Attribution modeling is the set of rules you use to decide which marketing touchpoints get credit for a sale. A customer rarely clicks one ad and buys. They see an Instagram post, read a blog article, open an email, search your brand name on Google and then convert. Attribution modeling decides how that single sale is split across those five moments, and different rules produce wildly different answers.
That is exactly why your Google Ads dashboard, your Meta Ads Manager, your email platform and your analytics tool all claim credit for the same revenue. Below we break down each model in plain English, then score one identical customer journey under five models so you can see, in dollars, how much the choice of model changes your reporting.
What is attribution modeling, in one paragraph?
Attribution modeling is the process of assigning conversion credit to the marketing channels and touchpoints a person interacted with before buying. A single-touch model gives 100% of the credit to one touchpoint (the first or the last). A multi-touch model spreads the credit across several touchpoints using a fixed rule (linear, time-decay, position-based). A data-driven model uses your own conversion data to work out how much each touchpoint actually contributed, rather than applying a rule of thumb.
Attribution is not an accounting truth. It is a lens. The job is to pick the lens that matches how people really buy from you.
Why your Google Ads, email and social numbers never add up
If you add up the conversions reported by each platform, you will almost always get more sales than your bank account shows. Three reasons:
- Each platform uses its own model. Google Ads defaults to data-driven attribution across Google properties. Meta typically counts a conversion within 7 days of a click or 1 day of a view. Your email tool usually uses last-click within its own tracked links.
- Each platform only sees its own touchpoints. Meta cannot see the Google search that happened three days later, so it takes full credit. Google cannot see the Instagram ad that started everything.
- Lookback windows differ. A 90-day window catches journeys a 7-day window never will.
Platform numbers are useful for optimising inside that platform. They are not a budget allocation report. For that, you need one model applied consistently across all channels.
Four words you need before we start
- Touchpoint: any tracked interaction with your marketing (ad click, organic visit, email click, referral).
- Conversion: the action you care about (purchase, qualified lead, booked call).
- Lookback window: how far back a tool looks for touchpoints before the conversion. Commonly 30, 60 or 90 days.
- Credit: the share of the conversion (and its value) handed to a touchpoint. Usually expressed as a percentage or a fraction of a conversion.
One customer journey, five attribution models
The journey
Meet Sofia. She runs a small design studio and buys a $600 annual software subscription from you. Here is what her 18 days looked like:
- Day 1: clicks an Instagram ad (paid social). Browses, leaves.
- Day 4: finds your comparison blog post through Google organic search. Signs up to your newsletter.
- Day 11: clicks a link in your onboarding email sequence and watches a demo video.
- Day 18: searches your brand name, clicks your Google Ads branded search ad, and buys. Conversion value: $600.
1. First-click attribution
100% of the credit goes to the first touchpoint. Sofia’s Instagram ad gets the full $600. Organic, email and Google Ads get nothing.
Use it when: you are testing top-of-funnel channels and want to know what creates demand.
Blind spot: it tells you nothing about what closes deals, and it over-rewards cheap awareness clicks that never convert on their own.
2. Last-click attribution
100% of the credit goes to the final touchpoint. Google Ads branded search takes the full $600, which makes branded search look like your best performing channel in the entire business.
Note the nuance: most tools use last non-direct click. If Sofia had typed your URL directly on day 18, the credit would jump back to the email click, not to “direct”.
Use it when: your sales cycle is genuinely short (impulse ecommerce, local services) and most journeys are one or two touches.
Blind spot: it systematically over-credits branded search, retargeting and coupon sites, and it will push you to cut the channels that actually created the demand.
3. Linear attribution
Every touchpoint gets an equal share. Four touchpoints means 25% each, so $150 to Instagram, organic, email and Google Ads. quantummetric.com makes the same point with more data.
Use it when: you want a fair, easy-to-explain starting point for a multi-channel business, or you are reporting to stakeholders who distrust black boxes.
Blind spot: a throwaway 4-second visit counts as much as a 20-minute demo.
4. Time-decay attribution
Touchpoints closer to the conversion get more credit. The standard setting is a 7-day half-life: a touch 7 days before the sale is worth half of one that happens on the day of the sale.
