Small Business Marketing Attribution: What It Can Prove

Imagine a homeowner who needs a planned installation. On Monday, a neighbor mentions your company. On Wednesday, the homeowner searches for the service and reads one of your pages. On Friday, they see an ad while comparing providers. The following week, they type your business name into a search engine and call.

This is an invented example. Its purpose is to show why “Where did the customer come from?” can have several valid answers. Discovery, evaluation, and final contact are related but different questions.

When an owner asks which campaigns bring inquiries, we look at the recorded acquisition touchpoint and the lead quality that follows. If you want to understand how customers first discover you, include intake answers and earlier touchpoints where available. If you want to estimate the effect of additional spending, attribution alone may not resolve the question.

A practical report can show more than one view. For example, keep a recorded digital source, a customer reported discovery source, and the last known contact source. Label them clearly instead of forcing them into one supposedly definitive field.

Google explains that the selected attribution model affects how conversion credit is assigned and can affect bidding based on those conversion actions. See Google Ads' attribution model guidance. Changing the model changes the reporting lens; it does not rewrite what the customer actually experienced.

We use one simple naming rule for the campaigns you control. Decide how email, paid social, partnerships, and other tagged links will be named. We record those rules in one place so team members don't create several labels for one source.

Google Analytics supports campaign parameters such as source, medium, and campaign. Its campaign URL documentation also notes that parameter values are case sensitive. Consistent naming reduces avoidable fragmentation in the report.

We keep customer information out of campaign URLs. A campaign label should identify the campaign, not a person's name, phone number, or private project details. The link should remain safe to share and understandable when someone encounters it outside the original message.

We keep the original recorded source when a later touchpoint arrives. Overwriting every earlier value with the latest visit makes the customer journey disappear. Your CRM measurement design should state which fields are original, which can change, and which are separate observations.

Some calls will have no reliable digital source. Some customers will not remember where they first heard of you. Some visits will not be observable because of consent choices, device changes, blocking, or other measurement limits. An unknown category is more honest than distributing those outcomes across channels without a stated method.

We watch whether the unknown share changes much. That can reveal a broken form field, a call routing change, or a new acquisition route. It can also reflect normal limits. We investigate before calling the report either complete or useless.

Ask the intake question neutrally: “How did you first hear about us?” We avoid steering the person toward the channel we'd like to see win. We keep their answer next to the observed digital source, and neither one automatically corrects the other.

Two advertising systems may both claim credit for the same customer's outcome because they observe different parts of the journey and use different rules. Your business still completed one job. A combined report should reconcile outcomes at the opportunity or transaction level where possible.

We separate attributed conversions from unique business outcomes. Also state the reporting window and conversion definition. Our commercial roofing microsites logged four inquiries from 209 Indiana sites from June 29 through August 31, 2026. Missed calls aren't counted, and a TPO inquiry on September 2 stays out because it came after the window. A platform might report an inquiry, while the CRM reports an accepted project. Comparing those totals without context creates a false disagreement.

For paid advertising, use the platform view to manage the campaign and a reconciled business view to understand total results. For search visibility work, examine both discovery indicators and downstream opportunities, recognizing that some influence will remain unobserved.

A simple evidence label can improve the conversation. “Observed” means the event or relationship was recorded directly. “Customer reported” means it came from an intake answer. “Modeled” means a system allocated credit. “Inferred” means the team has a plausible explanation that has not been directly established.

Those labels do not rank every method from good to bad. They explain what kind of evidence you have. A customer interview can reveal something a tag never could, while a timestamped digital record can be more precise about an actual click.

• Which decision will this attribution view support?
• What events and people are included?
• What model, window, and source rules apply?
• How are duplicate outcomes reconciled?
• How much remains unknown?
• What additional evidence would change the decision?

Bring these questions to your agency reporting meeting. A useful attribution system makes uncertainty manageable and keeps the team focused on decisions it can support. It should help explain how marketing and sales work together without claiming to observe every reason a customer chose you.