Marketing Attribution for Energy B2B: Measuring the 12 to 24 Month Sale

The instrument was built for a different cycle

Marketing attribution is the practice of assigning credit for a sale to the marketing touches that influenced it. Almost every tool that does this was designed around a software buying pattern: a person feels a problem, researches for a few weeks, clicks an ad or a nurture email, and converts inside a quarter. In that world a 30 or 90 day attribution window captures most of the journey, and last touch is a crude but survivable shortcut.

Energy inverts every assumption in that sentence. A capital procurement decision, a framework agreement, a plant contract, a multi year service scope, routinely runs 12 to 24 months from first awareness to signature, and often longer when it is tied to a Final Investment Decision. Apply a 90 day window to a 20 month cycle and you have deleted the first seventeen months of influence by construction. The model then credits whatever happened in the final quarter, which is usually a demo request or a tender response, and declares that to be what won the deal. It is measuring the last visible step of a decision that was effectively made much earlier.

The buying group makes it worse. Forrester's State of Business Buying 2026 reports a typical B2B purchase now involves 13 internal stakeholders and 9 external influencers. Gartner's research on the buying journey finds buyers spend just 17 percent of their total time meeting with potential suppliers, and when weighing several vendors, only 5 to 6 percent with any one of them. Attribution software can only credit touches it can tie to a known contact. When more than twenty people influence a decision and most of them never fill in a form, the trace is broken before it starts.

So the failure is not that a particular model is badly tuned. It is that the whole apparatus, short windows, deterministic touch tracing, last click credit, was engineered for a cycle that energy does not have. Reading a working energy programme through that instrument produces a confident, precise and wrong number.

Project 54A long energy sale behaves like a walk down a plant gangway. The distance is the point, and the measurement has to span the whole span, not the last step.

You are mostly paid to influence people who cannot convert yet

Professor John Dawes of the Ehrenberg-Bass Institute published the finding that has reframed B2B measurement more than any other in the last five years. At any given moment, only around 5 percent of business buyers are in market, ready to buy now. The other 95 percent are out of market and will not buy for months or years. The LinkedIn B2B Institute carried the finding into practice.

In long cycle energy categories the arithmetic is arguably starker. When a service master agreement runs five to ten years and the underlying asset runs decades, the share of your addressable accounts that are genuinely in a buying window at any instant is plausibly below 5 percent. We treat that as an inference from contract tenor, not a measured figure.

The measurement consequence is direct and uncomfortable. Most of what marketing does, most of the year, is build memory and preference in buyers who cannot convert yet. That work is real, it is what puts you on the shortlist eighteen months later, and it is almost completely invisible to any attribution model that only credits in market clicks. If your dashboard rewards only demand capture, it will systematically defund the demand creation that fills the pipeline in the first place. This is the mechanism behind the familiar and destructive cycle where brand budgets get cut because they cannot be attributed, and pipeline quietly dries up two quarters later.

This is why a serious energy marketer measures two different jobs. Demand creation, aimed at the 95 percent, is judged on reach, memory and share of search over quarters. Demand capture, aimed at the 5 percent, is judged on conversion and pipeline over weeks. Collapsing both into one attributed ROI number guarantees you mismanage at least one of them.

Every single path model is a story you tell about incomplete data

It helps to be precise about the options, because most disappointment with attribution comes from expecting a model to do something it structurally cannot.

Single touch models assign all credit to one interaction. Last touch credits the final click before conversion, which in energy is almost always a tender portal or a demo form, so it systematically over rewards the sales team's own late stage activity and blinds you to everything that built the shortlist. First touch credits the initial known interaction, which over rewards whatever channel is cheapest at generating an early form fill and ignores the twenty months of influence in between. Both are single points of failure dressed up as insight.

Multi touch models spread credit across several touches, linear (equal weight), time decay (more to recent), or position based (weighted to first and last). These are more honest about the journey being multi step, but they share a fatal dependency: they can only distribute credit among touches they can see and tie to a known contact. Cookie deprecation and platform privacy controls have cut usable user level identity coverage to roughly 30 to 60 percent of the journey, down from more than 90 percent in the cookie era. A multi touch map built on a shrinking minority of the touches is a precise account of a fraction of the truth.