Applied to Sofia’s journey, the weights work out to roughly 9.6% / 12.9% / 25.8% / 51.7%.
Use it when: you run longer consideration cycles with nurture sequences, or seasonal pushes where recency matters.
Blind spot: the half-life is an assumption, not a measurement. A 90-day B2B cycle with a 7-day half-life will erase your entire top of funnel.
5. Position-based (U-shaped) attribution
40% to the first touch, 40% to the last, and the remaining 20% split among everything in between. Sofia’s journey: $240 Instagram, $60 organic, $60 email, $240 Google Ads.
Use it when: you believe discovery and closing are the two hardest jobs and the middle is nurture.
Blind spot: it flattens the middle even when the middle is where the real persuasion happens (in Sofia’s case, the demo video).
6. Data-driven attribution
Instead of applying a fixed rule, a data-driven model compares thousands of converting and non-converting journeys and calculates how much each touchpoint changed the probability of conversion. Touchpoints that reliably move people forward get more credit, regardless of position.
For a business like this one, the model might return something like 31% Instagram, 12% organic, 34% email and 23% branded search, because it notices that people who click the onboarding email convert far more often than people who do not, while branded search clicks tend to happen after the decision is already made.
Use it when: you have enough conversion volume for the model to learn from (Google Ads and GA4 both apply thresholds) and multiple channels running at once.
Blind spot: it is a black box, it still only sees the touchpoints your tracking captures, and it cannot credit anything offline, in a private group chat or in a podcast mention.
Side by side: the same $600 sale, five ways
| Model | Instagram ad (Day 1) | Organic blog (Day 4) | Email (Day 11) | Google Ads brand (Day 18) |
|---|---|---|---|---|
| First-click | $600 (100%) | $0 | $0 | $0 |
| Last-click | $0 | $0 | $0 | $600 (100%) |
| Linear | $150 (25%) | $150 (25%) | $150 (25%) | $150 (25%) |
| Time-decay (7-day half-life) | $58 (9.6%) | $77 (12.9%) | $155 (25.8%) | $310 (51.7%) |
| Position-based (40/20/40) | $240 (40%) | $60 (10%) | $60 (10%) | $240 (40%) |
| Data-driven (illustrative) | $186 (31%) | $72 (12%) | $204 (34%) | $138 (23%) |
What that table actually tells you
Same customer. Same $600. And the Instagram ad is worth anywhere between $0 and $600 depending on the model you happen to have open. idea2product2business.com goes into the numbers.
Three practical takeaways:
- Last-click will get your paid social budget cut. If you judge Instagram on last-click revenue, it looks like a money pit. Under every other model, it is the reason the sale exists at all.
- Branded search is the great illusionist. Under last-click it earns 100% of the credit. Under data-driven, it earns 23%, because the customer was already sold before they typed your name.
- Never compare a number from one model with a number from another. “Email ROI is 12x” means nothing without “under which model, and which lookback window”.
How to pick the model that matches your sales cycle
| Your situation | Typical cycle | Best starting model | Why |
|---|---|---|---|
| Low-price ecommerce, impulse buys | 0 to 2 days | Last-click (or data-driven if volume allows) | Most journeys are one or two touches, so extra complexity adds nothing |
| Local service business (plumber, clinic, salon) | 0 to 7 days | Last non-direct click plus call tracking | The bottleneck is offline conversion tracking, not the model |
| Considered ecommerce ($200+ basket) | 7 to 30 days | Time-decay or data-driven | Recency matters but discovery deserves credit too |
| Lead generation and SaaS trials | 14 to 60 days | Position-based or data-driven | First touch and final decision both carry real weight |
| B2B with sales calls and contracts | 60 to 180+ days | Linear or position-based, CRM-based, 90-day window | Web analytics alone cannot see the deal, so credit must flow through the CRM |
| Brand building, offline, retail | Variable | Any model plus geo holdout tests | Click attribution cannot measure what it cannot track |
Rule of thumb: the longer and more multi-touch your sales cycle, the more damage last-click does to your budget decisions.
Attribution in the tools you already have
Google Analytics 4
GA4 retired first-click, linear, time-decay and position-based models back in 2023. You now choose between data-driven attribution (the default), cross-channel last click and Google paid channels last click, set in Admin under Attribution settings. You can also set the acquisition conversion lookback window to 30, 60 or 90 days. Pick 90 if your cycle runs longer than a month.
