You rank #2 for the right term. Almost nobody searches it.

Green Vault Systems builds automated weighing, batching and packaging equipment for licensed cannabis operators, from Mukilteo, Washington. The Precision Batcher and AirKush are the products the market knows you for.

This report measures one thing: what happens when someone who needs that machine goes looking for it. It’s built from your own site, your own search footprint and the live results pages for every term in your category. I parsed 59 pages of your website, pulled your 165 ranked keywords, and checked all 27 terms that make up the commercial category.

The short version: the machine is not the problem, and the marketing execution is not really the problem either. You rank #2 for “cannabis packaging machines”, which is exactly the term a buyer would type. It’s searched about 70 times a month. That single pair of facts is what this whole report is about.

Everything here was verified on 6 August 2026. Where something couldn’t be measured, it says so rather than guessing.

22 URLs under /products/ return “not found”, and Google still has 72 of your 165 ranked keywords pointing at them.

These were the spare-parts listings from the Shopify store on shop.greenvaultsystems.com — conveyor belts, motor drivers, nylon washers, thermocouples, hose clamps. Every one of them now returns HTTP 404 with no redirect behind it. Someone searches for a replacement part, finds you, clicks through, and gets an error page.

Fix them anyway. The case for the redirect map is not traffic recovery — on these numbers it is worth roughly 61 visits a month, and that is not a project worth selling you. It is that 22 live doors on your domain open onto an error, every external link ever pointed at those parts is now dead weight, and it costs half a day. Do it because it is cheap and correct, not because it is large.

The 22 URLs

• /products/1-pex-hose-clamps
• /products/100g-test-weight
• /products/18-conveyor-belt-replacement
• /products/24-volt-10-amp-power-supply
• /products/3-4-id-clear-reinforced-tubing-feet
• /products/6-bank-solenoid-rebuild-kit
• /products/ball-joint-ring
• /products/brushless-dc-motor-driver
• /products/conveyor-90-degree-radius
• /products/conveyor-belt
• /products/conveyor-clamp
• /products/conveyor-transfer-plate
• /products/denester-assist
• /products/homing-sensor
• /products/infeed-belt
• /products/multifunctional-timer-relay
• /products/nylon-plastic-washer
• /products/pack-funnel
• /products/photoelectric-sensor-reflector
• /products/pneumatic-thruster-slide
• /products/thermocouple
• /products/thermocoupler

What I can and cannot tell you about why

The product section is gone from the site’s content management system entirely. There is no product sitemap, no /shop/, and the sitemap index lists only posts, pages, categories and authors. No redirects were left behind for any of it.

Either way, the recommendation is the same: point each of these 22 URLs at the closest page that still exists. That’s a few hours of work — and it’s the cheapest item on this list, not the biggest one.

Your site isn’t thin and it isn’t neglected. It is 59 pages with a median of 2,115 words, published steadily, with schema markup in place. The problems are mechanical, and mechanical problems are the cheap kind.

Two test pages are live and indexed: /test-2/, titled “test popup 2”, and /test-popup/. So is an internal /category/for-review/ archive. Five assets 404 on their own pages, including two customer logos, Kings-Garden.png and Phatpanda-copy.png. Phat Panda is the subject of one of your own case studies.

Google holds no Chrome User Experience Report (CrUX) field data for this domain. There isn’t enough real traffic on the origin to populate one. Read every traffic estimate in this report against that.

Your About page works perfectly for people and does not exist for Google. Both halves of that sentence are true at the same time, which is why nobody has noticed.

Verified live on 6 August 2026: greenvaultsystems.com/about/ returns HTTP 500 while serving 76,351 bytes of perfectly good HTML. WordPress streams the page, hits a fatal PHP error near the end of the document, and sets the error status on shutdown, after the content has already gone out. Browsers render what arrived, so every human visitor sees a normal page. Crawlers read the status code, treat the URL as a server error, and drop it.

The page runs five Elementor add-on plugins on one page builder, the signature of an update breaking a single widget. That is a diagnosis to start from, not a certainty.

The visible consequence: your CTO and your Production Manager appear zero times across 59 pages of your own website, because the page that introduces them is the one page search engines have thrown away.

You rank for 165 keywords, worth an estimated 389 organic visits a month. 11 of those keywords sit in the top ten.