Data driven or algorithmic attribution uses machine learning to assign weights from observed conversion patterns. It is the best of the touch based family and the most oversold. It still cannot credit a touch it never observed, it needs a high volume of conversions to train on, which long cycle energy categories with a few hundred accounts simply do not produce, and it treats an unobservable dark funnel as if it did not exist.

The uncomfortable summary is that every touch based model, however sophisticated, is a narrative constructed over the subset of the journey your tools happened to capture. In a short cycle, high volume, fully digital purchase, that subset is large enough for the narrative to be useful. In a long cycle, low volume, human and offline energy purchase, the subset is small enough that the narrative is mostly fiction. The fix is not a better single model. It is to stop relying on any one of them.

Triangulation, not a silver bullet

The teams shipping pipeline numbers a finance director will sign off on have stopped searching for the one correct model. They run several imperfect methods in parallel, each strong where the others are weak, and reconcile them. The 2026 consensus term for this is triangulation, and it has four legs.

Marketing mix modelling is the strategic backbone. MMM uses statistical modelling to link marketing activity and spend to business outcomes at an aggregate level. Crucially it needs no cookies, no device IDs and no user level tracking, which makes it the natural privacy era answer to degraded multi touch attribution, and it can measure brand and out of home activity that touch tracking never could. Google's open source release of its Meridian model in late 2024 dropped the cost of entry from a six figure consulting engagement to a few weeks of in house data science work. For an energy company with long history and lumpy spend, MMM is how you decide next year's budget split.

Incrementality testing is the causal check. Attribution and MMM both infer; a holdout or geo test proves. Turn a channel off in a matched region, or hold back a matched set of accounts, and measure the difference in pipeline. It is the only method that answers the CFO's real question, which is not who gets credit but what would we have lost if we had not spent this. Reserve it for the decisions that carry real budget.

Pipeline and revenue attribution, run in your own warehouse, is the tactical layer. Warehouse first tools connect anonymous behaviour to known accounts and visualise long, complex journeys far better than the last click reporting inside an ad platform. This is where multi touch attribution earns its keep, optimising campaigns week to week, provided you remember it sees only part of the journey.

Self reported attribution recovers the dark funnel. Ask every inbound enquiry a single question at the point of contact: what made you get in touch now. HockeyStack and peers have made this a standard field, and the honest guidance from those same vendors is the point that matters most: self reported attribution alone is just as misleading as first or last touch alone, because it captures only the most remembered or misremembered touch. It has to be paired with the rest of the journey, never used on its own. But it is the only practical instrument that can see a peer recommendation, a conference conversation or a podcast that no pixel ever recorded.

None of these four is correct. Together they bracket the truth. When MMM, incrementality and self reported attribution all point the same way, you can allocate budget with confidence. When they disagree, the disagreement itself is the finding, and it is telling you where your measurement, or your strategy, has a blind spot.

Two ledgers: the machine and the money

The fastest way to lose a measurement argument with finance is to present a single blended return on ad spend number for a business with a two year cycle. It is either flattering and unbelievable, or honest and alarming, and in both cases it hides the two things a board needs to see. Split the report into leading indicators that prove the engine is running and lagging indicators that prove it pays.

Leading indicators prove the machine runs, months before revenue can. Reach into the target account list, the share of the 95 percent you are actually touching. Share of search, the most accessible proxy for mental availability and a decent leading indicator of market share. Pipeline created and its rate of creation. Qualification and shortlist coverage, the share of relevant tenders and buying groups where you are actually present. These move first, and in a long cycle they are the only evidence for a year or more that the strategy is working.

Lagging indicators prove it pays, and they are what finance ultimately underwrites. Sourced pipeline, meaning opportunities marketing originated, and influenced pipeline, meaning opportunities marketing touched, reported separately and honestly, because conflating them is how the credibility gets destroyed. Shortlist entry rate, which in a tender driven sector is close to the whole game. Win rate and sales cycle length, split by whether marketing was involved. And marketing sourced and influenced revenue, reconciled against MMM rather than asserted from last click.

