Who actually signs
Three people, and the sale is decided by the second one. The plant manager owns the number the site is measured on: yield, throughput, energy per ton, first pass quality, giveaway. That is your buyer.
The process engineer or technology manager owns the truth. They cannot sign anything and they can end you in one sentence at a meeting you are not in. They have watched an advanced control project decay to a low service factor and a corporate data science team deliver a dashboard nobody opened, and they will assume you are the next one until you prove otherwise.
Corporate operational excellence or digital manufacturing holds the budget line and the mandate. They will take your meeting fastest, which is the trap. A pilot funded by corporate at a site that never wanted it is the single most common way this sale dies at month nine.
There is a fourth person who is not a buyer and can still stop everything: the controls engineer who owns the distributed control system. To them, a model that changes setpoints is a safety question before it is an economics question.
The one sentence version
Your buyer is a plant manager who already hit the number once, on the best week of last quarter, with the same equipment and the same crew, and has no idea why that week was different.
How they think about it now, and where you need them
This is the part of the sale most vendors skip, and it is the whole thing. The plant is not undecided about your product. The plant holds a set of beliefs that make your product unnecessary, and every one of them was earned honestly.
What they believe today.
- Optimization is a capital problem. Getting more out of the unit means a bigger exchanger, a new column, a debottleneck study, and a place in a capital queue three years long.
- Control belongs to the control vendor. The distributed control system came with a name on it, the advanced control layer came from the same place or from a licensor, and anything touching setpoints is that vendor's territory.
- Our process is unique, so nobody's model applies to it. Twenty years of hard won knowledge sit in the heads of four people, and every outsider who arrived with a general answer has been wrong.
- Our data is a mess, so this is premature. The historian has gaps, two analyzers have been reading wrong since the last turnaround, and lab results do not line up with process time.
- AI means taking judgement away from the operator, which is the thing that keeps this unit from tripping.
What has to be true before they can buy.
- The problem is variability, not capability. The unit has already produced the better number. It did it on the good week, and nothing was purchased that week. The prize is not capacity, it is the distance between the good week and the ordinary one, and that distance is already in the historian.
- Uniqueness is the argument for the model, not against it. You are not carrying a model from another plant. You are training on their history, which is the only place their particular relationship between feedstock, ambient conditions and quality has ever been written down.
- The data does not need to be clean, it needs to be honest. A model that surfaces which instrument has been drifting is not a failed pilot, it is a finding the site can act on that week, and saying so before anyone asks is what separates you from the last vendor.
- Advisory comes before autonomous. The operator keeps the wheel. Closed loop is something the site earns after the recommendations have been right for months, not a switch you ask them to flip in the contract.
- The measure is what the plant kept. Not model accuracy, not a dashboard, not a percentage on a slide. The number the site already reports, moved, and still moved a quarter later.
Every email, meeting and proof point should be doing one of those five jobs. A sentence that is not moving one of those beliefs is a feature description, and it is costing you the reply.
The triggers, and where each one is visible
- Air permit applications and modifications. State environmental agencies publish them with the site, the unit, the throughput and the date. A plant filing to modify a permit is a plant changing how it runs, months before anyone announces anything.
- Capex and expansion announcements from the corporate parent, and the operations commentary in quarterly calls. These name the site and the number the company has publicly committed to, which is the number you should be writing about.
- Turnaround and shutdown schedules, visible in trade press and in contractor mobilization and local hiring. A turnaround is the one window when the historian gets touched, instruments get replaced and the budget is already open.
- Job posts naming the stack. A post for a process engineer, an advanced control engineer or a digital manufacturing lead that names a specific historian, control platform or manufacturing execution system tells you what you would integrate with and that someone approved headcount for the problem you solve.
- Energy and emissions commitments with a per unit target and a date, published in the parent's sustainability reporting. Somebody at the site now owns that number and has no plan for the last part of it.
- Ownership change. A carve out from a major or a private equity acquisition puts a new plant leader and a margin mandate in place at once, and the first hundred days is when unfamiliar ideas get heard.
