Linear Attribution: The Neutral Baseline Model

What Linear Attribution Measures

Linear attribution answers: "How should I credit all the marketing touchpoints that contributed to this conversion?"

Its answer: give everyone equal credit.

Four touchpoints, 25% each. Ten touchpoints, 10% each. The math is straightforward, and that's the point.

Why Linear Matters

Linear attribution makes no assumptions about which touchpoints are more important. Compare this to:

Linear's "bias" is assuming equal importance—which may or may not match reality. But it's the only model that doesn't bake in a predetermined hierarchy.

When Linear Attribution Is the Right Choice

1. Your First Multi-Touch Model

If you're moving from last-touch or first-touch to multi-touch, start with linear:

The shift reveals what last-touch was hiding: Paid Social introduces far more customers than it closes. Linear exposes the full journey.

2. Long Consideration Journeys

For products with extended research phases—B2B software, high-ticket items, considered purchases—every touchpoint genuinely contributes:

• Initial ad creates awareness
• Content builds understanding
• Webinar demonstrates value
• Case study provides proof
• Email answers objections
• Demo enables evaluation

Giving 100% to any single touch ignores the others' contribution. Linear acknowledges the entire journey.

3. When You're Unsure What Matters Most

If you don't have strong hypotheses about which touchpoints drive conversion, linear is the honest choice. It says: "We don't know the relative importance, so we'll treat them equally until we learn more."

This is intellectually honest. Better to be neutrally wrong than confidently biased.

4. Baseline for Comparison

Run linear alongside other models to understand how each model's biases affect your data:

Linear becomes your neutral reference point.

When Linear Is the Wrong Choice

1. Single-Session Conversions

If most conversions happen in one session with one touchpoint, linear adds complexity without value:

For impulse purchases or simple conversions, the overhead of multi-touch isn't worth it.

2. When Touchpoints Clearly Differ in Impact

Some touchpoints genuinely matter more:

• A 30-minute demo vs a banner impression
• A personalized email vs a generic newsletter
• A sales call vs an automated drip

Linear treats these equally, which may misallocate credit. If you have strong evidence about touchpoint impact, use a weighted model.

3. High-Volume Data-Driven Scenarios

If you have 5,000+ monthly conversions, algorithmic models (Markov, Shapley) can learn actual touchpoint impact from data rather than assuming equal contribution.

How Linear Attribution Works

The Math

Linear attribution divides credit equally:

Credit per touchpoint = 1 / Number of touchpoints

For a journey with n touchpoints:

Implementation

class LinearAttribution
  def initialize(lookback_days: 30)
    @lookback_days = lookback_days
  end

  def attribute(conversion)
    touchpoints = conversion.user.touchpoints
      .where("occurred_at >= ?", conversion.occurred_at - @lookback_days.days)
      .where("occurred_at <= ?", conversion.occurred_at)
      .order(:occurred_at)

    return [] if touchpoints.empty?

    credit_per_touch = 1.0 / touchpoints.count

    touchpoints.map do |touchpoint|
      {
        channel: touchpoint.channel,
        source: touchpoint.source,
        medium: touchpoint.medium,
        campaign: touchpoint.campaign,
        credit: credit_per_touch,
        touchpoint_at: touchpoint.occurred_at
      }
    end
  end
end

Aggregating to Channels

To get channel-level credit:

SELECT
  channel,
  SUM(credit) as total_credit,
  SUM(credit * conversion_value) as attributed_revenue,
  COUNT(DISTINCT conversion_id) as assisted_conversions
FROM attribution_results
WHERE model = 'linear'
GROUP BY channel
ORDER BY attributed_revenue DESC;

Key Implementation Decisions

Linear vs Other Models

Linear vs Last-Touch

Verdict: Linear is better for budget decisions. Last-touch remains useful for conversion optimization specifically.

Linear vs First-Touch

Verdict: Use both. First-touch answers "where do leads come from?" Linear answers "what's the full contribution?"

Linear vs Time-Decay

Verdict: Time-decay makes sense when recent touches genuinely matter more (e.commerce, urgency products). Linear is more neutral for considered purchases.

Linear vs Position-Based

Verdict: Position-based is better when you have strong reason to believe introduction and conversion are the key moments. Linear is better when the middle of the journey matters.

Common Linear Attribution Mistakes

Mistake 1: Treating Linear as "Truth"

Linear is not the correct answer—it's a neutral starting point. All attribution models are simplifications.

Fix: Use linear as a baseline, then validate with incrementality tests.

Mistake 2: Too Many Low-Quality Touchpoints

If you include every page view, email open, and ad impression, you dilute credit across noise:

Fix: Define what counts as a touchpoint. Usually: clicks, meaningful engagements, not passive impressions.

Mistake 3: Not Accounting for Journey Length

Conversions with 2 touchpoints vs 10 touchpoints get the same total credit (100%), but distributed very differently:

Fix: Segment analysis by journey length. Compare channels within similar journey lengths.

Mistake 4: Ignoring Conversion Value

Linear credits all touchpoints equally, but if conversion values vary, weight by value:

# Credit weighted by conversion value
attributed_value = credit_per_touch * conversion.value

# Channel gets sum of attributed values
channel_revenue = attributions.sum { |a| a[:credit] * a[:conversion_value] }

Linear Attribution and GA4

Google Analytics 4 removed linear attribution in 2023. Your options:

1. Use a Third-Party Tool

Attribution platforms like mbuzz support linear and other models out of the box.

