Attribution Without a Pixel or Engineering Ticket

A practitioner's guide to measuring channels this week, why most tools need a pixel, and which pixel-free options actually see what you need.

Updated 8 September 2026 · Joris van Huët, founder, Causality Engine

If you want channel-level attribution running by Friday and you do not have a developer, the direct answer is this: yes, it exists, but the pixel-free tools split into two very different camps, and picking the wrong one wastes your week. One camp reads data you already have (a GA4 export, a CRM). The other builds a statistical model from aggregate spend and revenue over many months. They answer different questions and see different things.

The fastest route for most ecommerce brands is exporting a GA4 file and running a causal read on it. Causality Engine does exactly that: upload the export, get per-channel causal attribution with confidence intervals in about five to ten minutes, €99 for a single read or €299 per month for the Pro plan. No pixel, no code, no onboarding call. The tradeoff is real and I will name it up front: a GA4 export cannot see what GA4 never recorded, which means offline sales and platforms that live outside GA4 are simply not in the picture unless you feed them separately.

Why most attribution tools ask for a pixel or server-side tag

Conventional attribution follows events. To attribute a sale to a channel, the tool wants to observe the visit, the click, the session, and the purchase, then connect those to an ad platform's identifier. That observation is the job of the tag. Google's own documentation describes its tag as measuring page views, clicks, scrolls, conversions, and campaign performance. Meta's Conversions API sends website, app, messaging, and offline events from a server back to the platform.

So the pixel or server-side tag is doing two things at once. It records the journey for the attribution vendor, and it sends conversion signals back to the ad platforms for optimization. Meta says server events through its Conversions API are processed like browser pixel events and can be used for measurement, attribution, and delivery, and recommends running both together.

There is a third quiet requirement people forget: the identifier has to survive. A click ID like gclid or fbclid is only useful if something captures it, stores it, and later ties it to an order or lead. When it is not captured, the fallback is weaker matching on hashed email and phone, with lower match rates.

That is the whole reason "no-code attribution" is a category worth writing about. Every one of those steps normally means an engineering ticket.

What a pixel-free approach can and cannot see

Pixel-free does not mean magic. It means you are working with data that already exists in some system, and your visibility is bounded by that system.

Before you compare tools, measure the boundary. Take the conversions your analytics could attribute to any source and divide them by the orders your commerce platform recorded. That ratio is your coverage, and every attribution number you will ever read is the real one divided by it. If coverage fell from 0.8 to 0.6 over a year, your reported cost per acquisition rose a third from the decay of your own visibility alone, before any change in media price. In our own analytics warehouse, 48.6% of sessions between 17 October 2025 and 2 September 2026 had no resolvable source. That is one company's census, not a benchmark. Yours will differ, and no tool on this page, pixel or not, sees what your systems did not record.

There are two honest ways to skip the pixel:

• CRM matching. Attribute known leads and customers using CRM records, captured click IDs, or hashed contact details.
• Aggregate modeling (MMM). Estimate channel contribution from time-series spend, revenue, seasonality, and promotions, with no user-level data at all.

And there is a third path that is neither of those, which is reading an analytics export you already have:

• GA4-export causal analysis. Model the channels and sessions GA4 already recorded, using causal inference instead of last-click rules, without installing anything new.

Each sees a different slice of reality.

CRM matching sees identifiable people: leads, qualified leads, closed-won deals, revenue tied to a contact. It does not see the anonymous visitor who never filled in a form. That is not a defect. A CRM-first tool is measuring known contacts on purpose.

MMM sees the shape of the whole business over time, including offline and untrackable channels, but it never tells you which individual saw which ad. It trades granularity for coverage.

A GA4 export sees whatever GA4 recorded: paid channels that were tagged, organic, direct, referral, email when it is captured, and your conversion events. It does not see a sale that happened in a store, over the phone, on Amazon, or on a platform that GA4 never touched.

The pixel-free tools worth knowing

Four tools come up repeatedly when brands want measurement without a new tracking install. They are not interchangeable.