What you actually holdRanked by position, not by volume — these are the terms you own

Read the top two rows again. You are #2 and #5 for “cannabis packaging machine(s)”, the exact phrase a buyer would type. There’s nothing wrong with how that page is optimised. The term is simply worth 70 searches a month.

What you do not holdThe high-volume terms in your footprint, and where you actually sit on them

These are home-grower and hobbyist terms. They carry real volume, and you’re on page five to page ten for all of them. They aren’t your buyers, and chasing them wouldn’t help. I list them because they are the reason a traffic-shaped report would tell you to write more blog posts. That would be the wrong advice.

You run no paid search at all: 0 paid keywords. That’s a decision worth revisiting, and sheet 06 explains why.

I built a thousand-term commercial slate from your own market and scored every term. 27 of them are genuinely cannabis-equipment terms. Together they are worth 3,790 searches a month.

This is the finding the rest of the report hangs on. The whole cannabis packaging equipment category is 3,790 searches a month nationally. That is every term, not a sample. The generic packaging-automation category next door is 136,970. You’re competing in a room roughly 36 times smaller than the one your machine already qualifies for.

Your categoryThe biggest terms in it, and what advertisers pay for a click

The category next doorSame machine, different vocabulary

The strategic read: you sell a precision weighing and batching machine. Cannabis is one industry that needs it. Your equipment already qualifies for the far larger food, nutraceutical and general packaging-automation market. Your own site says you have decades of experience in food and agriculture. The category, not the ranking, is the constraint.

I bought the full ranked-keyword footprint of every equipment manufacturer you named. Then I checked who holds page one for all 27 terms in your category.

Who actually holds page oneAcross all 27 category terms — the census, not a sample

Canapa by PaxiomYour closest true competitor, measured properly

Paxiom’s cannabis brand trades as Canapa at canapasolutions.com. It ranks for 120 keywords worth 21,020 monthly searches, and it holds page one for the terms that describe the machine itself:

Your competitors are not invisible, so the honest comparison is this: Canapa owns page one for “pre-roll machine” and you rank for “clear trichomes”. They’re answering the buyer’s question. Your best-ranked content answers a grower’s.

Pre-roll is the largest live sub-cluster in your category: 8 of the 27 terms. My crawl of your site found no pre-roll product line, so the gap is in what you sell as much as in what you rank for. Worth a conversation before it becomes an SEO task.

Sheet 06 delivered the bad news: 3,790 searches a month across your entire category. This sheet is the answer to it.

You can’t wait for 3,790 searches to find you. You don’t have to. Every licensed cannabis producer and processor in the United States is a matter of public record, filed with each state’s regulator, updated continuously, and free to obtain.

As a demonstration, I pulled Washington from the Liquor and Cannabis Board’s own renewal file: 365 businesses holding a Producer or Processor privilege, each with trade name, licence number, street address, city, county and phone number. Not a purchased list, not scraped, not inferred. It’s the state’s published record.

Then I found the file behind the fileSame agency. Not linked from the open-data portal. Three times the size, and it carries email.

Then I kept going, and the renewal file turned out to be a slice — it only holds licences sitting inside a renewal window, which is why it returns 365 rows. The same agency runs its own mapping service, and the layer behind the public Cannabis Retailer Map is an open, read-only API that returns every active licensee in the state in a single request. No key, no login, no records request, no scraping: one URL, 1,684 licensees, 1,673 of them with a phone number and 1,683 with an email address, across 231 cities.

The list itself isn’t in this report, and that’s deliberate. Those are real businesses’ direct contact details, and pasting an extract of them into a sales document would be the wrong way to demonstrate that I can find them. The endpoint is named above. It answers to anyone, including you, and it will answer the same way tomorrow. That’s the point — you don’t need me in order to have this.

Then a third source, and it is the one that qualifies a buyerThe mapping layer says who is licensed. It does not say who weighs.

The Washington Department of Revenue publishes an endorsement register for every licensed location in the state. I read all 2,017 of them — one detail page at a time, 1,683 distinct businesses, nothing sampled and nothing estimated. What the register adds is the column neither of the other two sources has: which endorsements each location actually holds, and whether they are active.

Three sources, three different numbers, and I have not merged them. The renewal file gives 365. The WSLCB mapping layer gives 1,024 businesses. The DOR register gives 982 with an active producer or processor endorsement. They disagree because they record different things — a renewal window, a licence footprint, and an endorsement status — and averaging them into one confident figure would be the dishonest move. Each is named with its source above. If you want one number to work from, it is 821: active, and already weighing.