2. Build in BigQuery

Export GA4 data to BigQuery and calculate linear attribution:

WITH touchpoints AS (
  SELECT
    user_pseudo_id,
    event_timestamp,
    traffic_source.source,
    traffic_source.medium
  FROM `project.dataset.events_*`
  WHERE event_name IN ('page_view', 'click', ...)
),
conversions AS (
  SELECT
    user_pseudo_id,
    event_timestamp as conversion_time,
    event_params.value.double_value as value
  FROM `project.dataset.events_*`
  WHERE event_name = 'purchase'
),
journeys AS (
  SELECT
    c.user_pseudo_id,
    c.conversion_time,
    c.value,
    t.source,
    t.medium,
    COUNT(*) OVER (PARTITION BY c.user_pseudo_id, c.conversion_time) as touchpoint_count
  FROM conversions c
  JOIN touchpoints t ON c.user_pseudo_id = t.user_pseudo_id
    AND t.event_timestamp < c.conversion_time
    AND t.event_timestamp > TIMESTAMP_SUB(c.conversion_time, INTERVAL 30 DAY)
)
SELECT
  source,
  medium,
  SUM(value / touchpoint_count) as linear_attributed_revenue
FROM journeys
GROUP BY source, medium
ORDER BY linear_attributed_revenue DESC;

3. Use Segment or CDPs

Customer data platforms often include attribution modeling as a feature.

Validating Linear Attribution

Compare to Other Models

If linear produces wildly different results from time-decay or position-based, investigate why:

Large gaps suggest the channel has a specific role (introducer vs closer).

Validate with Incrementality

Linear might over- or under-credit certain channels. Test with holdout experiments:

1. Pause a channel in test markets
2. Measure actual revenue impact
3. Compare to linear-attributed revenue
4. Calculate calibration factor

Linear attributed: $100K/month to Paid Social
Incrementality test: $80K actual lift
Calibration factor: 0.8x

Apply calibration to adjust linear attribution toward truth.

Implementing Linear Attribution in mbuzz

mbuzz uses AML (Attribution Modeling Language) — a small Ruby DSL — to define how credit is distributed. Linear is the simplest model in the language.

Basic linear

within_window 30.days
  apply 1.0 / touchpoints.length to touchpoints
end

One unit of credit, split equally across every touchpoint in the 30-day lookback. Change 30.days to lengthen or shorten the window.

Linear with channel filtering

To exclude direct and branded organic from the credit pool, filter touchpoints first:

within_window 30.days
  marketing = touchpoints.reject { |tp| %w[direct organic_brand].include?(tp.channel) }
  apply 1.0 / marketing.length to marketing
end

The full AML reference — including segment weights, conditional logic, and time-decay primitives — is at docs/attribution-models.

Tuning Linear Parameters for Your Business

Linear attribution has few parameters, but the ones it has matter. Here's how to tune them for different contexts.

By Business Type

In AML, the lookback is the within_window duration. To exclude direct, filter it out of touchpoints before the apply call (see channel-filtering example above).

Seasonal Adjustments

Buying behavior compresses during sale periods. Shorten the AML window from 30.days to 14.days for BFCM and the holiday peak; stretch it to 45.days post-holiday to credit gift card redemptions back to the original campaigns. Run the seasonal variant alongside the standard one and compare the credit split.

Big Campaign / Launch Adjustments

For a launch attribution view, narrow the lookback and restrict touchpoints to the launch campaigns:

within_window 14.days
  launch = touchpoints.select { |tp| tp.campaign.starts_with?("product-launch-2024") }
  apply 1.0 / launch.length to launch
end

Parameter Tuning Cheatsheet

Summary

Linear attribution distributes credit equally across all touchpoints. It's simple, transparent, and unbiased—making it the ideal baseline for multi-touch attribution.

Use linear when:
- Moving from single-touch to multi-touch attribution
- You want a neutral, unbiased view of contribution
- Journey touchpoints are roughly equal in importance
- You need a baseline to compare other models against

Don't rely solely on linear because:
- Some touchpoints genuinely matter more than others
- It may dilute credit across low-impact touches
- It doesn't capture recency or position effects

Best practice: Start with linear, understand your baseline, then explore time-decay or position-based if you have strong hypotheses about touchpoint importance. Validate any model with incrementality tests.

Further Reading

On Attribution Models:
- First-Touch Attribution — Credits the introduction
- Last-Touch Attribution — Credits the close
- Time-Decay Attribution — Credits recency
- How to Choose the Right Attribution Model — Decision framework

On Validation:
- MTA vs MMM: What's the Difference? — Where attribution fits in measurement
- Triangulating MTA, MMM, and Incrementality — Validating with multiple methods

Holly Henderson

Co-Founder, mbuzz

Holly Henderson is Co-Founder of mbuzz. With 10+ years in marketing including roles at Westpac, Avon, and Forebrite, she's obsessed with making measurement actually useful.

Harvard Extension School Forebrite Westpac Avon