Landing page software attribution methods compared on one event and window

Landing page software attribution methods allocate credit according to recorded rules. They do not, by themselves, show that a page or channel caused a transaction. Compare methods against one fixed event, eligible population, unit, period and data boundary, then keep causal evaluation in a separate field.

This desk comparison was researched on 6 September 2026 for an England operator. No product account, journey or conversion record was tested. It uses blank values and has no overall winner.

What to take away

The fixed comparison case: one accepted transaction, September 2026

Define one outcome event from the finance or customer system, such as an accepted transaction under a named status definition. Fix its eligible population, currency or count unit, event date, Europe/London reporting period and duplicate rule. Use one declared lookback window and identical available touchpoint records for every attribution method.

The worked case uses an England operator, 1 to 30 September 2026, a single 30-day lookback and one set of touchpoint records. The page builders in play are Unbounce, Instapage and Leadpages. HubSpot captures form source on its free tier.

Worked example, illustrative numbers. One accepted order on 18 September 2026 is worth £240. Its recorded touchpoints are a paid search click on 2 September, an organic visit on 9 September, an email click on 14 September and a direct visit on 18 September.

Typical vendor list prices for the tools named here are below. Annual billing usually lowers them.

Also record privacy and technical gaps. A path may be incomplete because a visitor objects, a tag does not run, devices cannot be joined or an offline outcome lacks an identifier. Keep an unattributed category rather than forcing credit.

Direct response source: HubSpot, Leadpages and Unbounce forms

This method assigns the outcome to a source captured explicitly with the form or transaction, such as a declared referral code. HubSpot stores the original source and hidden fields on a free tier, and Leadpages and Unbounce forms can carry the same fields.

CallRail and WhatConverts track the phone source on paid plans; their starting prices sit in the table above.

Its rule is easy to audit when the source field and eligible values are stable. It can miss prior or parallel influences, and self-reported or copied codes may be wrong.

Use the same outcome unit as the other methods. Report missing and invalid source values separately. Audit the source field itself, because a broken field reads as a channel with no demand.

Last eligible touch: GA4 last non-direct and Google Ads last click

Credit goes to the final recorded eligible interaction before the outcome within the fixed window. Google Analytics 4 reports last non-direct click inside the free property. Google Ads records its own last-click credit and cannot see paths that never touched a Google ad. Adobe Analytics offers a last-touch model in its report suite.

The rule is reproducible, but it ignores earlier eligible touches and depends on what the system can observe. Direct visits, repeat devices and channel exclusions can change the result.

Google Ads' current first-party attribution-model documentation describes its supported model choices and how credit enters product reporting. Those rules are global product records. They do not establish independent performance or suitability for an England buyer.

Buyer-authored fractional rules: Adobe Analytics custom attribution

A buyer may assign fractions across declared positions or touch types. Adobe Analytics and Customer Journey Analytics support a custom model with user-set weights and a chosen lookback window.

Google Analytics 4 does not expose user-defined fractional weights; it offers rules-based and data-driven models instead.

The shares for each outcome must use one unit and reconcile to the agreed total. Publish the weights, eligibility, order rule and treatment of repeated touches.

This method can show sensitivity to a stated policy, but the weights are assumptions unless backed by a separate design. Run the fractional model against a second period before relying on the weights.

Platform data-driven attribution: GA4, Google Ads and Meta

A platform can calculate credit from the data and model available within its product. Google's Analytics attribution guide says its data-driven model is specific to an advertiser and key event and records channel and lookback settings. Google Ads and Meta Ads Manager run comparable models on their own data.

Triple Whale and Northbeam apply similar models to direct-to-consumer stores. Treat every result as platform-attributed credit. Store the product, property, event and model. Also record the documentation date, eligible channels, window and reporting-date basis.

Opaque or changing model behaviour can limit comparability over time. Multiple platforms may each claim credit for the same outcome because they observe different paths. Never add their credited conversions as if they were unique transactions.

Controlled incrementality

Incrementality compares observed outcomes with a credible estimate of what would have happened without the intervention. It requires a pre-specified design and a declared comparison population. Add the assignment method, analysis plan, deviations, missing outcomes and uncertainty.

Incrementality design sketch, illustrative. The assignment unit is the signed-in user, randomised on the first eligible visit. Hold back 10% from paid search for six weeks, from 22 September to 1 November 2026, with the two weeks before that as a pre-period.

The primary metric is accepted transactions per assigned user, reported with a 95% confidence interval. If user-level splitting is not possible, assign whole postcode areas instead, and allow two extra weeks for baseline stability.

HM Treasury's Quality in policy impact evaluation guidance applies to government policy evaluation. Its main distinction is useful here. Monitoring shows whether an outcome changed; a strong counterfactual is needed for a confident causal claim. A qualified evaluator must judge feasibility and proportionality.

A holdout needs enough assigned units to detect the effect it is meant to find.

Select the reporting label

Choose the method answering the stated reporting question with evidence the buyer can govern. Report observed outcomes once, show credit under each selected rule in separate columns, and put causal estimates in their own table.

The ICO's final analytics exception guidance shows why collection purpose and visitor controls can affect coverage. Privacy permission is a separate gate from analytical usefulness. A free tier carries the same legal duty as an enterprise contract.

Before you act

Common questions

Why is assigned credit not the same as incrementality?

Incrementality compares observed outcomes with a credible estimate of what would have happened without the intervention. That requires a pre-specified design, comparison population, assignment method and analysis plan.

What must stay fixed when comparing attribution methods?

Use one outcome event from the finance or customer system, with a named status definition. Fix its eligible population, currency or count unit, event date, Europe/London reporting period and duplicate rule. Then use one declared lookback window and identical available touchpoint records for every method compared.

In the case above that is the accepted transaction, 1 to 30 September 2026 and a 30-day window.

Can credited conversions from several platforms be added together?

No. Google Analytics 4, Google Ads and Meta each claim credit for the same outcome because they observe different paths. Treat each result as platform-attributed credit and store the product, property, event and model. Also record the documentation date, eligible channels, window and reporting-date basis. Adding their credited conversions would treat them as unique transactions.