This part doesn’t depend on search at all. It’s why a small category is survivable: in a market this size, you go and get the list.

You have two Google Business Profiles. One is unclaimed. Neither shows the Mukilteo address you actually trade from. All four of your reviews sit on the one with no address at all.

The four reviews are dated 24 December 2017, one in 2019, one in 2021, and the most recent on 16 January 2022. Every one is five stars. That’s a good record, and it’s four and a half years cold.

Fixing this is free and takes an afternoon: claim listing A, merge or remove the duplicate, set the real address, set a category, set hours, and ask the last dozen customers for a review.

Your link profile is healthy and genuinely earned. This is the sheet with good news on it.

242 referring domains and zero broken backlinks is a better profile than most manufacturers of your size carry. Nothing here needs fixing.

I checked 41 directories: cannabis trade, industrial marketplaces and general business listings. You are present on 6.

Present

Cannabis & Tech Today, Crunchbase, Google Business Profile, LinkedIn, MapQuest, ZoomInfo

MissingIncluding every cannabis trade directory except one, and every industrial marketplace

BBB, Facebook, YouTube, Manta, Bing Places, Apple Maps, Chamber of Commerce, Dun & Bradstreet, Bloomberg, Yelp, Yellow Pages, Superpages, Foursquare, Citysearch, Hotfrog, Cylex, MerchantCircle, Local.com, Brownbook, BizProfile, OpenCorporates, USA Cannabis Directory, MJBiz Directory, MJBizDaily, mg Magazine, Ganjapreneur, Cannabis Business Times, Marijuana Venture, Thomasnet, GlobalSpec, Machinio, Exapro, Packaging Digest, PMMI, Explorium

One axis of this could not be measured. Checking directories by phone number returns a persistent server error from the data provider; I made four attempts across two number formats. That axis is not measured, which is not the same as clean, and I am not reporting it as a result.

Increasingly, the buyer never reaches a results page. They ask an assistant and take the answer. So I asked the assistants directly.

I put 10 real buying questions to four answer engines: Google’s AI Mode, ChatGPT, Gemini and Perplexity. I phrased each one the way a licensed operator actually would: “Who are the best manufacturers of … for a licensed cannabis operator?”. That is 40 answers. You are named in 3 of them.

Perplexity names you twice and Gemini once. Google’s AI Mode and ChatGPT do not name you at all across ten questions each. When they do name you, they cite your domain. So the citation works. The problem is that you are not in the answer to begin with.

Everything on this sheet is my own measurement, made on 6 August 2026. There is no industry average anywhere on this page. I asked the questions, I paid for the answers, and the counts above are what came back.

The questions I asked: pre-rolls packaging; weed machine; equipment for growing weed; automated pre-roll machine; equipment for weed; pre-roll machines; hemp cultivation equipment; hemp farming equipment; pre-roll joint machine; custom pre-roll packaging.

On 29 July 2026 you posted a Marketing & Business Development Manager role — Mukilteo, full-time, $65,000–$80,000. It lists 12 responsibilities. 7 of them describe something measured independently in this report, before I had read the posting.

Your own HR department published a list of what the company is missing. This sheet sets it beside what the crawl found. Two sources that never spoke to each other, one conclusion — which is the only reason it is in the report rather than in a sales letter.

Nothing on this sheet is an opinion about the role. It is your list and my measurements, printed next to each other.

Four moves, in the order I would do them. The first two are maintenance. The third and fourth are the actual strategy.

Move 1 — redirect the 22 dead product URLsA few hours. Recovers the link equity draining with them, and about 61 visits a month.

Point each dead /products/ URL at the closest page that still exists. Nothing else on this list pays back this fast. Before you start, settle internally whether that catalogue was meant to go. The answer changes what each URL should point at, though not whether to do it.

Move 2 — fix the About page and claim the second Google listingAn afternoon each. Both are trust, not traffic.

The 500 error is almost certainly one Elementor add-on. Deactivate them one at a time and watch the status code. Then claim listing A, set the Mukilteo address on whichever listing survives, add a category and hours, and ask recent customers for reviews. The newest one you have is from January 2022.

Move 3 — go and get the buyersThe category is 3,790 searches a month. Do not wait for them.

