Marketing Attribution: Models, Limits and How to Use It

Marketing attribution is the practice of assigning credit for a sale or lead to the marketing touchpoints that came before it. It answers a plain question: which ads, searches, emails and visits actually helped this customer say yes? Get it right and you move budget toward what works. Get it wrong and you cut the channel that was quietly doing the heavy lifting.

We're a marketing agency in London, Ontario, and we work with businesses across several industries, not one niche. This guide explains the main attribution models in plain language, shows where each one misleads, and covers the privacy rules Canadian businesses need to respect when they track.

TL;DR

• Attribution assigns credit to touchpoints. It describes the customer journey. It does not prove what caused a sale.
• Last click and first click each give 100% of the credit to one touchpoint. Both are simple, and both leave out most of the journey.
• Multi-touch models (linear, time decay, position-based, data-driven) spread credit across several touchpoints. They are only as good as your tracking.
• In Canada, tracking has to follow PIPEDA and CASL. Consent matters more than the tool you pick.
• Start with clean conversion tracking and one or two models side by side. Test before you make big budget moves.

Table of Contents

• What marketing attribution actually does
• The main attribution models, one by one
• Last click vs. first click
• Why multi-touch attribution is worth the effort
• Where attribution goes wrong
• How to measure attribution as a Canadian business
• Privacy rules that shape your tracking in Canada
• Our perspective
• Work with a local team on measurement
• FAQ
• Sources

What marketing attribution actually does

A customer rarely buys after one ad. They see a Facebook post, search your name a week later, click a Google ad, then call. Attribution is the rulebook for deciding who gets credit for that call.

The output is a credit allocation. According to Ipsos MMA, that allocation feeds reporting dashboards and channel-level ROI calculations. It often drives tactical choices such as bid adjustments, budget shifts and audience changes.

That matters because money is on the line. Martech.org cites Proxima research estimating that companies waste 40 to 60% of their digital marketing budgets on ineffective spending, poor targeting and inefficient strategies. Attribution is one way to find where that waste sits. It won't fix it by itself.

Here is how it works in practice. The model looks at the sequence of interactions before a conversion and spreads credit by rules or by statistics. Inputs can include ad exposures, clicks, site visits and conversion events, connected by a cookie, device ID, logged-in profile or probabilistic match, as Ipsos MMA explains. If those inputs are patchy, the output is patchy.

The main attribution models, one by one

The textbook Foundations in Digital Marketing describes first or last touch, even (linear), time decay, starter/player/closer (U-shaped) and custom attribution. Shopify Canada lists seven common types: last click, first click, last non-direct click, linear, time decay, position-based and algorithmic.

Here is the short version.

Position-based and W-shaped splits

The common position-based setup gives 40% to the first touch, 40% to the last and 20% to the middle touches, according to Statsig. When there are no middle interactions, first and last each get 50%. The W-shaped version adds a third milestone, usually lead or opportunity creation. A common split is 30% each to first touch, mid-funnel event and last touch, with the remaining 10% shared among other touches. It suits B2B companies with defined sales funnels.

Those percentages are conventions. Nothing in your customers' behavior says the first touch is worth exactly 40%. Treat them as a lens, not a law.

Rules-based versus algorithmic

Nielsen separates rules-based models (first touch, last touch, even weighting, position-based, time decay) from algorithmic ones. Algorithmic models use machine learning to assign fractional credit based on each touchpoint's observed statistical contribution. Nielsen notes these models can be rebuilt as often as daily, so you can adjust while a campaign is still running.

For a local service business with a modest ad budget, algorithmic models are usually more machinery than the data can support. A clear rules-based view, checked against real phone calls and booked jobs, tends to be more useful.

Last click vs. first click

These two get the most searches and cause the most arguments, so they deserve a closer look.

Last click gives 100% of the credit to the final click. Shopify Canada calls it the most commonly used model and the default for most marketing platforms. It answers "what closed this customer?" That makes it handy for direct-response campaigns and short sales cycles.

First click gives 100% to the first click. It answers "what introduced this customer to us?" It is useful when you are deciding where to spend to find new audiences.

Both have a blind spot. The Foundations in Digital Marketing text points out that first-touch attribution under-represents the branding and influence effects other campaigns had in convincing the customer. On the other side, a Communications of the ACM guide says last touch routinely undervalues awareness-building channels and pushes investment toward end-of-funnel tactics. That can distort strategy and slow growth.