Tool	Method	What it needs	What it sees	Speed and cost	Main caveat
Causality Engine	Causal inference on a GA4 export	A CSV export from GA4 (also supports a Shopify + GA4 workflow)	Per-channel causal attribution, incremental ROAS, and confidence intervals across channels visible in GA4: paid, organic, direct, referral, email	5 to 10 minutes; €99 per read or €299/mo Pro with automated GA4 ingestion, AI chat, and a developer API	Only models what the export contains. Offline sales and platforms absent from GA4 need a separate ingestion path
Attribi pixel-less mode	CRM matching and offline conversion sending	Connected CRM and ad platforms; stronger matching needs forms that capture click IDs	Known leads, qualified and closed-won outcomes, CRM attribution fields, conversions sent back to platforms	Vendor says roughly 15 minutes using existing form code; hashed email/phone fallback when no click ID	Not a replacement for anonymous site-journey tracking. Quality depends on CRM completeness
Recast	Bayesian marketing mix modeling	Daily spend by channel and a daily KPI; recommends 18+ months of history, ideally ~27	Baseline, paid contribution, total outcome, channel-level contribution from aggregate data	No pixel; weekly ingestion. Needs substantial history and prepared input tables	Strategic aggregate model, not a user-level report. Data prep required
Cassandra	MMM plus broader measurement and planning	Aggregated spend, impressions, revenue, pricing, promotions, seasonality, external factors	Channel contribution across online and offline media, including Amazon and retail	No cookies or user tracking; public pricing starts at €2,700/mo	A larger MMM engagement, not a one-off read. Aimed at bigger data and budget bases

Read that table by intent, not by feature count. If your question is "which of my GA4 channels caused incremental revenue," the first row answers it in an afternoon. If your question is "how do I allocate next year's budget across TV, retail, and paid social," the bottom two rows are the right shape and the top row is not enough on its own.

The specific limitation of a GA4 export, said plainly

I want to be exact here because vendors are usually vague about it, and vagueness is what gets people burned in a budget meeting.

A GA4 export can support analysis of:

• Paid channels that are correctly tagged or otherwise visible in GA4
• Organic search, direct, and referral traffic
• Email traffic when GA4 recorded it
• Ecommerce or conversion events already in GA4
• Historical periods already present in the property

A GA4 export cannot, by itself, supply:

• Offline sales that never entered GA4
• Retail, Amazon, phone, or field-sales revenue
• Spend or activity from platforms not represented in GA4
• Untracked ad impressions
• A complete record of exposure when there was no click and no GA4 session
• CRM or finance-system data unless you join or upload it separately

This is why the GA4-export read is honest about its scope. It answers "given what GA4 already knows, which visible channels appear to have driven incremental revenue." It is not a business-wide model built from finance, retail, and offline data. If most of your revenue happens offline, no GA4 read will fix that, and any tool that implies otherwise is overselling.

There is a second limit that has nothing to do with pixels, and it applies to every method on this page. The effect you are trying to detect has to be larger than the noise your own business generates. Multiply a channel's share of revenue by the return you would honestly defend for it; that is roughly how much total revenue would move if the channel stopped. If that number is smaller than the smallest lift your data can distinguish from noise (about 8% for a typical DTC brand with six months of history and an eight-week test), the channel is not measurable at your scale, by anyone, and the honest move is to manage it on judgement rather than on a null result.

If you want to understand how the causal number differs from the ROAS your platforms report, the incremental ROAS explainer covers where the gap comes from. And if you are weighing an export read against a full model, the MMM explainer lays out what marketing mix modeling needs before it produces anything useful.

Why I keep recommending the export read for a one-week deadline

The pattern I see over and over: a brand owner walks into a budget conversation with three dashboards that disagree. Meta claims one number, GA4 claims another, and the finance spreadsheet claims a third. Nobody trusts any of them, so the budget stays frozen or moves on gut feel. Those three numbers are answers to three different questions, and only the finance one is a fact about money, because it is the only one the seller of the media did not produce.

The reason I lean toward a GA4-export causal read in that situation is not that it sees everything. It does not. It is that it needs nothing from your engineering team, it uses data you already have, and it puts a confidence interval on every channel estimate. A number with an interval is defensible in a way that a bare ROAS figure is not. You can say "paid social contributed somewhere in this range, here is the methodology" instead of "the dashboard said 4x."

The results back this up in ways I find more convincing than any accuracy claim. In the published causal read for Me Gorgeous, a Dutch DTC brand, around €2,000 a month of Meta spend showed no incremental contribution and was reallocated. Another brand was told by last-click to cut Pinterest, and the causal read kept it alive because the platform number was hiding the real contribution. Those are the two failure modes of platform attribution in one sentence each: paying for spend that does nothing, and killing a channel that quietly works.

