The Tender Was Decided at FEED, Not at the Bid

Price competition happens inside a field the specification already chose

The claim is narrow and it is documented rather than rhetorical. On an energy capital project, the set of suppliers who can win is fixed by technical documents written during front end engineering design, and the tender that follows distributes work inside that set. Competing harder on price at tender is competing for the remaining margin in a contest whose entry list closed earlier.

Operators do not run capital projects as a single procurement event. They run a phased, gated front end. The naming varies by operator, decision gate zero to four in the Norwegian tradition, front end loading one to three in the Independent Project Analysis model, identify and assess and select and define in others, but the shape is consistent: appraise the opportunity, select a concept, define the design, authorise, then award and execute. Teaching material from NTNU's field development course describes the pre-engineering and FEED gate as the point at which the final technical solution is defined and the basis for awarding contracts is established.

The mechanism by which vendor freedom closes is written into a public standard. AACE International Recommended Practice 18R-97, the cost estimate classification system for the process industries, states that the level of project definition correlates with the percentage of engineering and design complete, and its estimate input checklist and maturity matrix tracks each deliverable through the classes as none, started, preliminary or complete.

Two rows of that matrix carry the argument. Specifications and datasheets are started at Class 4, preliminary to complete at Class 3, and complete at Class 2 and Class 1. Contracting strategy is assumed at Class 5 and Class 4, preliminary at Class 3 and defined at Class 2. A Class 3 estimate is described as prepared to form the basis for budget authorisation, appropriation or funding. A Class 2 estimate is, in AACE's own words, often used by contractors as the bid estimate to establish contract value, and its input list already includes vendor quotations.

Put those together and the sequence is unambiguous. Before a bid estimate exists, the datasheets are finished and vendors have already been asked for numbers. The procurement machinery then enforces the outcome. Technical bid evaluation precedes commercial evaluation, and grounds for technical rejection routinely include deviation from mandatory material requisition requirements and simple absence from the client's approved vendor list. Price is a conversation among the survivors of a technical filter.

Project 54By the time steel is in the ground, the specification that chose the equipment was signed off months earlier.

Read the estimate class and you can read the door closing

The useful discipline is to stop thinking in terms of sales stages and start thinking in terms of estimate class, because estimate class is the operator's own measure of how much of the design is already decided. It is published, it is standardised across the process industries, and the project team will tell you which one they are working to if you ask.

The table below maps the stage gates to AACE's classes, to the design maturity each implies, and to what has happened to the specification by that point. It is the single most useful thing a commercial team in this sector can put on a wall.

One caveat worth stating plainly. There is no published figure we could verify for the share of total project cost committed by the end of FEED, and readers should treat any confident number of that kind with suspicion. What is verifiable is the proxy chain: a conventional FEED itself costs roughly 2 to 3 per cent of total installed cost, on the account of Peter Laing of ABB Consulting, and by the Class 2 bid estimate vendor quotations are already inputs. The often quoted finding that 86 per cent of programme cost is determined before design freeze comes from research into unmanned aerial vehicle development published in Design Science, not from energy plant, and should be used to illustrate the mechanism rather than to size the energy case.

It is measured, and the industry is getting worse at it

Independent Project Analysis has benchmarked capital projects for decades and states that the completeness of front end loading is the single best predictor of safety, cost, schedule and operability outcomes, on a database of more than 25,000 projects. That is a strong claim, and it is IPA's own, but it is made against a benchmarking set with no real rival in the sector.

The distribution is the interesting part. IPA reports that more than 30 per cent of all projects pass the FEL 2 gate with incomplete FEL 2, which it links to underestimated capital costs, and that only 16 per cent of projects in its 2025 benchmarking sample entered authorisation with Best Practical front end loading. Aditya Munshi of IPA has put it bluntly: front end loading has not improved for more than a decade and has stayed in the fair range. Average schedule slip has roughly doubled to around 18 per cent, and execution speed has declined about 20 per cent over two decades.

The size of the prize is documented too. Work by Edward Merrow of IPA, cited by Owner Team Consult, found projects that reached authorisation at 18 to 30 per cent engineering complete averaged 3 per cent cost deviation and 7 per cent execution schedule slip. Projects authorised at limited or screening level definition averaged 29 per cent and 21 per cent. Operability problems ran at 11 per cent against 50 per cent, and success rates at 70 per cent against 20 per cent. Gibson and Hamilton, writing for the Construction Industry Institute in 1994, recommended that design work hours before authorisation run at 10 to 25 per cent of total design effort. Later CII research on the Project Definition Rating Index across 140 projects worth roughly five billion US dollars found that projects scoring 200 or lower outperformed those above 200 on both cost and schedule.

Merrow's own formulation is the one to keep. Front end loading, he wrote in Industrial Megaprojects, does not guarantee success, but it does enable it. Pam Wertz of IPA has made the commercial consequence explicit: committing to a project before the end of scope development adds substantially to the risk of cost growth and schedule slip.

For a supplier this matters twice over. It is the reason operators hold the design open long enough for you to influence it, and it is the reason a well evidenced total installed cost argument lands during FEED in a way it never lands at tender. The operator's own benchmark data says definition is where their outcomes are decided. You are not interrupting their process by turning up early. You are arriving at the only point where you are useful to it.

