One purchase. Five plausible explanations.
An operations director hears about a supplier from a colleague. Later, an engineer finds its application guide through search. Two people attend a webinar, someone forwards a follow-up email, and procurement eventually clicks a branded search advertisement before submitting an inquiry.
This is a hypothetical buying journey, not an observed PaperBoat account. It illustrates why several people can present convincing reports about the same sale. The event team sees attendance. Search reporting sees a visit. The paid platform sees a measured conversion. Sales remembers the conversation that resolved the final concern.
Each account may contain useful evidence. None gives the business permission to count the revenue five times. Nor does the final recorded interaction establish that the earlier work was unnecessary.
The executive question is usually more demanding than “Which channel gets credit?” It is “What should we fund, improve or stop next?” Answering that requires a clear definition of the outcome and an honest account of what the evidence can establish.
For the individual reports and tool checks, see what to look for in Analytics, Search Console and SEO tools. This article starts where the reports need to become a commercial decision.
Keep three questions separate.
What did we record?
A measured visit, registered attendee, email reply or CRM stage change establishes an event in a particular system. Check its definition, date and identity before connecting it to the opportunity.
What did the buyer say mattered?
A customer may explain that a colleague’s recommendation created interest or that a technical document resolved a concern. That account is valuable, but memory is selective and one person’s view may not represent the buying group.
What changed because of the activity?
A suitable comparison can help estimate incremental impact. The design must address the alternative explanation that customers would have acted anyway.
Google defines attribution as assigning credit to interactions along a measured path. A model supplies rules or an algorithm for that allocation. Its result should be interpreted within the data and settings available to it, not as a complete biography of the buying decision.
Keep the distinction visible in your own language. “This opportunity included a webinar attendee” is an observation. “The buyer said the demonstration reduced uncertainty” is a reported explanation. “The webinar generated incremental sales” is a stronger causal claim that requires different evidence.
Reconstruct one opportunity without inventing the missing parts.
Begin with a commercially meaningful record, such as a qualified opportunity, completed sale or renewal. Agree what qualifies and which date places it in the reporting cohort. A form submission and a signed agreement should not share a label that makes them look equivalent.
For the illustrative supplier journey, the record could look like this:
| Interaction | Available evidence | What remains uncertain |
|---|---|---|
| Peer recommendation | The director names the recommendation in an interview. | Its exact timing and relative influence are not independently measured. |
| Application guide | The engineer says it answered a compatibility question. Aggregate search data shows visits to the guide. | The aggregate data does not identify that person’s visit. |
| Webinar | Named registrations and attendance records match people attached to the account. | Attendance does not establish attention, understanding or causation. |
| Forwarded email | A buyer describes sharing the message with procurement. | The internal discussion and downstream readership may be unobserved. |
| Paid search inquiry | A tagged inquiry is connected to the CRM opportunity under the implemented measurement rules. | The final click does not establish that the ad created the original demand. |
Leave unknowns as unknowns. Do not join unrelated records simply because they share a company name or a convenient sequence of dates. Keep the distinction between an individual, an account and an opportunity. Several people may participate in one purchase, and one account may contain several unrelated projects.
Use the minimum personal information needed for a legitimate measurement purpose, with appropriate access and handling. A commercial record can describe the evidence and uncertainty without copying a customer’s private correspondence into every reporting tool.
Ask about the decision, not the channel you hope to defend.
“Did our webinar help?” gives the customer an easy route to a polite answer. A more useful conversation begins with what was happening in the business when the search for a solution started.
Ask what prompted the investigation, which alternatives entered the shortlist, who else participated and what needed to become clear before approval. Then ask where the buyer looked for answers and which material they shared internally. Invite them to distinguish what they remember from what they are reconstructing.
A technical guide might matter because it helped an engineer explain the purchase to finance. A webinar might establish that the supplier could answer difficult questions. An email might simply provide a convenient link. These roles are different, and none needs an invented percentage of revenue to be useful.
Include lost opportunities and stalled decisions where appropriate. Interviewing only enthusiastic customers can conceal why apparently engaged accounts never progressed. Keep each observation attached to its context instead of turning one memorable comment into a universal customer journey.
Count the outcome once. Let several sources explain it.
Suppose a hypothetical company closed ten new contracts in a defined cohort. Six involved people who attended an event, seven had a recorded search interaction and four had a paid campaign touchpoint. Those counts overlap. They describe associations with ten contracts, not seventeen contracts generated by channels.
A finance-facing report should use the agreed commercial source of truth for bookings or revenue. Platform reports can sit beside it as diagnostic views. If a platform assigns a conversion value, label that value according to its actual definition; do not silently add it to recognized revenue.
Keep the same discipline with costs. A campaign may use media, creative work, staff time, software and sales follow-up. Decide which costs belong in the question being asked. Cost per lead and customer acquisition cost answer different questions because their denominators and cost scopes differ.
A useful statement might be: “These contracts totaled the recorded bookings shown here. The associated interactions suggest the application guide and event supported evaluation. We have not estimated the incremental contribution of either.” That is a more reliable basis for discussion than a precise-looking return assembled from incompatible reports.
What should change in the next budget?
First identify the decision. Continuing a useful technical library, repairing an inquiry handoff and expanding an advertising audience are different commitments. They should not all be judged by the same last-click threshold.
Protect a capability with a documented job.
If buyers and salespeople repeatedly use a technical explanation to resolve an important question, maintaining it may be justified by that role. Record the evidence, the cost and the expected use. Do not convert “supports evaluation” into a guaranteed revenue contribution.
Fix an observed break.
If qualified inquiries are reaching the wrong team or disappearing before follow-up, repair the verified defect. You do not need a complete attribution model to recognize that an agreed customer handoff fails. Check the repaired journey and follow its subsequent outcomes.
Test an uncertain expansion.
If the proposal is to increase spending, define what the additional investment is expected to change. Where volume and implementation permit, use a suitable controlled experiment. Google describes Conversion Lift as a comparison between exposed and control groups to estimate incremental conversions; availability is limited to eligible accounts.
A low-volume technical business may not have enough independent outcomes for a useful test at the desired scale. A before-and-after comparison can still inform judgment, but seasonality, sales effort, pricing and market conditions can confound it. Record those limits. A bounded investment with a planned review may be more honest than pretending a small sample has settled causation.
For the next increment, assess qualified opportunities, progression, eventual economics and the time needed for outcomes to mature. A channel’s historical average return does not establish the return on spending more.
Bring an evidence note to the meeting.
Choose cohorts that have had a realistic opportunity to progress. Recent inquiries can be monitored for early quality, but comparing their closed sales with an older cohort gives the older group more time to succeed. Show the observation window and the proportion still open.
A useful budget recommendation contains five things
- The decision: continue, repair, expand, reduce or stop a defined activity.
- The evidence: commercial outcomes, recorded interactions and customer accounts, labeled separately.
- The uncertainty: missing identity, immature opportunities, alternative explanations or limited sample size.
- The commitment: cash, people and displaced work required.
- The review: the owner, timing and observation that would change the recommendation.
The Campaign Measurement Plan Builder can help define the question before the next campaign starts. The marketing decision-ownership guide explains who should act when the evidence points toward a change.
The company does not need to assign every sale perfectly before making a responsible decision. It does need to distinguish a useful observation from a confident story about what caused it.
