A customer sees your ad on Monday, reads a blog post on Wednesday, clicks an email on Friday, and buys on Sunday. Which of those touches gets the credit for the sale? That question is the whole problem of marketing attribution, and getting it wrong means you cut the channel that was quietly doing the real work.
I wrestle with this on my own businesses constantly, because attribution decides where the next marketing dollar goes. Spend it on the channel that looks good in a simple report, and you might be pouring money into the last click while starving the thing that actually introduced the customer. This isn’t about finding the one true answer, because there isn’t one. It’s about understanding the trade-offs well enough that you stop fooling yourself, and spend based on a fuller picture than the default report hands you.
What attribution is
Marketing attribution is the practice of assigning credit for a sale to the marketing touchpoints that led to it. Every customer usually interacts with you several times before buying, and attribution is how you decide which of those interactions deserves the credit, and how much. It sounds like bookkeeping, but it directly shapes strategy, because the channel you credit is the channel you fund. Get the credit assignment wrong and you systematically starve whatever your model happens to undervalue, no matter how important it really was to the customer’s journey.
The customer journey is messy
The reason this is hard is that real journeys are tangled. People bounce between your ad, your social posts, a friend’s recommendation, a search, and an email over days or weeks before they buy. Some of those touches you can track, and many you can’t, because a conversation over coffee or a glance at a billboard leaves no data trail. So before you even choose a model, accept that you’re working from a partial map of the journey, and any credit you assign rests on the touches you happened to be able to see.
Last click attribution
The default almost everywhere is last click, which gives all the credit to the final touch before the purchase. It’s simple, and it’s what most tools show you by default. It’s also badly biased, because it hands all the glory to whatever happened to be last, usually a branded search or a direct visit, while ignoring everything that made the customer want to come in the first place. Last click makes your bottom-of-funnel channels look like heroes and your awareness channels look useless, which quietly leads you to defund the very things that fill the top.
First click attribution
The mirror image is first click, which gives all the credit to the very first touch that brought a customer into your world. This flatters your awareness channels and ignores everything that actually closed the deal. It answers a genuinely useful question, where do my customers first discover me, but it’s just as lopsided as last click, only in the opposite direction. Neither extreme is right, because a real sale is almost never the work of a single touch. Both models pick one moment and pretend the rest of the journey didn’t happen.
Linear attribution
A fairer middle ground is linear attribution, which splits the credit evenly across every touch in the journey. If a customer had four interactions before buying, each gets a quarter of the credit. This at least acknowledges that the whole journey mattered, not just one end of it. Its weakness is treating every touch as equally important, when clearly some moments matter more than others. A throwaway glance at a social post isn’t really equal to the email that finally convinced them. Still, for a small business, linear is often a saner default than either extreme.
Time decay and position models
There are smarter splits. Time decay gives more credit to touches closer to the purchase, on the logic that recent interactions did more of the convincing. Position-based models give extra weight to the first and last touches, the introduction and the close, and split the rest among the middle. These are reasonable attempts to match how influence probably works, but notice they’re all just different guesses about how to divide credit. None of them knows the truth. They’re opinions about human behavior, dressed as arithmetic.
Why no model is correct
Here’s the honest heart of it. No attribution model is right, because the true influence of each touch is unknowable. You can’t rerun a customer’s life without the ad to see if they’d still have bought. Every model is a simplifying assumption about something you fundamentally can’t observe. This isn’t a flaw in the tools, it’s the nature of the problem. Once you accept that attribution is a set of useful fictions rather than a measurement, you stop chasing the perfect model and start using several of them to triangulate.
Use several models together
The practical move is to look at more than one model at once and see how the story changes. If a channel looks great under last click but terrible under first click, you learn it’s a closer, not an introducer. If another looks strong under first click and weak under last, it’s an introducer that needs help closing. The disagreement between models is itself the insight. No single view is the answer, but the pattern across views tells you what role each channel actually plays in the journey.