- Conference papers and speaking slots. A process engineer presenting at a technical conference has told you what they are working on and, just as usefully, that their employer lets them talk to outsiders about it.
Permit filings and stack specific job posts are the two to build on. One tells you the unit is changing. The other tells you who to write to and in what language.
Qualify in sixty seconds
- Is there variability to harvest? Variable feedstock, seasonal ambient effects, grade changes and long campaigns are fuel. A short run job shop with tight inputs is not your account, whatever the headcount says.
- Is there a historian with a few years of data at a usable frequency, and can one identifiable person be named who knows where it lives? If nobody can be named, the pilot has no floor.
- Is the number owned at the site? Find out who is measured on yield or energy per ton at that plant. If the answer is only corporate, you have a pilot and not a customer.
- Is there an incumbent advanced control layer, and is anyone maintaining it? A decayed controller is a better opportunity than a blank sheet, and a completely different conversation from a greenfield one.
The angle that gets replies
Never lead with the technology. The plant manager received four notes about AI this week and deleted all of them, because none mentioned a unit, a feedstock or a number they recognise.
Lead with the reframe, applied to something specific and public about their site, and make the first ask small enough that saying yes needs neither a budget nor a data agreement.
Three openers you can adapt
- On a permit modification"Saw the permit modification filed for the reactor train at the plant. Anything that changes throughput usually widens the spread between your best month and your average one, because the operating envelope everyone learned no longer applies. That spread is already sitting in your historian and almost nobody reads it that way. Two paragraphs on how we quantify it, no data required from you."
- On a job post naming the stack"You are hiring a process engineer and the post names your historian and your control platform, which tells me the data is there and somebody wants something done with it. The first question we would ask is not what a model could predict. It is what your best quarter looked like against your average one on the same equipment. If that gap is worth more than the hire, it is worth twenty minutes."
- On a turnaround"With the turnaround coming up you will have instruments out and a rare chance to fix what the historian has been recording wrong. Most sites find the drifting analyzer afterwards. We run a short read on the last ninety days that flags which signals disagree with each other before the unit comes down, and it is useful whether or not you ever buy anything from us."
Each one names something real about their site, states the variability reframe in a sentence, and offers a small piece of work rather than a demonstration. Process engineers forward those. Nobody forwards a demo request.
What not to send
- A percentage improvement with no base case attached. The process engineer stops reading at the number, because a yield gain means nothing without the unit, the feedstock and the period it was measured over, and quoting one without them marks you as a marketer.
- The words AI powered in a subject line. Every plant manager in the industry receives that sentence weekly and has learned to archive on sight.
- A request for their process data in the first email. That is a legal conversation and an intellectual property conversation, and asking early tells them you have not done this before.
- Anything implying the operators are the problem. The operators are the reason the unit is still running, everyone at the site knows it, and one sentence in that direction ends the account.
- A dashboard screenshot. The site has dashboards. Nobody at a plant has ever bought one twice.
The objection you will hit
We already have advanced process control. Treat that as the qualification, not the objection, and say so. Controllers hold constraints against a model of the plant as it was when they were commissioned, and the plant drifts. Ask two questions: when were they last retuned, and what is the current service factor. An engineer who knows both numbers is a technical buyer worth pursuing. An engineer who has to go and find out has just discovered the gap themselves, which is far more persuasive than you describing it.
Our process is unique. Agree completely, then invert it. Uniqueness is the reason a model trained on their own history beats any general answer, and it is the reason the four people who understand the unit are the ones you want in the room. This is the objection most likely to become the reason they buy, and vendors keep arguing with it instead.
Our data is not good enough yet. Never argue, and never promise the data does not matter. Offer the readiness read as the first step, be explicit that a drifting analyzer or a lab timestamp misalignment is a real finding, and let them see that you will tell them the unwelcome version. Waiting for good data is a way of waiting forever, and the site usually knows it.
We tried a pilot and nothing came of it. The most common objection in this niche now, and the answer is a question: who owned the number. In almost every dead pilot, corporate owned it and the site was a venue. Design the next one so the plant manager reports the result, and say that out loud, because it is the difference the buyer has felt but never put into words.
