Back to all articlesCRM measurement

Allocate Attribution Credit When Several Sources Contribute

Alactic Editorial10 min read

Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.

If six new customers become twelve attributed customers when you add the source rows, your report is counting source participation rather than allocating customer credit. Both measures can be useful, but they answer different questions. They should not share the same label or grand total.

To make credited outcomes reconcile to the underlying business total, give each outcome either one exclusive source or fractional shares that sum to one. Keep full-credit participation counts separate. The essential check is local before it is global: each eligible outcome must contribute exactly one unit of credit across all sources, including an explicit unknown bucket where necessary.

The worksheet below compares these choices using the same fictional customer records. It also shows why a report can balance numerically while still using an allocation rule that does not suit the decision.

1. Define the total you intend to reconcile

Start with the business outcome, not the source field. A lead, an opportunity and a customer are different counting units. Allocating one unit for every lead record cannot be reconciled directly to a total of unique customers.

Write a short measurement contract before building the report. It should specify:

  • Counting unit: the unique entity or event receiving one unit of credit.
  • Eligibility: the status and date conditions that put an outcome in the report.
  • Reporting period: based on the outcome date, rather than whichever touch dates happen to be available.
  • Eligible touch window: the interval in which a recorded interaction may enter the allocation.
  • Source definitions: the categories and mapping rules used consistently across records.
  • Unknown treatment: how outcomes with no eligible recorded source remain visible.

For the fictional worksheet, the contract is: count six unique customers whose first customer date falls in a hypothetical April reporting period. Include recorded touches in the 30 days ending at conversion. Exclude later touches. Deduplicate records before allocating. A customer without an eligible source receives one unit in Unknown.

These are hypothetical design choices, not universal defaults. A different business process may require another window or counting unit. What matters is that the denominator and the allocation use the same eligible population.

2. Choose exclusive credit or fractional credit

Exclusive credit assigns the whole outcome to one source. A transparent rule might select the first eligible source, the last eligible source or a designated primary source. It is simple to explain and produces whole-number source totals. Its limitation is deliberate: other recorded sources receive no allocated credit.

Fractional credit divides the outcome among eligible sources. Equal-share allocation treats each included source equally. Explicit weighting assigns different shares using a documented rule. Both can reconcile, provided every outcome's shares sum to one.

Full credit for every source is different. If three sources appear on one customer's eligible path, all three receive one participation count. Summing those rows produces three source participations, not three customers. Use labels such as “customers with this source recorded” and explain that rows overlap.

Choose exclusive credit when the operational question requires a single owner or source assignment. Choose fractional credit when you want a partition of outcomes that preserves several recorded sources. Retain participation counts when the question is how often sources appear, but do not present their sum as an allocated customer total.

None of these choices demonstrates causal contribution. Recorded sequence and assigned credit do not show whether a customer would have converted without a source.

3. Set the allocation rules before calculating

The fictional example uses four known categories: Search, Webinar, Referral and Sales conversation. These are recording categories, not claims that acquisition sources and sales interactions are interchangeable. Including them in one allocation is a conscious reporting choice; a team interested only in acquisition sources should define a narrower eligible set.

For this worksheet, repeated occurrences of the same source within a customer's eligible window collapse into one source entry. The earliest eligible occurrence determines its position. This prevents repeated activity in one category from automatically earning extra shares.

The three comparison rules are:

  1. Full-credit participation: give every distinct eligible source one participation count.
  2. Equal-share allocation: divide one unit by the number of distinct eligible sources.
  3. Explicitly weighted allocation: for three sources, allocate 20% to the first, 30% to the middle and 50% to the last; for two, allocate 40% to the first and 60% to the last; for one, allocate 100% to that source.

For a customer with no eligible source, all three views place one unit in Unknown. Unknown is a completeness bucket, not an observed touch.

The weighted percentages are hypothetical policy choices. They illustrate a preference for later recorded sources, not measured influence. This example permits at most three distinct sources. Before applying the rule to other data, define what happens with longer paths and tied timestamps. Until then, flag those records for review rather than silently inventing weights.

4. Filled hypothetical multi-touch allocation worksheet

Fictional worked example: every record and number below is hypothetical. Each customer is eligible once, and source order runs from earliest to latest. Fractions are retained exactly in the worksheet so rounding does not obscure the reconciliation.

ELIGIBLE OUTCOME REGISTER
Customer | Ordered eligible sources
C01      | Search > Webinar > Sales conversation
C02      | Referral > Sales conversation
C03      | Search > Sales conversation
C04      | Webinar > Search > Sales conversation
C05      | Referral
C06      | No eligible source: use Unknown

Control total: 6 unique customers

Per-customer allocation

C01
Full credit: Search 1; Webinar 1; Sales 1           Sum: 3
Equal share: Search 1/3; Webinar 1/3; Sales 1/3     Sum: 1
Weighted:   Search 0.20; Webinar 0.30; Sales 0.50  Sum: 1

C02
Full credit: Referral 1; Sales 1                  Sum: 2
Equal share: Referral 0.50; Sales 0.50            Sum: 1
Weighted:   Referral 0.40; Sales 0.60             Sum: 1

C03
Full credit: Search 1; Sales 1                    Sum: 2
Equal share: Search 0.50; Sales 0.50              Sum: 1
Weighted:   Search 0.40; Sales 0.60               Sum: 1

