-decay and position-based models are useful concepts for custom analysis. Google notes that these older models are no longer available in GA4 attribution reports. Do not plan an implementation around an option your tool no longer offers.

Data-driven: useful only within the data available

Google’s data-driven model uses observed converting and non-converting paths to assign credit. It cannot turn an untracked WhatsApp conversation into a known interaction. Check collection coverage, identity matching and model availability before interpreting the output. When comparing enterprise email platforms, ask vendors to demonstrate the events and exports needed for your analysis.

Build a measurement record you can reconcile

1. Pick the outcome. Use a confirmed order, a qualified lead or a booked appointment. Keep form submissions separate from sales-qualified leads.
2. Define the value. Decide whether revenue includes tax, shipping, refunds and cancellations. Use the same definition across reports.
3. Tag outbound links. Record source, medium, campaign and message variation consistently. Never put names, email addresses or phone numbers into tracking URLs.
4. Join only justified identities. Use an internal customer or order identifier where lawful and available. Do not assume two devices belong to one person.
5. Set a lookback window. State how far before the outcome an interaction can receive credit. Keep it unchanged during the comparison.
6. Deduplicate outcomes. Reconcile order IDs before adding channel totals. One order claimed by both platforms is still one order.

For offline sales, agree with sales operations on the event, timestamp and matching method before importing anything. Review your first-party measurement practices rather than expecting an import to recover every missing interaction.

Engagement is up, but sales are not: what to check

Before changing the attribution model, locate the break. Work through these checks with whoever owns analytics, checkout and sales follow-up.

• Orders are stable, but attributed revenue fell. Compare the same order records across both reports. Check tracking parameters, event collection, consent settings, reporting windows and refunds before blaming the campaign.
• Opens rose, but visits did not. Check verified clicks and landing sessions. Background image loading can inflate opens; more opens alone are not a reason to send more messages.
• Visits rose, but completed orders fell. Separate traffic quality from purchase friction. Check the offer, stock availability, delivery charges and payment failures. Compare campaigns with automated journeys rather than combining their totals.
• WhatsApp replies rose, but sales are missing. Review response ownership and the handoff to sales. Match completed orders or qualified opportunities only where a lawful, reliable identifier exists. Leave unmatched outcomes unknown.

Keep a short record of the symptom, evidence, owner and next test. If the problem is campaign ownership rather than credit allocation, use the campaign management evaluation guide. Change one suspected cause at a time; switching models cannot repair a broken checkout or an unanswered reply.

Handle the gaps without inventing precision

Email opens: Apple explains that Protect Mail Activity downloads remote content in the background. A tracking pixel can therefore load without a person reading the message. Keep opens as a diagnostic signal, not proof of engagement or buying intent. Our email marketing guide covers the broader channel context.

Private sharing: a copied link can lose its original tracking context. Label unattributed traffic as unknown instead of assigning it to WhatsApp by assumption. Even a tagged forwarded link identifies the campaign, not necessarily the original recipient.

Device changes: access across devices and matching a buyer across devices are different problems. The historical WhatsApp device report covers the former. For attribution, validate which interactions your own system can reliably join.

Case studies: WhatsApp’s campaign measurement article can suggest questions to investigate. A named brand’s result is not a general conversion forecast for your business.

Use credit to guide a test, not declare a winner

Compare last-click and your selected alternative using the same orders, audience and period. Inspect which campaigns gain or lose credit. If the total recognised revenue changes, check eligibility, window settings and deduplication before treating the difference as performance.

Then test the decision. For example, randomly assign eligible, consenting customers to email-only and email-plus-WhatsApp journeys. Keep the offer and outcome window comparable. Measure completed outcomes per assigned customer, including customers who never click. Track opt-outs and support workload alongside revenue. Choose sample size and uncertainty reporting with your analyst before launch.

Use send-time testing for a separate scheduling question, and the email engagement funnel to diagnose where customers stop progressing. Changing timing, audience and channel together makes the result harder to interpret.

For tooling research, PushEngage, MarTech and Coupler.io offer further product or practitioner material. Use product documentation to check integration requirements and practitioner material to frame your evaluation questions.

Bring your measurement requirements to the platform review

CampaignHQ provides customer engagement automation for email + WhatsApp. Its email product page identifies it as a Meta Tech Provider, with AWS supporting its infrastructure. Bring your event sources, permissions, reporting needs and export requirements to a demo. Ask which integrations and reporting functions are available for your setup. You can also visit the CampaignHQ signup page.

Frequently asked questions

Can I add revenue from email and WhatsApp dashboards?

Not safely without checking overlap. Reconcile order IDs and value definitions first, then apply one agreed credit rule.

Should I count every message as a touch?

Only if that is the explicit purpose of the model. Delivery, an observed read, a click and a reply are different events. Combining them without rules makes comparisons misleading.

What should I do with unmatched conversions?

Keep an unknown category and report its size. Investigate broken tags and identity gaps without guessing the missing source.

How often should I change the attribution model?

Change it when the reporting question or evidence warrants it. Preserve the previous definition and reprocess a comparable period before using the new report for budget decisions.

Does attributed revenue prove incremental revenue?

No. Attribution assigns credit to recorded interactions. A controlled experiment estimates outcomes that would not have happened without the tested intervention.

Written by CampaignHQ Team