The landscape: five kinds of 'attribution tool'
“Attribution software” is a label stretched over genuinely different products, and most bad purchases start by comparing tools from different categories as if they were substitutes. The useful taxonomy:
- Free analytics with attribution built in — GA4. The default baseline every client already has.
- Ecommerce attribution platforms — Triple Whale, Northbeam. First-party pixels plus modeled multi-touch, built around Shopify-centric DTC.
- Click-tracking / funnel trackers — Hyros and kin. Long-window click and call tracking for high-ticket funnels, feeding corrected conversions back to ad platforms.
- Enterprise unified measurement — Rockerbox (acquired by DoubleVerify in 2025): MTA, MMM, and incrementality in one governed platform.
- Incrementality-first platforms — Measured: experiments as the primary source of truth rather than touchpoint tracking.
Note what that taxonomy implies: the categories differ not in UI polish but in measurement philosophy — whose numbers you'll be defending in the client QBR. For the methodological background on those philosophies, see MMM vs. MTA vs. incrementality; this page is about the software.
The comparison table
| Tool | Method | Indicative pricing | Best-fit client | Watch out for |
|---|---|---|---|---|
| GA4 | Data-driven & last-click attribution on consent-gated, partially modeled data | Free (GA 360 for enterprise needs) | Every client, as the shared baseline | Thresholding, consent gaps, no ad-cost truth outside Google without imports |
| Triple Whale | First-party pixel + selectable attribution windows/models, ecommerce analytics suite | Free tier; paid plans commonly hundreds of $/mo, scaling with revenue | Shopify DTC brands that want one operating dashboard | Pixel numbers will disagree with platforms and GA4 — pick a source of truth before the client asks |
| Northbeam | ML multi-touch on server-side first-party data + MMM-style modeling ('MMM+') | Roughly $1k+/mo, scaling with traffic/spend | Mid-to-large DTC spending ~$50k+/mo across several channels | Real onboarding lift; needs volume before the modeling stabilizes |
| Hyros | Long-window click/call/email tracking with 'print' attribution, pushes corrected conversions to ad platforms | From a few hundred $/mo, scaling with ad spend | High-ticket funnels: info products, coaching, calls-based sales | Philosophy is precise click-tracking, not modeled influence; dated UI; overkill for standard ecommerce |
| Rockerbox (DoubleVerify) | Unified MTA + MMM + incrementality with identity resolution | Enterprise custom (typically five figures/yr and up) | Enterprise omnichannel brands incl. offline media | Procurement-grade sales cycle; more than mid-market clients need |
| Measured | Incrementality experiments (geo, audience) as primary method, with MMM | Enterprise custom | DTC/omnichannel spending enough to power continuous experiments | Experiment-led means slower answers; requires org buy-in to act on holdout results |
Tool by tool
GA4 — the default everyone shares
GA4 is where every attribution conversation starts because it's free, ubiquitous, and integrated with Google Ads. Its data-driven attribution is a legitimate methodology (we dissect it in the models guide), and custom channel groups let you do genuinely useful AI-era segmentation (setup here).
- +Free, universal, and the lingua franca of client reporting
- +Data-driven attribution across Google and non-Google touchpoints it can observe
- +Custom channel groups handle AI referral segmentation well
- +BigQuery export enables serious custom analysis at zero license cost
- −Consent gaps and thresholding quietly shrink the observed dataset
- −Modeled conversions blend measurement and estimation without clear seams
- −No independent view of ad spend outside Google without manual imports
- −Google grading Google: structural conflict of interest on paid search
Triple Whale — the Shopify operating dashboard
Triple Whale grew from a Shopify profit dashboard into a broad ecommerce analytics suite: its own first-party pixel, selectable attribution models and windows, creative analytics, and an AI assistant (“Moby”) for querying the data. Its center of gravity is operational clarity for DTC founders — blended ROAS/MER, profit after costs, cohort views — with attribution as one feature among several rather than the entire product.