A practical example. A roofer runs Facebook video ads and a Google search ad. Homeowners see the video, forget about it, then search "roof repair near me" months later when a leak appears. Under last click, Google gets all the credit and Facebook looks like a waste. Cut Facebook and, a few months later, the searches dry up. That is the mistake to avoid.

Good answer: "We look at last click and first click side by side, then check both against actual booked jobs."

Walk-away answer: "The platform dashboard says this channel is the winner, so we're moving everything there."

If you are choosing where to spend between the two big ad platforms, our comparison of Google Ads vs Facebook Ads walks through what each is good at.

Why multi-touch attribution is worth the effort

Multi-touch attribution looks across the whole journey instead of crowning one click. People really do move across channels. A Salsify survey of more than 2,700 shoppers in the US, Canada and the UK, reported by Shopify, found more than half check two to three channels before buying mid-range items like apparel, beauty or sports equipment. Your customers may behave differently, but the point stands: one-touch models describe a journey that mostly doesn't exist.

A sensible path, as the ACM guide frames it, is gradual. Use last touch or first touch as a baseline, then move to linear, where credit is split equally across touchpoints. Add more complex models only when your data supports them.

Where attribution goes wrong

Attribution looks precise. It isn't. Know these limits before you act on a report.

Credit is not causation

Attribution describes the journey. Ipsos MMA draws the line clearly: incrementality testing and marketing mix modeling are the tools that measure causation. A customer who was going to buy anyway still registers as an attributed conversion under most models. They don't register as an incremental one, because removing the ad would not have changed what they did, as Ipsos MMA explains in its piece on incrementality.

Branded search is the classic case. Someone types your business name into Google and clicks your ad. The ad gets credit, but that person had already decided.

Tracking is leakier than it looks

Ipsos MMA also notes that the reliability of attribution depends on the quality of the identity data and the validity of the credit logic, and that both have become harder to trust in the post-cookie era. A buyer who browses on a phone, then buys on a laptop, can look like two people. Calls and walk-ins often don't show up at all.

That last point is serious for local businesses. If most of your leads phone you, a click-based model is missing the most valuable part of the picture unless you track calls and ask people how they found you.

Platforms grade their own homework

Every ad platform reports conversions using its own rules, and each wants the credit. Add up the numbers from several dashboards and the total can easily exceed the leads you actually received. Your own records are the referee: booked jobs, signed contracts, calls answered.

The lookback window

The "first" click is only the first click inside the lookback window, not the first time someone heard of you. If your sales cycle runs longer than the window, early touches vanish from the report.

How to measure attribution as a Canadian business

You don't need an enterprise stack. You need a short list done properly.

1. Define the conversion. A form submission, a tracked phone call, a booked appointment. Pick the actions that mean money.
2. Set up tracking correctly. Our guides on GA4 conversion tracking and Google Ads conversion tracking cover the setup step by step.
3. Tag your campaigns. Use consistent UTM parameters so every link you share says where it came from.
4. Ask the customer. Add "How did you hear about us?" to forms and intake calls. Self-reported answers catch word of mouth, podcasts and offline touches that tools miss.
5. Compare two models. Look at last click and first click, or last click and linear. Where they disagree, you've found a channel worth investigating.
6. Tie it to revenue. Credit to a click is not credit to cash. Our explainer on ROAS vs ROI shows how to connect the two.
7. Test before big moves. Before you cut a channel, pause or change it in one area and watch what happens. Geo holdout tests, where some markets get the marketing and matched control markets don't, are described by Ipsos MMA as among the most reliable evidence for a channel's incremental contribution. A small business can run a simplified version: change spend in one city and compare it with a similar one.

Privacy rules that shape your tracking in Canada

Tracking is personal data. In Canada, that means rules you need to respect, and you should confirm the details for your situation with a privacy lawyer.

• PIPEDA governs how private-sector organizations collect, use and disclose personal information in commercial activities, including marketing, as BLG's guide explains.
• CASL prohibits sending a commercial electronic message unless the recipient has consented and the message meets form and content requirements. Its commercial message provisions took effect on July 1, 2014, per the same BLG guide. The Privacy Commissioner's guide for businesses notes CASL also amended PIPEDA to add restrictions on address harvesting and on collecting personal information through spyware.
• Sharing data with ad platforms needs care. In its Home Depot investigation, the Office of the Privacy Commissioner found the company should have obtained express opt-in consent before sharing customers' hashed email addresses and offline purchase details with Meta for ad measurement. An opt-out after the fact was not enough.
• Collect only what you need. In the Tim Hortons findings, the Commissioner concluded the company had no legitimate need to collect vast amounts of sensitive location data it never used for its stated purpose.