Deal shape
- Data readiness review: two to four weeks, free or nominal. Produces a document the process engineer can circulate, and tells you whether the account is real.
- Paid pilot on one unit against one number: commonly $50K to $200K over three to six months. One unit, one metric, one named owner at the site.
- Site licence after a successful pilot: commonly $150K to $500K per site per year, priced on the units under model and the value of the metric rather than on seats.
- Multi site rollout: negotiated with corporate once a site is defending the result, priced per site with a declining rate and a services component for onboarding each unit.
- Signer: the plant manager for the pilot, corporate for the rollout, with the process engineer holding a veto at every stage. Cycle: six to eighteen months, and budgets are set in the autumn, so a proposal landing in January waits a year.
Two things about this shape are worth internalising. The pilot is not a discount, it is a qualification round you get paid for. And the rollout is a different sale to a different person, which means the day the pilot succeeds is the day you start selling again.
A cadence you can actually run
- Monthly, pull new and modified air permits in the states you can serve, and add the sites to a named account list rather than a lead queue.
- Weekly, pull job posts naming historians, control platforms and manufacturing execution systems at those sites and at their corporate parents.
- Quarterly, review turnaround schedules and mark the sites whose window is six months out, which is when the conversation is useful and not yet frantic.
- Continuously, watch for plant manager changes at accounts already on the list. A new plant manager inside ninety days is the best moment this niche offers, and it arrives on its own schedule.
- Twelve to twenty accounts a week is a full program. This universe is a few hundred sites, not a lead flow, so accuracy matters more than volume and a burnt account stays burnt.
- Three touches over two weeks, each with something new in it, then leave the account alone. The permit, the turnaround and the new plant manager will each give you a fresh and legitimate reason to write again.
The site already produced the better number once. Everything you are selling is the difference between that week and the rest of the year, and the evidence is sitting in a historian nobody has read that way.
The sending mechanics most people get wrong
Everything above is about who and what. This is about how, and it is where most outbound in this niche quietly dies. Seven rules. None of them are optional.
1.Three to five sentences. That is the whole email.
Your reader is on a phone between meetings. One observable fact about their company, one consequence they have not thought about, one specific thing you would do. Anything past five sentences is a memo, and memos get archived unread.
2.Lead with a technical differentiator that turns into a number.
The messages that work best name something concrete you do differently and translate it into time or money saved. In this niche the differentiator is the model and how it is deployed, so name it plainly: what class of model, whether it runs advisory or closed loop, what horizon it predicts over, and whether it reads their historian directly or needs a parallel stack. Then translate it into their units. Dollars per ton, tons per hour, points of first pass quality. A process engineer will forgive a modest number described exactly and will never forgive an impressive one described vaguely.
Most services firms do not have a technical differentiator, and pretending to have one reads as exactly that. The substitute is a verticalized case study: a company like theirs, what you did, what happened, in one sentence. For this niche the line is: a plant on the same unit type with the same feedstock variability, the metric that moved, the months it took, and whether it held through the next turnaround. The turnaround detail is the one that lands, because it answers the question every process engineer is actually asking, which is whether the result survives the plant changing. Get written permission to describe it at pilot close, while the plant manager is pleased.
3.Ten to twenty emails a day per mailbox. Not a hundred.
Sender reputation is scored per mailbox and per sending domain. One inbox pushing a hundred cold emails a day looks like exactly what it is, and the penalty lands on the domain, which means it lands on your client correspondence too.
If the math says you need more volume, the answer is more mailboxes on more warmed sending domains, separate from the domain you invoice from. It is never more volume per mailbox. Fifteen named accounts a week at three touches is roughly nine emails a day from one mailbox, comfortably inside a warmed sending domain. This is a named account program and not a volume program, so if the math ever pushes past twenty a day the problem is targeting rather than capacity.
4.Write ten versions of every step and test them.
