Data-Driven (DD) Attribution

Static attribution models — First Touch, Last Touch, Linear, U-Shape — all apply one fixed rule to every customer journey. That's the problem.

Take a first-touch interaction that happened 60 days ago, after which the customer was completely inactive for 50 days before re-engaging and converting 10 days later. Is that first touchpoint really responsible for the conversion? Probably not — but First-Touch gives it 100% of the credit regardless.

Think of it like a football goal: every player in the move had some role, but their contributions weren't equal — and the value of each contribution changes from one goal to the next. Static models ignore this. They hand all credit to one player, or spread it evenly, regardless of what actually happened.

Data-Driven attribution solves this by evaluating each touchpoint individually and adjusting its weight based on what it actually contributed.

Take this example journey:

TikTok → Facebook Paid → Influencer → Branded Paid Search → Email Automation → Purchase

Data-Driven doesn't start from zero — it starts from a U-shape, reflecting the general principle that the first and last touchpoints tend to have the most impact on a conversion. This is the default, not the final answer.

For each touchpoint session, Klar looks at how much time the customer spent on the site. A session where someone spent 10 minutes browsing products is weighted significantly higher than one where they immediately bounced. This adjusts the baseline weighting up or down per session.

Beyond how long someone spent on a session, Data-Driven also looks at how much time passed between touchpoints. A very short lag — a customer clicking an email two minutes after visiting your site, right before checkout — often signals a mechanical trigger (e.g. a discount pop-up they just subscribed to) rather than a genuinely persuasive touchpoint, so it's weighted down. A touchpoint followed by a long stretch of inactivity (e.g. an ad clicked three months before purchase, with nothing happening in between) is penalised too — it never actually kicked off the near-term journey that led to the sale. A touchpoint that re-engages an otherwise dormant customer at the right moment, on the other hand, is weighted up.

Different channels play fundamentally different roles in a customer journey — broadly, some generate new demand, some amplify existing interest, and some simply capture demand that already exists elsewhere and Data-Driven accounts for this:

Branded paid search is boosted heavily down. A customer who types your brand name into Google and clicks an ad already knew your brand from somewhere else. That search didn't create the demand — it captured it. Giving branded search full credit would misrepresent where your marketing is actually working.

Email and owned channels (e.g. WhatsApp) are a good example of the time-lag adjustment from Step 3 in practice: a click two minutes after a site visit likely just redeemed a welcome discount mid-journey and contributed very little, while an email that brings back a customer after two months of silence deserves significant credit.

If a discount code is linked to a channel in Klar — via the Influencer CRM or the Influencer Sheet — Data-Driven uses it as an additional signal. Two things can happen:

This matters most for influencer marketing, where creators often drive conversions that never generate a trackable click.

When a customer completes a post-purchase survey and names a channel, Data-Driven uses that zero-party data:

This is how channels like Pinterest or organic social — which rarely generate trackable clicks — still get credited when customers say that's how they found you.

For the journey TikTok → Facebook Paid → Influencer → Branded Paid Search → Email Automation, a typical Data-Driven output might look like this:

Compared to static models or last-click, Data-Driven typically produces these directional shifts:

These are directional patterns, not benchmarks. The actual redistribution depends on your channel mix, customer journey length, and how well your PPS and discount codes are set up.

Data-Driven is the right model when you want to understand what's happening within your tracked journeys — which channels are contributing, which touchpoints are doing the work, and how specific ads or campaigns are performing. For the bigger question of where to allocate budget across channels — especially accounting for demand that didn't generate a trackable click — switch to MMM.

Whatever model you use, it's worth sanity-checking the output against blended metrics or incrementality tests before shifting real budget on it.

Data-Driven works on the touchpoints Klar can observe. When a customer journey consists of a single branded search click with nothing before it, Data-Driven will still attribute the conversion to that click — because that's all it can see. This is where MMM extends beyond it, modelling which channels likely created that demand in the first place.

If you're making cross-channel budget decisions and a significant share of your new-customer conversions sit in Direct or Branded Search, Data-Driven alone may undercount top-funnel channels. Use MMM for those decisions.

This is a structural limit of any click-based model, not just a Klar quirk: channels that create awareness without generating a trackable click — word of mouth, some organic social, offline touch points — are systematically undervalued. That's exactly the gap MMM is built to close.

Go to Tracking → Attribution → Attribution Deep Dive. Use the model selector at the top and choose Data Driven.

For background on how Klar tracks visitors and resolves identity in the first place, see How Klar Attribution Works.