Attribution is the rule that decides which partner gets paid: credit goes to whoever's link helped drive the sale inside the agreed window. Cookie windows set how long that credit lasts. Last-click still dominates because it is simple and universally understood; multi-touch is fairer in theory but needs clean data and a client willing to read it.
01What does attribution actually decide?
Who gets paid, and for which sale. Strip out the network jargon and affiliate attribution is one working rule: a partner gets credit if their link helped drive the sale within the agreed window. Everything else – cookie windows, click versus view tracking, last-click versus multi-touch, deduplication logic – is detail sitting on top of that single idea. If you understand the core rule, the rest is just how strictly and how broadly it gets applied.
It matters because attribution is the mechanism that turns a click into a commission payment. Get it wrong and two things go wrong with it: publishers who genuinely earned a sale do not get paid, which makes them stop promoting you, or publishers get paid for sales they did little to influence, which quietly inflates your commission spend without adding real growth. Attribution rules are not a technical afterthought – they are the commercial terms of the whole program, written into tracking logic instead of a contract clause.
02How do cookie windows actually work?
A cookie window is the length of time a partner's click stays eligible for credit after it happens. Someone clicks a partner's link, a cookie (or a device-level equivalent on platforms without cookies) records that click, and if the customer buys within the window that follows, the partner is credited. Buy after the window closes, and the click no longer counts – whoever gets credit next, if anyone, depends on what happened after.
There is no single correct window length, and treating one as a category standard is a mistake. The right window is a judgement based on your own purchase cycle, not a number copied from another brand's program. A few principles hold regardless of category:
- Match the window to how long people actually take to decide. An impulse category with a short path to purchase does not need the same window as considered categories where customers research, compare and come back days or weeks later.
- A window that is too short under-credits real influence. Partners who genuinely move a customer toward a purchase – content, comparison, review sites – often do their work well before the final click. Cut the window too tight and you stop paying for exactly the influence you want more of.
- A window that is too long over-credits low-effort activity. The longer the window, the more it favours whichever partner happens to get the last click before purchase – frequently a cashback or coupon site the customer was going to use anyway, regardless of whether that partner influenced the decision at all.
- Review it, do not just set it once. Purchase behaviour shifts as a program matures and as marketing mix changes elsewhere. A window set at launch and never revisited is one of the more common blind spots we see in program audits.
03Why does last-click attribution still dominate?
Because it is simple, and simple is genuinely valuable at scale. Last-click gives every sale's credit to whichever partner's link the customer clicked most recently before purchasing. Every major network supports it as the default, every publisher already understands exactly how they get paid, and when a dispute comes up, it is easy to check: pull the last click, confirm the timestamp against the window, done. That clarity is worth more to most programs than the theoretical fairness gains of a more complex model.
The trade-off is real, not imaginary. Last-click systematically over-credits the partner positioned closest to the point of purchase – often a cashback or coupon site – and under-credits partners who influenced the decision earlier in the journey, such as content or comparison publishers. Most brands accept that trade-off because the alternative brings its own cost, which is where multi-touch comes in.
| Last-click | Multi-touch | |
|---|---|---|
| Simplicity | High – one rule, easy to explain and audit | Lower – credit-splitting logic needs defining and maintaining |
| Data required | Standard network tracking is enough | Clean, consistent tracking across every touchpoint in the path |
| Who it suits | Most programs, especially newer or smaller ones | Larger programs with genuine multi-partner customer journeys |
| Watch-outs | Over-credits the last touch, usually cashback or coupon | Needs a client engaged enough to interpret and act on split reporting |
04When is multi-touch actually worth it?
When the customer journey genuinely involves multiple partners and you have the data quality to split credit between them meaningfully. Multi-touch attribution shares commission across more than one partner in the path to purchase, rather than handing it all to the last click. Done well, it can be a fairer reflection of who actually contributed – a content partner that built awareness, a coupon site that closed the sale, both credited for their part rather than one getting everything.
Done without the right conditions in place, it is complexity for no real benefit. Multi-touch needs clean data – tracking that reliably captures every touchpoint, not just the last one – and it needs a client who understands what the split reporting is telling them and is willing to act on it. Without both, you end up running a more complicated model that produces numbers nobody trusts and nobody uses to make decisions. That is a worse outcome than a simple last-click program that everyone understands.
Our honest read: most Australian programs are better served staying on last-click until volume and data quality genuinely justify the move. Multi-touch is an upgrade you grow into, not a default you should feel behind for not having on day one.
A good affiliate program is boring in the best way. Clean tracking, good partners, clear offers and consistent management.
05What does poor-quality activity look like inside attribution rules?
It looks like activity that technically satisfies your attribution rule while adding little or no real value – and it is worth watching for specifically because last-click rewards exactly this kind of behaviour. Two patterns show up most often in Australian programs.
Coupon poaching is a code appearing on a coupon site that was never actually issued or approved for that partner, sometimes an expired or generic code presented as if it were exclusive. A customer already intending to buy searches for a discount, clicks through the coupon listing, and the site claims last-click credit for a sale it did nothing to influence. Brand bidding is a partner bidding on your own brand terms in paid search, intercepting customers who searched for you directly – traffic you were going to capture anyway, now costing you a commission it did not need to.
Both problems sit outside what the attribution model itself can catch – they are a monitoring problem, not a tracking-window problem. The full list of what to check, and how often, is in our affiliate program audit checklist, which covers coupon quality and brand-bidding rules alongside the other seventeen checks that keep a program's commercials honest.
06How does Olivestreet set attribution rules for clients?
We start with fair rules and clean tracking, then keep watching. Attribution windows get set to match how a brand's customers actually buy, not copied from a template or left on a network default. Tracking gets tested end to end before it is trusted, the same way we approach it in any new program launch. And once a program is live, we watch for the poor-quality activity that exploits attribution rules – coupon poaching, brand bidding, validation gaps – as an ongoing part of management, not a one-off check.
Most Australian programs we work with are well served by last-click done properly, rather than multi-touch done badly. If a brand's volume and data genuinely support a more complex model, we will say so – but we would rather a client run a simple attribution rule that is clean and trusted than a sophisticated one that nobody on the team can actually interpret.