The practical takeaway: if you upload customer lists or share offline sales with an ad platform to improve measurement, get clear, prior consent first, and keep your privacy policy honest about what you collect.

Our perspective

We see the same pattern across the businesses we work with in London and the surrounding Ontario service areas. The owner opens an ad dashboard, sees one channel taking the credit, and starts moving money. Six weeks later the phone is quieter and nobody knows why.

Our view is that attribution is a decision aid, not a verdict. We'd rather start with the basics: conversions that are tracked properly, calls and forms counted, and a website that gives visitors a reason to act. Then we read the numbers from more than one angle, because a single model will always flatter somebody.

That is also why we like to work across channels. Search, social ads, a site built to rank, and professional photo and video all touch the same customer at different moments. Judging each in isolation is how good work gets cut. Judging them together, against real leads and real revenue, is how budgets get smarter.

We work with multiple industries, so we don't hand out a template. What counts as a conversion for a contractor is not what counts for a clinic or a retailer, and we scope our work to fit.

Launch & Found

Work with a local team on measurement

If attribution feels like a pile of dashboards that don't agree, a good starting point is a plain look at what you run today and what it's producing. We support businesses in London, Ontario first, and also serve companies across surrounding Ontario service areas. Our work covers Google Ads management, Facebook and Instagram advertising, SEO, website design and media production.

Pricing is by custom quote after we assess fit, because scope depends on your business, your market and your goals. Contact Launch & Found for a quote and we'll tell you honestly whether we're the right fit. You can also see an overview of what we do on our main site.

FAQ

What is marketing attribution in simple terms?

It is the method for deciding which marketing touchpoints get credit for a sale or lead. A customer might see an ad, search your name and then call. Attribution is the rule that splits the credit among those steps, so you can judge which channels earn their budget.

What is the difference between first-click and last-click attribution?

First click gives all the credit to the first touchpoint, which shows what introduced the customer to you. Last click gives all the credit to the final touchpoint, which shows what closed the deal. Both ignore the steps in between, so most businesses look at both together.

Which attribution model is best for a small local business?

There is no single best model. Many local businesses do well starting with last click and first click side by side, then adding a linear or position-based view as tracking improves. Whatever you choose, verify it against booked jobs and phone calls, since click data alone misses offline leads.

Does attribution prove which channel caused a sale?

No. It assigns credit based on the path a customer took. Customers who were already going to buy still show up as conversions. Measuring cause takes experiments such as holdout tests or marketing mix modeling, which Ipsos MMA treats as separate from attribution.

Do Canadian privacy laws affect attribution tracking?

Yes. PIPEDA covers how businesses collect and use personal information in commercial activity, and CASL covers commercial electronic messages. The Privacy Commissioner has found that sharing customer data with an ad platform for measurement called for express opt-in consent. Check your setup with a privacy professional.

Sources

• What Is Marketing Attribution? | Ipsos MMA
• What Is Marketing Incrementality? | Ipsos MMA
• What Is Marketing Attribution? Models & Best Practices | Martech.org
• Attribution, Foundations in Digital Marketing
• Marketing Attribution: Models and Best Practices | Shopify Canada
• Multitouch Attribution: Models and Tips | Shopify
• Digital marketing attribution models: A tech survey | Statsig
• Methods & Models: A Guide to Multi-Touch Attribution | Nielsen
• A Practical Guide to Last-Touch and Multi-Touch Attribution | Communications of the ACM
• Targeting the Canadian Consumer | BLG
• Privacy Guide for Businesses | Office of the Privacy Commissioner of Canada
• PIPEDA Findings #2023-001: Home Depot | Office of the Privacy Commissioner of Canada
• PIPEDA Findings #2022-001: Tim Hortons | Office of the Privacy Commissioner of Canada

Recommended

• ROAS vs ROI: What Each Tells You and Which to Track
• GA4 Conversion Tracking: How to Set It Up Right
• Google Ads Conversion Tracking: Set It Up Right
• Marketing Automation for Small Business: Where to Start