Google Ads
Data-driven attribution is the default for most conversion actions. Your default click-through conversion window is 30 days and can be extended up to 90. Bidding strategies learn from whatever model and window you set, so changing the model changes how the algorithm buys.
Meta Ads
Meta reports on its own attribution setting, commonly 7-day click and 1-day view. View-through conversions are the main reason Meta’s reported sales exceed what you see anywhere else.
Email and CRM
Email platforms count a conversion when someone clicks a tracked link and later buys within their own window. That is a last-click view of a single channel. If you sell through a sales team, your CRM (with source and campaign fields on the contact record) is a better source of truth than any web analytics tool.
Six steps to get usable attribution without a data team
- Fix your UTMs first. No model can rescue traffic labelled “facebook”, “Facebook” and “fb” in three different reports. Agree a naming convention and write it down.
- Pick one lookback window that is longer than 90% of your real sales cycles, and use it everywhere.
- Choose one model as your source of truth for budget decisions, using the table above. Keep platform reports for in-platform optimisation only.
- Add a self-reported source field to your checkout or lead form (“How did you hear about us?”). It catches podcasts, word of mouth and dark social that no pixel will ever see.
- Send offline and CRM conversions back into your ad platforms so that closed deals, not form fills, drive optimisation.
- Validate with a holdout test once a quarter. Turn one channel off in a set of regions for two to four weeks and measure the change in total revenue. That is incrementality, and it beats any model.
Where attribution modeling breaks down
- Cross-device journeys: phone discovery, desktop purchase, two separate “users” unless someone logs in.
- Consent and privacy: tracking prevention and cookie banners mean a meaningful share of journeys is partly invisible, filled in with modeled data.
- Untrackable influence: a recommendation in a WhatsApp group, a podcast mention or a billboard will show up as “direct” or “organic”.
- Correlation, not causation: attribution tells you who was present at the sale, not who caused it. Only experiments do that.
The honest position: use an attribution model to allocate budget week to week, use holdout tests and, if your spend is large enough, marketing mix modeling to sanity-check the big picture once or twice a year.
FAQ
What does attribution mean in simple terms?
Attribution means answering “what made this sale happen?”. In marketing, it is the practice of linking a conversion back to the ads, emails, posts and searches that led to it, and deciding how much credit each one deserves.
What are the types of attribution models?
The main ones are first-click, last-click (and last non-direct click), linear, time-decay, position-based (U-shaped) and data-driven. The first two are single-touch models. The next three are rule-based multi-touch models. Data-driven is algorithmic and uses your own conversion data.
What are the four types of attribution?
In a marketing context, people usually mean four families: first-touch, last-touch, multi-touch (linear, time-decay, position-based) and algorithmic or data-driven. Note that if you searched this expecting psychology, attribution theory is a different subject and refers to how people explain the causes of behaviour. hockeystack.com makes the same point with more data.
Which attribution model is best?
There is no universally best model. The best model is the one that matches your sales cycle and your data volume. Short cycles can live with last-click. Cycles longer than two weeks with multiple channels need a multi-touch or data-driven model. If your journeys average more than three touchpoints and you have consistent conversion volume, data-driven attribution usually gives the most realistic split.
What is the difference between attribution and incrementality?
Attribution distributes credit among touchpoints that were present. Incrementality measures what would have happened if a channel had not run at all, using holdout or geo tests. Attribution guides day-to-day decisions, incrementality validates them.
How long should my attribution lookback window be?
Look at the report of days-to-conversion in your analytics tool and pick a window that covers roughly 90% of journeys. For most small ecommerce brands that is 30 days. For lead generation and B2B, 90 days is safer.
The takeaway
Attribution modeling will not give you a perfect number. What it will give you is a consistent, defensible way to compare channels so you stop cutting the campaigns that quietly create demand and stop over-funding the ones that simply show up last. Pick the model that fits your sales cycle, apply it everywhere, and test it against reality once a quarter.
Need a hand mapping your own customer journey and choosing a model that fits it? Get in touch with the team at Love Camels and we will walk through your data with you.