The holdout test
The one method that comes closest to truth isn’t a model at all, it’s an experiment. Turn a channel off for a while, or hold it back from a portion of your audience, and watch what happens to your overall sales. If turning off a channel barely dents your revenue, it wasn’t doing as much as its attributed credit suggested. This is more disruptive and slower than reading a report, but it measures real incremental impact rather than assigned credit, and for a big spending decision it’s worth the temporary discomfort.
The small business reality
For most solo operators, the elaborate models are overkill anyway, because your data is too thin for their precision to mean much. Often the most honest question you can afford to answer is simply, how did you hear about us, asked directly at checkout or in a quick survey. It’s imperfect and people misremember, but a rough signal you actually collect beats a precise-looking number built on tracking you can’t really see. Sometimes the low-tech answer is the most truthful one available to a small business.
Beware double counting
One trap to watch as you add channels and tools. Each platform loves to claim credit for the same sale, so your ad tool, your email tool, and your analytics can each report the conversion as theirs. Add up what every channel claims and you’ll find you apparently sold your product two or three times over. This double counting makes every channel look more effective than it is. A single, consistent source of truth for conversions, even a rough one, keeps you from funding channels based on credit they all claimed at once.
What attribution can’t do
The limit, stated plainly. Attribution can inform where you spend, but it can’t prove that spending more on a credited channel will produce more sales, because credit isn’t the same as cause. It works from the touches you can see, misses the ones you can’t, and rests on assumptions you can’t verify. Treat it as one input to a judgment call, never as a guarantee of return, and never as a precise measurement of something that is, at bottom, unmeasurable.
The view-through blind spot
A huge share of influence never registers as a click at all. Someone sees your ad, doesn’t click, and searches for you directly two days later. Every click-based model credits that final search and gives the ad nothing, even though the ad did the real work of planting the idea. This view-through influence is genuine and almost invisible to tracking, which is why channels that build awareness are chronically undervalued by the tools. Remember that the touches you can measure are only the ones that happened to leave a click behind, and plenty of real influence leaves none at all.
Privacy has broken the tracking
It’s worth knowing that the ground under attribution has shifted. Tighter privacy rules, cookie restrictions, and people moving across devices have made the detailed cross-site tracking these models rely on far less complete than it once was. Gaps in the data are now the norm, not the exception, so even your best-looking attribution report is built on a partial picture. This isn’t a reason to give up, it’s a reason to lean harder on experiments and direct questions, which don’t depend on the fragile tracking that privacy changes keep chipping away at every year.
Match the effort to the spend
Be practical about how much attribution rigor a decision actually deserves. If you’re deciding whether to keep spending fifty dollars a month on a channel, don’t build an elaborate multi-touch model to justify it, just make a reasonable call and move on. Save the real effort, the holdout tests and careful analysis, for the channels where you’re spending enough that being wrong genuinely hurts. Matching the depth of your attribution work to the size of the spending decision keeps you from drowning a small choice in analysis it never warranted.
Track the trend, not the perfect number
Since no attribution number is truly accurate, stop chasing the perfect one and watch the trend instead. Pick a consistent model, imperfect but the same every month, and follow how each channel moves within it over time. A channel steadily climbing under a fixed model is genuinely improving, even if the absolute credit is fuzzy. Consistency over time turns unreliable numbers into a useful signal, because the errors stay roughly constant and the movement is real. The direction a channel trends is far more trustworthy than the exact credit any single report assigns it on any given day.
Marketing attribution assigns credit for a sale across the touches that led to it, and every model, from last click to linear to time decay, is a useful fiction rather than a measurement. Read several together, run the occasional holdout test, and ask your customers directly how they heard about you. That combination will get you further than trusting whatever your default report happened to credit last.
For more breakdowns like this one, come back to the homepage.
Get new guides and videos first — join the Telegram channel.