- +Fast time-to-value on Shopify; strong out-of-the-box dashboards
- +First-party pixel recovers some post-iOS14 signal that platforms lose
- +Whole-business view (profit, LTV, creative) rather than attribution in isolation
- +Accessible pricing path from small brands upward
- −Shopify-centric: value drops quickly off-platform
- −Its pixel's numbers are another opinion, not ground truth — expect three-way disagreements with GA4 and ad platforms
- −Attribution methodology less transparent than open documentation would allow
Northbeam — modeled MTA for scaling DTC
Northbeam pairs server-side first-party tracking with machine-learning attribution models and an MMM-style layer, aimed at brands where multi-channel budgets justify modeling over rule-based credit. It is the archetype of the “serious DTC measurement stack” purchase: powerful, opinionated, and demanding — of budget, data volume, and analyst attention.
- +Server-side collection is more resilient to browser privacy limits
- +Multiple model lenses (click-only vs. modeled) encourage healthy skepticism
- +Built for multi-channel scale where GA4 and native pixels diverge most
- −Price and onboarding effort put it out of reach below mid-market
- −Modeled numbers require trust in a proprietary black box
- −Teams sometimes buy it hoping for certainty; it delivers better estimates, not truth
Hyros — long-window tracking for high-ticket funnels
Hyros dominates a specific niche: businesses selling via calls, webinars, and multi-week email sequences, where the money moment happens far from the ad click and standard pixels lose the thread. Its “print” tracking follows identities across long windows, ties phone sales back to ads, and feeds corrected conversion data to ad platforms to improve their optimization.
- +Best-in-class for call/high-ticket funnels with long lags
- +Feeding corrected conversions back to ad platforms improves algorithmic bidding
- +Email and organic touch tracking most ecommerce pixels ignore
- −Click-identity philosophy inherits every blind spot of click tracking (view-through, AI-assistant influence, dark social)
- −Pricing scales with spend; ROI case is weakest for standard ecommerce
- −Interface and reporting feel dated next to the DTC suites
Rockerbox — enterprise unified measurement
Rockerbox, acquired by DoubleVerify in 2025, represents the enterprise consolidation thesis: MTA, MMM, and incrementality testing reconciled in one platform, spanning digital and offline (TV, audio, direct mail). For organizations with measurement governance requirements — and budgets to match — the pitch is one vendor, one methodology conversation, one source of reconciled truth.
- +Triangulation (MTA+MMM+experiments) as a product, not a project
- +Offline channel coverage most mid-market tools lack
- +Governance and support model suited to enterprise procurement
- −Cost and complexity far beyond mid-market needs
- −Unified platforms trade methodological transparency for convenience
- −Post-acquisition roadmaps warrant watching for strategic drift
Measured — incrementality as the product
Measured inverts the standard architecture: instead of tracking touchpoints and modeling credit, it runs a continuous program of geo and audience experiments and uses the results to calibrate channel-level reporting. Philosophically it's the purest expression of “stop asking what touched the conversion, ask what caused it” — with the costs that entails: experiments take time, consume test budget, and answer one question at a time.
- +Causal answers — the only category on this page that truly provides them
- +Immune to tracking signal loss by design
- +Excellent counterweight to platform-reported ROAS inflation
- −Needs conversion volume and spend to power experiments
- −Not a daily optimization tool; cadence is weeks, not hours
- −Enterprise pricing
The B2B corner
Everything above is consumer-centric. Agencies with B2B SaaS clients face a different problem — long cycles, committees, and journeys dominated by untrackable touches (communities, podcasts, peer recommendations, and increasingly AI assistants). The dedicated stack there: Dreamdata and HockeyStack (account-level journeys stitched to CRM revenue), Factors.ai (similar, with intent-data angles), alongside CRM-native attribution in HubSpot. All are directionally useful and all inherit the click-tracking blind spots; the correction that matters most in B2B is a mandatory self-reported attribution field — in B2B the gap between what buyers report (“heard on a podcast, asked ChatGPT, saw it in a community”) and what click paths show is the largest in all of marketing.