The shape is familiar, the timescale is not

Buying research outside energy has been converging on the same finding for a decade, and it is worth using carefully rather than as decoration. 6sense's 2025 Buyer Experience Report, based on nearly 4,000 buyers, found first contact with sellers now sits at about 61 per cent of the journey, that the eventual winner was already on the Day One shortlist 95 per cent of the time, and that the average cycle ran 10.1 months. Its 2023 study of more than 900 buyers found 84 per cent of deals were effectively decided by the time buyers made first contact, and that vendors who had not influenced buyers during the first two thirds of the process had roughly a 16 per cent chance of winning. Kerry Cunningham of 6sense summarises the mechanism: buyers rarely shift off their 70 per cent solution once that choice has been made.

Forrester's State of Business Buying 2024 adds the committee shape: an average of 13 people participate in an organisational buying decision, 89 per cent of purchases involve two or more departments, and 86 per cent of purchases stall. The LinkedIn B2B Institute, working with the Ehrenberg-Bass Institute, frames the same problem from the demand side in the 95-5 rule, that around 95 per cent of potential buyers are not in market today.

The honest caveat is that none of this research was conducted on energy capital project buying groups. It skews heavily to technology purchasing. The structural parallel is sound, because both describe a committee that forms a preference before it contacts suppliers, but the percentages should not be presented as if they were measured on EPC procurement. In energy the effect is almost certainly stronger, not weaker, because the preference is not merely a preference. It is written into a datasheet and countersigned by a technical authority.

There is also a live distortion pushing the decision even earlier. Equipment lead times in power have decoupled from project definition entirely. POWER Magazine reported in November 2025 that power transformer lead times averaged 128 weeks, switchgear 44 weeks, and gas turbines more than five years, and that GE Vernova opens its funnel with slot reservation agreements taking customer deposits on day one. Entergy had secured more than 19 gigawatts of generation components, including 4.5 gigawatts of power island equipment for 2031 to 2032. Brendan O'Brien of Burns and McDonnell describes the result as the traditional ready, set, go development cycle flipped to ready, go, set. When the equipment is reserved before the design is finished, the specification is being written around a machine that has already been bought.

Four stages, four different buyers, four different assets

The work is not more marketing earlier. It is different work, aimed at different people, producing different artefacts. At concept select the audience is the operator's discipline engineers, the study leads and the pre-FEED study contractor, and the job is to be a concept in the options register rather than a product in a brochure. That means offering an architecture comparison, reference designs and typicals that make your configuration the natural base case, and a shared cost technical study, sizing, heat and material balance, footprint and weight, that the study team can lift straight into their option register. The category manager is usually not in the room yet.

At pre-FEED the technical authority appears, which is the role that owns the standards and approves deviations, and often sits outside the project team entirely. Here the job is admission: get onto the approved vendor list before the material requisition is written, get named in the operator's specification as an acceptable manufacturer, supply preliminary datasheets in the operator's own template, and provide budgetary pricing with a stated validity period so that your numbers are the ones carried into the Class 3 estimate. Given transformer and turbine lead times, a delivery lead time letter is now a commercial document, not an administrative one.

At FEED the FEED contractor's discipline engineers are writing the datasheets and material requisitions that will become the tender. The objective is to be the source of the datasheet. That means an issued for design vendor data package the FEED engineer can build the requisition from, model files, interface and utility consumption data, support through HAZOP and layout reviews, spare parts and lifecycle cost data, and a total installed cost model that includes installation hours, foundation loads and power draw rather than equipment price alone. Then confirm, explicitly, that you appear on the final bidder list.

At invitation to tender there is no design influence left to win. The work is to avoid a technical exception and to make the commercial submission consistent with the total installed cost case already circulating inside the operator. Compete on compliance and on the schedule argument you built earlier.

None of this is free, and it should not be spread evenly. It is a portfolio decision: pick the operators and the project types where you can plausibly be in the options register, and accept that everywhere else you are bidding to fill capacity.

Track states, not events

Marketing qualified leads, form fills, demo requests and content downloads all measure one thing: the moment a member of a thirteen person, multi department buying group raises a hand. In energy capital projects that moment usually arrives at invitation to tender, which is Class 2, which is the point where datasheets are complete and vendor quotations are already inside the estimate. Optimising for lead volume is therefore optimising for arrival at the point of least influence. Quarterly attribution windows make it worse, because a multi year cycle guarantees that the last touch gets the credit and the first touch gets cut.

The replacement is a small set of state measures, tracked per project rather than per campaign. Specification presence rate: across the live projects you track, in what share is your product named or your architecture the base case in concept or FEED documentation. Approved vendor list coverage: what share of target operators list you, by equipment category. Datasheet origination: how many material requisitions in the period carry technical content that started with your data. Pre-tender engagement depth: how many named operator and FEED contractor engineers you have engaged per project before the tender date.

Two lagging measures make the case internally. Stage of first contact, recorded for every pursuit won or lost, tells you whether the distribution is moving earlier. Win rate split by stage of first contact tells you what that movement is worth. If the second number does not separate, the strategy is not working and you should say so rather than defend it.

One diagnostic is worth more than the rest combined. Count technical deviations raised against your bids at technical evaluation. A falling deviation count is direct evidence that your influence on the specification is real, because it means the specification is moving toward what you actually make. A flat deviation count with rising bid volume means you are getting better at answering documents written for somebody else.