Versions A through J, not A and B. Rotate subject lines and bodies. You learn which angle is actually working instead of guessing, and there is a second reason that matters more: identical bodies going out over and over is one of the patterns postmaster tools flag. Variation is a deliverability tool as much as a testing one.
Subject line seeds for this niche, each of which should become several variants: "the permit modification", "best month versus average month", "your historian and the new req". Lower case, no punctuation tricks, and nothing that would look odd in a reply from a colleague.
5.Stop at three.
Most replies arrive on the first and second email. The third is already thin. Every touch past that raises the odds the whole thread gets classified as spam, and that classification follows the mailbox to the next person you write to. The long cadence is over. Three touches, each with something new in it, then leave them alone for ninety days.
6.Know what good looks like.
A one percent reply rate with a quarter of those replies positive is a healthy trigger based program. Anyone quoting you double digit reply rates is counting out of office messages or selling a course.
7.LinkedIn Sales Navigator is not optional.
Every other data source tells you who held a title at some point. Sales Navigator tells you who holds it today, because the person maintains it themselves. That is the difference between a three percent bounce rate and a fifteen percent one, and bounces are scored against the mailbox the same way spam complaints are. Verify the name there before anything goes out.
It is also the cheapest trigger detector you will own. The job change filter surfaces people who arrived in a role in the last ninety days, which is the moment they have budget and no incumbent. The posted recently filter surfaces companies talking about the exact problem you solve. Account lists with headcount growth alerts tell you who is scaling before the press release does. For this niche the saved search is titles Plant Manager, Operations Manager, Process Engineer, Technology Manager, Advanced Process Control and Operational Excellence, at an account list you build from permit dockets rather than from an industry filter. Switch on job change alerts above everything else, because a new plant manager inside ninety days is the highest value trigger available here. Navigator confirms who holds the role today at sites where titles rarely appear anywhere else.
Use it for the research and the verification, not for the message. InMail reply rates are a fraction of email, and the person who replies to a thoughtful email is the same person who ignores a connection request with a pitch attached. Pull the work email from a data provider once Navigator has confirmed the person is real and current.
None of this is specific to your niche. All of it is specific to whether anyone ever reads the angle you spent an hour getting right.
If you would rather not run it yourself
That is what we do. ExpertLayer runs this exact loop for firms with proprietary technology: the permit and turnaround pull, the account list built site by site, the job posts that reveal the stack, the angle written per account in the language of that unit, the sending across warmed mailboxes, and the reply reading. You take the technical conversations, which is the part nobody can do for you.
The first step is free and it is the same research described above. Send us your website and we will come back with 10 companies that hit these triggers right now, with the filing, the post or the schedule, the contact, and the opening line for each.
Questions from people running this
Do we sell to the plant or to corporate?+
Corporate takes the meeting faster and funds the pilot, which is exactly why so many pilots die. The site never owned the number, so when the pilot ends nobody at the plant is worse off if it stops. Sell the plant manager the number, let the process engineer verify it, and bring corporate in to fund the rollout once one site is defending the result. The order matters more than the entry point.
Every prospect asks for a free pilot. What do we do?+
Give away the data readiness review, never the pilot. The review costs you two weeks, tells you whether the account is real, and produces a document the process engineer can circulate. A free pilot costs you a quarter of engineering time and teaches the buyer that the technology has no price. Firms that give away pilots in this niche spend the following year trying to charge for the same work.
Our best results are under NDA. How do we prove anything?+
By being specific about everything except the name. Same unit type, same feedstock variability, the number that moved, the months it took, and what happened at the next turnaround. A process engineer will accept an unnamed reference described in real units and reject a named logo described in adjectives. Ask for written permission at pilot close, when the result is fresh and the plant manager is pleased, not eighteen months later.
The controls engineer treats this as a safety question. Is that unreasonable?+
No, it is the correct reflex, and the firms that argue with it lose. Anything that writes to the control system sits inside a safety case, and the person who owns that case is right to be slow. Lead with advisory mode, keep the operator holding the wheel, and describe the path to closed loop as something the plant earns rather than something you switch on. That answer converts the controls engineer from a blocker into the person who explains you to the rest of the site.