One discipline protects the whole report. Never present influenced pipeline as if it were sourced revenue. The credibility of energy marketing measurement is destroyed more often by overclaiming than by underdelivering, because a finance team that catches marketing crediting itself with a deal the sales director owns will discount every number that follows. Report the honest, defensible, triangulated figure, flag every estimate as an estimate, and you build the one thing that actually protects the budget, which is trust in the measurement itself.

For the shape of the pipeline that feeds these numbers, our pipeline velocity framework sets out the four levers, and our dossier on buying signals in energy covers the causal, public events that actually trigger a purchase and should anchor any account scoring model.

The observable funnel is thinning, so causal and self reported signals matter more

The trend that will reshape energy attribution over the next two years is the migration of research into answer engines and away from clickable web pages. Gartner's March 2026 survey of B2B buyers found 67 percent now prefer a rep free experience, up from 61 percent a year earlier, and 45 percent used generative AI tools during a recent purchase. Every one of those AI mediated research sessions is a touch that influenced the decision and emitted no click your analytics could capture.

This is the quiet crisis for touch based attribution. Multi touch models depend on buyers browsing pages, opening emails and clicking ads. As synthesis moves into a chat window, the observable substrate thins, and the identity coverage that was already down to a minority of the journey erodes further. The rational response is not to buy a more sophisticated touch tracker. It is to lean harder on the methods that do not depend on the click, MMM, incrementality and self reported attribution, and to make yourself visible inside the model's answer in the first place.

That last point connects measurement to a new discipline. If a buyer's shortlist is increasingly assembled by an AI before any human visits your site, then being cited by the model is now part of being in the consideration set. That is a measurable objective in its own right, and we cover the practice in our dossier on generative engine optimisation for energy B2B.

Gartner also flags a second order effect worth watching. It found that buyers who research alone report high rates of purchase dissatisfaction and inconsistency between what they read and what they later learn, and predicts a partial swing back toward valuing human guidance by 2030. For measurement, the implication is that self reported attribution and post sale interviews will become more valuable, not less, because they are the only instruments that capture the human validation steps, the peer calls, the reference checks, the conference conversations, that increasingly decide a rep free purchase.

The one line strategy: as the funnel becomes less observable, stop trying to observe more of it, and start proving causation and asking buyers directly. The energy marketer who wins the measurement argument in 2027 is not the one with the most complete click map. It is the one who can show, through triangulated and honest evidence, that the marketing system is filling the shortlist.

Seven failure modes

Trusting last touch because it is the default in the ad platform. It will consistently credit the tender response and the demo form and tell you to defund everything that put you on the shortlist. It is the single most expensive habit in energy marketing measurement.

Applying a 90 day window to a 20 month cycle. If your attribution window is shorter than your sales cycle, you are not measuring your sales cycle. Match the analysis horizon to the real buying horizon, or accept that the number is meaningless.

Judging demand creation by demand capture metrics. Asking brand and category work to show last click ROI guarantees it looks like a failure and gets cut, after which pipeline dries up two quarters later and nobody connects the two events.

Chasing a single perfect model. There is no attribution model that is correct in energy. The teams that succeed run several imperfect methods and triangulate. The teams that fail keep buying the next tool that promises the one true number.

Overclaiming influenced pipeline as sourced revenue. The moment finance catches marketing crediting itself with a deal sales owns, every subsequent number is discounted. Underclaim before you overclaim.

Ignoring the dark funnel because it is not in the dashboard. The peer recommendation, the conference conversation and the podcast are often the touches that actually won the deal. If you do not ask, you will never see them, and you will keep crediting the last email instead.

No owner, no cadence, no honesty about estimates. A measurement system without a named owner becomes nobody's job, without a reporting cadence becomes a scramble, and without explicit labelling of what is measured versus modelled becomes a liability the first time a number is challenged in a board meeting.