Attribution is the topic that creates more arguments inside SaaS marketing teams than any other. The reason is that every model is wrong in a particular direction, and people pick the model that makes their channel look best. That is not a bad faith decision; it is a structural problem with how attribution is taught.
The right way to think about attribution is to start from the way your customers actually buy and pick the model that distorts that reality the least. Below is a plain language tour of the common options and the SaaS scenarios each one fits.
First touch
First touch credits the very first interaction a contact had with your brand. Useful for understanding which channels are creating new awareness in net new accounts. Misleading for understanding which channels are converting researched prospects into closed deals.
For an early stage SaaS company that needs to know where to find new accounts, first touch is genuinely informative. For a later stage SaaS company where the same channels keep introducing new contacts at known accounts, first touch will overstate top of funnel discovery channels and understate everything that closes deals.
Last touch
Last touch credits the final interaction before conversion. Useful for understanding which channels are doing the closing work. Misleading because it makes branded search and direct traffic look heroic, when both are mostly capturing demand created elsewhere.
For most SaaS companies, last touch is the worst standalone model because it systematically overpays the channels that close deals and underpays the channels that create the conditions for those closes to happen.
Linear
Linear credits every touchpoint along the path equally. Useful as a counterweight to first or last touch when you want to acknowledge that every interaction did some work. Misleading because it pretends the customer thoughtfully weighted each touch the same way, which is not how human attention works.
Linear is a fine secondary lens. It is rarely the right primary model for a SaaS funnel because it flatters every channel equally and therefore guides few real budget decisions.
Time decay
Time decay credits later touchpoints more than earlier ones, with the weighting tunable. Useful for SaaS funnels with long consideration windows because it acknowledges that the touch a week before the demo had a bigger role in the outcome than the touch nine months ago.
Time decay is a sensible default for B2B SaaS once you have enough data to fit the decay curve to your actual sales cycle. The most common mistake is to leave it on the platform default, which is usually a seven day half life that is wildly too aggressive for B2B.
Position based
Position based credits the first and last touch heavily, with the middle touches splitting the remainder. Useful for funnels where awareness creation and closing are both genuinely valuable and the middle touches are mostly research behavior.
For a SaaS funnel where discovery and decision are the two moments that matter most, position based is often the cleanest single model. Default weights of forty, twenty, forty are a fine starting point.
Data driven
Data driven attribution uses your own conversion data to assign weights based on which paths are most associated with conversion. Useful when you have meaningful conversion volume and want the platform to do the math.
The catch is that data driven models inherit the biases in your data. If your tracking systematically misses certain channels or device handoffs, the model will too. Audit your tracking before you trust the output.
How to actually use these models
Pick one primary model that matches your customer journey and use it to make budget decisions. Use one or two secondary models to sanity check what the primary model is telling you. When the primary and secondary models disagree, the disagreement itself is signal worth investigating; it is rarely a tie that needs to be broken with a new model.
And accept that no model is correct. The goal is a model that is consistently wrong in known ways, so your team can adjust for the bias rather than chase a perfect answer that does not exist.
The right cadence
Recalibrate your attribution model annually, not monthly. Marketing teams that change models every quarter never accumulate the comparison data that makes attribution useful in the first place. Pick a model, defend it, and let it run long enough to learn from.
Want to talk about applying any of this to your account? Send us a note and a senior expert will reply within one business day.