Sheet 07b ends at 821 licensed Washington locations that hold an active producer or processor endorsement and a scale registered with the state — 739 businesses that have already told the government, in writing, that weighing accuracy is a compliance cost worth paying. Addresses and phone numbers included, pulled free from the state’s own record. The same records exist for every state you sell into. In a category this size, outbound to a qualified list beats waiting on search, and it is the one channel whose size you control.

Move 4 — decide whether you are a cannabis companyThis is the one that is worth more than the rest combined.

Your entire category is 3,790 searches a month at roughly zero cost per click, because nobody is bidding. Directly adjacent, general packaging automation runs 136,970 searches a month at $12–$91 a click. That is roughly 36 times the demand, from buyers with procurement budgets. Your machine already does that job, and your own site claims decades in food and agriculture.

If you do only one thing from this report, do Move 1. It takes a few hours. If you do only one thing that matters, have the Move 4 conversation.

This is a sequence, not a price list. Each stage is only justified by the measurement the previous one produced — if a stage cannot point at a number the stage before it generated, it does not get bought.

Stage 0 — you, freeThe three fixes on sheet 12. No invoice, no conversation required.

Redirect the 22 dead product URLs, deactivate the Elementor add-ons one at a time until /about/ stops returning 500, and claim the Google listing. Your own team can do all three this week. I have written them out in enough detail to execute without me, and they are worth doing whether or not we ever speak.

Stage 1 — measure before optimisingGA4 and Search Console in your Google Cloud project; server-side tagging into your BigQuery.

Sheet 13 says Google holds no real-world speed data for your origin, which means every traffic number in this report is modelled. That is honest, and it is also not good enough to optimise against. Stage 1 replaces modelled numbers with your own, and baselines the zero-click surface so the 3-of-40 answer-engine count becomes a trend instead of a snapshot. Nothing gets optimised until it is measured.

Stage 2 — move into the adjacent categoryThe pages that do not exist yet, structured data, entity resolution.

This is where the 36× lives. Your category is 3,790 searches a month at about $0.00 a click; general packaging automation next door is 136,970 at $12–$91. The machine already qualifies for it. What does not yet exist is a single page on your site that says so in the language that market searches in.

Stage 3 — answer engines, ongoingThe one number on this page that is a scoreboard.

You are named in 3 of 40 answer-engine responses today. That number is measured monthly and reported whether it moves or not. It is the only metric here that cannot be flattered by a chart, which is exactly why it is the one I would be judged on.

Stage 0 is the only one I would push you to do this week. Everything after it waits on a measurement that does not exist yet.

Douglas Mallett. Self-employed since 17; twenty-five years across industrial automation, data engineering, design, and now AI. Kirkland, Washington.

Industrial automation is where I learned verification — a machine that misreports its own state causes real damage. That turned out to be exactly what AI needs, because it fails the same way: confidently, and silently. You build machines that weigh and batch to a tolerance, so this is a standard you already hold your own equipment to. It is the standard this report was built to: every number in it traces to a file, and the ones I could not source are named on sheet 13 instead of estimated.

Today I run my own platform — thousands of domains scanned every 72 hours into a warehouse I built and operate alone. The audit you are reading came out of that platform, not out of a template. This is not a service I read about.

The toolsMarks shown are what the work is actually done in, grouped the way the career ran.

Photography is on this list on purpose: DSLR and lighting, which is a real skill and not a rounding error — your own posting asks for it twice. Drone videography is not something I do.

I am not applying for the role on sheet 11b. What I do sits alongside a hire rather than instead of one: the measurement, data and search infrastructure as a system, so that whoever you hire spends their time on the company instead of on the plumbing.

How every number in this report was produced, and what I could not measure.

What I could not measure

Directory presence by phone number. The data provider returns a persistent server error on that query. Four attempts, two number formats, same error. Reported as not measured, not as clean.

Whether the product catalogue was retired deliberately. There is no archived copy of those pages and no redirect was configured, so intent cannot be established from outside your business. Sheet 02 asks rather than assumes.

Real-world speed data (CrUX). Google’s Chrome User Experience Report holds nothing for your domain. There is not enough real traffic on the origin to populate it. Every traffic estimate here is modelled, and should be read that way.

Two dates on your own site. Your pages attach “75 years” of experience to two different fields, and third-party profiles say 65. I have not resolved which is right, so no figure for it appears in this report.

Prepared for Green Vault Systems, 6 August 2026.