Attribution vs contribution in marketing measurement: the difference, and when to use each

Attribution, defined

• Starts from the conversions and touchpoints a tracking system can observe: clicks, views, emails, visits.
• Rules: last click, first click, linear, time decay, or data-driven weights learned from paths.
• Same total: every rule divides the tracked conversions, using cookies, pixels and identifiers, often daily.
• Ad platforms each count the conversions linked to their own ads; summed, they often exceed recorded conversions.

Contribution, defined

• Starts from total revenue or orders, and estimates how much each driver produced, usually with a marketing mix model.
• Drivers: channels, plus price, promotions, distribution, seasonality and the economy.
• Baseline: sales that arrive without marketing, from returning customers, brand demand, habit and word of mouth.
• Also captures delayed response, diminishing returns and channels supporting each other, which a budget plan needs.

The differences, dimension by dimension

Attribution, then contribution:

• Total divided: tracked conversions, versus total revenue including the baseline.
• Coverage: trackable digital, versus every channel with a spend record, plus price and seasonality.
• Data and speed: user-level events daily, versus aggregate weekly data refitted each planning cycle.
• Grain: down to ad and keyword, versus channel and campaign.
• Best for: tactics inside channels, versus budget allocation, forecasting and finance review.

Why the same channel gets two different figures

Both figures can be correct for their own scope. Attribution's total leaves out:

• Untracked conversions, such as those driven by television or completed in a store.
• The baseline: sales that would have happened anyway are credited to touchpoints on the path.
• Usually most affected: channels near the purchase, such as brand search and retargeting.

A worked example

Illustrative figures: 1,000 orders in a month, 700 of them visible on a digital path.

• Attribution divides the 700. Paid search is the last touch on 350, so last click gives it 50%.
• Contribution divides all 1,000. The model puts 550 in the baseline and 120 down to paid search.
• So paid search is 12% of total orders, about a quarter of what marketing added.
• A budget decision needs the second figure: how many orders it produced.

How large the baseline is

• Across 138 advertisers and more than 870 models, the median advertiser had 56% of revenue arriving as baseline.
• Middle half: between 30% and 80%.
• On a typical business it is the largest single component of revenue, varying by category, brand strength and repeat customers.

Where mixing the two causes problems

• Budgets set on attributed share can favour the channel best placed to capture demand already coming.
• Brand spend looks weak on attributed figures, because brand builds the baseline.
• Growth targets on attributed totals overstate what a larger budget can move.
• Summed platform totals count shared sales more than once; contribution shares cannot exceed actual orders.

Changing the attribution rule does not change the total

• A better attribution model, such as data-driven over last click, improves how credit is shared.
• The total stays the same, so the business share needs a model of total revenue.
• Better tracking narrows the tracking gap; the baseline gap remains.

What contribution does not fix on its own

• Correlation: alone, a contribution model can struggle to separate a cause from a driver that moved with it.
• Experiments such as geo tests separate them; calibrating with them makes the shares hold up.
• Cadence: on its own it does not guide creative rotation or daily bidding.
• Thin history gives a wide range, the accurate result for that data.

How to choose

• Has to add up to the business (revenue, growth, a budget change), or covers offline and brand media: use contribution.
• About a path (which touchpoints, in what order, over how long): use attribution.
• Two reports disagree: check what each sums to; tracked conversions describe the tracked subset.

Reporting both numbers

• Label each report: share of tracked conversions, or share of total revenue.
• Agree in advance: contribution for budgets and growth plans; attribution for creative, bidding and funnels.
• Share the baseline figure, so attributed totals are read for what they are.

Questions

Related comparisons

• MMM vs multi-touch attribution
• First touch vs last touch vs multi touch
• MMM vs incrementality testing