Choosing by client profile
| Client profile | Sensible stack | Reasoning |
|---|---|---|
| Shopify DTC, < ~$100k/mo spend | GA4 (clean setup) + Triple Whale | Operational clarity per dollar; attribution precision beyond this rarely changes decisions at this scale |
| DTC, multi-channel, $50k–500k/mo | GA4 + Northbeam (or Triple Whale) + quarterly geo tests | Enough spend that modeled MTA plus periodic ground-truthing pays for itself |
| High-ticket / call-based funnels | GA4 + Hyros | Long-lag, off-platform conversions are precisely the gap Hyros exists to close |
| B2B SaaS | CRM-integrated attribution (Dreamdata/HockeyStack) + mandatory self-reported attribution | Click paths capture a minority of B2B influence; the self-reported field is non-negotiable |
| Enterprise omnichannel | Rockerbox or Measured (philosophy choice: unified vs. experiment-led) + retained analyst capacity | At this scale, methodology governance matters more than any feature list |
AI-era questions to ask every vendor
The category's weakest spot in 2026 is the one this series exists to examine: journeys that begin inside AI assistants, where there is no click to track (background here). Put these to every vendor on your shortlist and grade the candor of the answers as much as the content:
- “How do you classify referrals from ChatGPT, Perplexity, Gemini, and Copilot — out of the box, today?”
- “What's your position on journeys that start in an AI assistant and arrive as branded search? Where does your model put that credit?”
- “Can I ingest external signals — branded search volume, AI share-of-voice, self-reported attribution — alongside your tracked data?”
- “Which of your numbers are measured and which are modeled, and how would I tell them apart in the UI?”
- “How do you validate your attribution against incrementality experiments — and will you show me a case where your model was wrong?”
A vendor who answers the last one honestly is a vendor whose other numbers you can calibrate. A vendor who claims their model has no blind spots has just told you the size of theirs.
Frequently asked questions
Is GA4 good enough, or does my client need a paid attribution tool?+
For a majority of SMB clients, a well-configured GA4 — clean channel groups, AI referral segmentation, conversion imports — plus a self-reported attribution field answers the decisions that matter. Paid tools earn their cost when spend is high enough that better allocation recovers the fee (usually $50k+/mo), when the business model breaks standard pixels (calls, long lags), or when platform-vs-analytics discrepancies are causing real strategic arguments.
Why do Triple Whale, GA4, and Meta all report different numbers for the same campaign?+
Because they measure different things: different attribution windows, different credit models, different identity graphs, and different exposure to consent and browser privacy limits. Meta counts view-through conversions GA4 never sees; GA4 applies data-driven credit across channels; a first-party pixel sees traffic ad blockers hide from others. The discrepancy is structural, not a bug — the professional move is designating one source of truth per decision type and reporting the others as lenses.
Do any attribution tools track influence from ChatGPT and other AI assistants?+
Attribution tools can classify AI referral clicks (visits from chatgpt.com, perplexity.ai, etc.) if configured for it — but no click-based tool can see recommendations that never produce a click, which is the majority of AI influence. That layer requires proxies: self-reported attribution, branded search trends, and AI answer share-of-voice monitoring, which is a separate tool category we compare in the AI visibility guide in this series.
What does attribution software cost in 2026?+
Indicatively: GA4 is free; ecommerce suites like Triple Whale run from free tiers to several hundred dollars monthly scaling with revenue; Northbeam-class modeled MTA starts around $1,000/month; Hyros scales with ad spend from a few hundred monthly; enterprise unified or incrementality platforms (Rockerbox, Measured) are custom-priced, typically five figures annually and up. Most vendors scale pricing on spend, traffic, or revenue — model the cost at 2× current client scale before signing.
Should an agency standardize on one attribution tool across all clients?+
Standardize the framework, not the tool. One measurement philosophy (what's a source of truth, how experiments validate models, how AI-era blind spots are reported) applied across a small approved toolset per client category beats forcing Shopify DTC tooling onto a B2B SaaS client. Agencies that standardize the QBR narrative — attribution for tactics, experiments for truth, proxies for invisible influence — scale measurement without scaling chaos.
Stay informed on life in Europe as an expat. The Local delivers daily news, guides and essential info across 9 European countries. Sign up for the free newsletter at thelocal.com/free-newsletter