C04
Full credit: Webinar 1; Search 1; Sales 1          Sum: 3
Equal share: Webinar 1/3; Search 1/3; Sales 1/3    Sum: 1
Weighted:   Webinar 0.20; Search 0.30; Sales 0.50  Sum: 1

C05
Full credit: Referral 1                          Sum: 1
Equal share: Referral 1                          Sum: 1
Weighted:   Referral 1                           Sum: 1

C06
Full credit: Unknown 1                           Sum: 1
Equal share: Unknown 1                           Sum: 1
Weighted:   Unknown 1                            Sum: 1

Source totals and control checks

Source              Full credit   Equal share   Weighted
Search                    3           7/6          0.90
Webinar                   2           2/3          0.50
Sales conversation        4           5/3          2.20
Referral                  2           3/2          1.40
Unknown                   1             1          1.00
------------------------------------------------------
Column total             12             6          6.00
Underlying customers      6             6          6.00
Difference               +6             0          0.00

The full-credit column has twelve source participations across six customers. That is valid as an overlapping participation measure, but it fails as a partition of customer credit. Both fractional columns reconcile to six.

For an exclusive last-source comparison, Sales conversation receives four customers, Referral receives one and Unknown receives one. Search and Webinar receive zero. The exclusive total is also six, though it tells a substantially different story from either fractional view.

This distinction is the practical decision: reconciliation constrains the arithmetic, but does not choose the policy. In this hypothetical example, Sales conversation receives 5/3 units under equal shares and 2.20 under the weighted rule solely because the allocation policy changes.

5. Build checks that expose errors rather than offset them

A balanced grand total is necessary but insufficient. One outcome could receive too much credit while another receives too little, leaving the report apparently correct. Check at three levels.

Outcome-level checks

  • Every eligible outcome appears in the allocation register.
  • No ineligible outcome appears.
  • Every credit is nonnegative and no greater than one.
  • Each outcome's fractional or exclusive credit sums to one.
  • Every outcome without a known eligible source is assigned to Unknown.

Population and aggregation checks

Compare the distinct allocated outcome IDs with the control register in both directions. A matching count alone can hide one missing customer and one incorrectly included customer. Then sum allocation rows by source and confirm that their total equals the control total.

For the fictional worksheet, the acceptance checks are six matching customer IDs, zero extra IDs, zero missing IDs, zero outcome-level sum failures and six total allocated units.

Precision checks

Calculate with sufficient precision and round only for display. Three displayed shares of 0.33 sum to 0.99 even when their underlying values are exactly one-third. Do not repair this by changing the business total.

If displayed rows must add exactly to the displayed total, use a documented rounding-residual rule and retain the unrounded values. Distinguish a display adjustment from a change to allocation policy.

6. Recover from the failures that commonly inflate credit

Repeated source activity changes the denominator. If one customer has several Search records, a touch-based equal-share calculation differs from a distinct-source calculation. Neither is automatically wrong, but switching between them silently is. Recover by rebuilding the eligible source list according to the declared unit.

Record joins multiply outcomes. Combining customer, contact and activity records can create several rows for the same customer-source pair. Recover by checking uniqueness at the intended allocation grain before assigning credit, not by removing duplicates from the final source totals.

Missing sources disappear. Dropping customers with blank source fields makes the credited total smaller than the business total. Recover by assigning Unknown and reporting its share separately. Do not redistribute its credit among known sources merely to make their percentages look complete.

Filters apply after allocation. A report that hides Unknown will show five credited units in the hypothetical example. That is a known-source subtotal, not the six-customer total. Either retain the full population or label the reduced denominator explicitly. Renormalizing visible shares answers a different question and must be disclosed.

Late records rewrite history. A newly recorded earlier touch can alter source order and weights. Choose either a frozen reporting snapshot or a documented restatement process. Preserve the previous version so changes can be traced to corrected evidence rather than mistaken for new business outcomes.

7. Publish the meaning alongside the numbers

A usable report needs a compact policy note, not just a chart legend. State the counting unit, outcome period, touch window, repeated-source treatment, allocation rule, unknown treatment and rule version.

Keep participation and allocated credit in separate columns. Label fractional results “allocated customer credit,” rather than implying that a decimal represents a partial person. Label exclusive results with the selection rule, such as “customers assigned to last eligible source.”

Also state what the report cannot establish. Missing offline interactions, inconsistent source capture and the chosen lookback window can change the recorded path. Reconciliation proves that credit was conserved within the selected population; it does not prove that the path is complete, the weights are fair or the sources caused the outcomes.

If the report will guide a consequential operational decision, compare at least one alternative rule. Large shifts between exclusive, equal-share and weighted results are a reason to examine sensitivity and recording practices, not evidence that one rule has discovered the truth.

8. Run a bounded reconciliation before the next report

For the next reporting cycle, select one closed outcome cohort and produce a control register plus an allocation sheet. Use the hypothetical worksheet structure above, replacing its records and policy choices with your documented definitions.

  1. Confirm the unique outcome IDs and freeze the eligible source records.
  2. Calculate one exclusive rule and one fractional rule.
  3. Resolve every missing ID, extra ID and outcome-level credit failure.
  4. Publish the source totals with Unknown visible and participation counts clearly separated.
  5. Record the rule version, unresolved data issues and any reporting restatements.

The measurable completion condition is straightforward: the allocated ID set matches the intended outcome set, every outcome receives exactly one unit within the declared numerical precision, and source credit totals equal the business total. Only after those checks pass should readers compare how the chosen rule distributes that credit.

CRM attributionData qualityReportingRevenue operations

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