One more thing, and it applies to us as much as to the other three tools in the table. Between 14 August and 2 September 2026 we audited thirty-one commercial measurement vendors for a published validation of their method against randomised experiments, with the sample, the design and the discrepancies disclosed. We found none. Hold Causality Engine to the same question. The read states its counterfactual and its interval, and it is a model on observational data, not an experiment. Ask us, and every other vendor, which of your channels are not measurable at your current spend. The honest answer is a list.

MMM tools like Recast and Cassandra catch a different problem. They see the channels GA4 cannot, which matters enormously if you run TV, radio, direct mail, or heavy retail. But they are not a Friday-afternoon answer. Recast's own guidance points to 18 months of history minimum, daily spend by channel, a daily KPI, and promotional calendars. That is a data project, not an upload.

Attribi sits in the lead-generation lane. If your business is about pipeline and closed-won deals rather than anonymous ecommerce traffic, its CRM-first approach is a genuine fit, and its pixel-less mode removes the website tracker install. Just do not call it click-level attribution unless your forms are actually capturing the click IDs.

Plain checklist for an answer this week

Fastest path: GA4 export

• Confirm GA4 is collecting your conversion or revenue event
• Confirm the channels you care about appear in GA4
• Compute your coverage: attributed conversions divided by orders in your commerce platform
• Export the last useful period as CSV, including date, channel, campaign where available, sessions or events, and revenue or conversions
• Note any sales that happen offline or on another platform
• Upload to Causality Engine and pay €99 for one read
• Review per-channel causal attribution and confidence intervals
• Do not describe missing offline sales or absent platforms as measured
• If it becomes a recurring process, weigh the €299/mo Pro plan

Lead-gen path: CRM without a pixel

• Confirm the CRM records lead source, contact details, stage, and deal value
• Connect the CRM and ad platforms in Attribi
• Use hashed email and phone matching if you cannot change forms
• Check whether existing forms already store gclid, fbclid, or another supported click ID
• If they do, map those fields in; if they do not, do not call it click-level attribution
• Send one test lead through the full journey and confirm it reaches the ad platform
• Treat anonymous non-lead traffic as outside this view

Strategic path: aggregate MMM

• List every channel, including offline
• Gather daily or weekly spend by channel
• Gather a clear daily or weekly KPI: revenue, orders, or leads
• Add promotion dates, price changes, launches, and major external events
• Check how much history you have
• Use Recast or Cassandra if the question is budget allocation, not individual journeys
• Expect data prep even though no pixel is required

FAQ

Can I really get channel attribution without installing a pixel?

Yes, but be specific about which kind. A GA4-export read and a CRM-matching tool both work without a new pixel because they use data you already collect. MMM tools work without one because they never touch user-level data at all. What none of them can do is invent touchpoints that were never recorded anywhere.

Does pixel-free mean no data setup?

No, and this is where people get tripped up. "No new pixel" is not the same as "no preparation." A GA4-export read is genuinely close to zero setup if GA4 is already collecting cleanly. A CRM tool needs a complete CRM. An MMM model needs many months of organized spend and KPI data. The pixel is only one of the things that can require engineering.

What does a GA4 export miss that a pixel would catch?

Less than you might expect for online channels, because GA4 already records most of your web sessions and conversions. Where the export falls short is anything outside GA4 entirely: offline sales, retail, phone orders, Amazon, and platforms you never connected. A pixel would not fix most of those either, which is why offline-heavy brands should look at MMM instead.

Is causal attribution better than last-click if I only have a GA4 export?

For most brands, yes, and it is the whole reason to bother. Last-click just credits the final touch it happened to see. A causal read on the same GA4 data estimates incremental contribution and attaches a confidence interval, which is what lets you defend a reallocation decision. Same input file, a far more useful output.

When should I skip all of this and use MMM instead?

When a meaningful share of your revenue happens where GA4 and your CRM cannot see it. Heavy offline media, retail distribution, or long consideration cycles all point toward Recast or Cassandra. Just budget for the data work and the timeline. MMM is a planning tool, not a same-week diagnostic.

Sources and further reading

• The Price of Being Found (Causality Engine, Edition 2.10, September 2026): the coverage rate (Chapter 8), the scoreboard (Chapter 9), the measurability floor (Chapter 15), the vendor audit (Chapter 17)
• Causal attribution for ecommerce brands
• Incremental ROAS explained
• MMM explained for DTC brands
• Sixty-second versions of these ideas on YouTube Shorts

Vendor prices and features quoted in this article were taken from each vendor's own website on 8 September 2026 and may have changed since. Check the vendor's pricing page before relying on a figure.