Custom security guard software in Baltimore: a post is 168 hours, a guard is 40

Thirty-two companies, and the line that moved under all of them

There is a version of this article that opens with the global physical security services market and a compound annual growth rate, and it would tell a Baltimore guard company nothing whatsoever. So here is the local picture instead, counted rather than estimated.

The Census Bureau’s County Business Patterns file for 2023 — the most recent county‑level release — records 274 security guard and patrol establishments with paid employees in Maryland under NAICS 561612. Between them they employ 21,747 people and carry an annual payroll of $701.7 million. The caveat about paid employees matters here in a way it does not in most trades: a retired officer working weekend doors as a sole proprietor does not appear in this file, and neither does the one‑person patrol company that subcontracts every hour it sells. Everything below counts businesses that have made the jump to being an employer, and the real population of people who stand posts for money in this region is larger than the file admits.

Baltimore City has thirty‑two of them. They employ 3,558 people and pay out $96.8 million a year, which works out to 111 employees per establishment — by a wide margin the most labor‑dense figure we have found in four months of writing about Baltimore trades. The previous record holder, waste hauling, was 38. Montgomery County is denser still at 44 establishments and 7,489 employees, or 170 apiece; Prince George’s has 58 and 4,072; Baltimore County has 44 and 3,582.

Those averages hide the actual shape of the market, and the size bands are where the story is. Of the thirty‑two Baltimore City establishments, eleven — 34.4% — have fewer than five employees. Three more have between five and nine, three have between ten and nineteen, seven have between twenty and forty‑nine, and three have between a hundred and two hundred and forty‑nine. So a third of the city’s guard companies employ, between them, at most forty‑four of the 3,558 people in the trade. One in three companies employs about one in eighty guards. That is a barbell, and it is the single most useful fact about who these products are being sold to.

Look at the Baltimore City payroll‑per‑employee figure for a moment, because it is doing more work than it appears to. $27,216 a year divided by Maryland’s $15.00 minimum wage is 1,814 hours, which is 34.9 hours a week across a full year, or about forty‑five weeks of full‑time work. Nobody in this trade is working a tidy 2,080‑hour year. The average Baltimore guard’s payroll record is a part‑year, part‑week record, and it is that fact — not any vendor’s feature list — that determines what the software has to be good at.

For contrast, the eleven Baltimore City establishments in NAICS 561621, security systems services, employ 124 people on a payroll of $5.58 million: $45,000 per employee, or 1.65 times the guard rate. The people who install the cameras are paid two‑thirds more per head than the people who watch them, and they work in companies a tenth the size.

A post is 168 hours. A guard is 40.

Every argument in this article comes out of one piece of arithmetic that almost nobody writes down, so let us write it down.

A twenty‑four‑hour post is a place, and it needs covering every hour of every day. That is 168 hours a week and 8,760 hours a year. A guard is a person, and under Md. Lab. & Empl. §3‑415 read with §3‑420 every hour over forty in a workweek is paid at one and a half times the usual rate. Divide one by the other and a single continuous post requires 4.2 people. You cannot hire two‑tenths of a person. Everything after that is a choice between two imperfect roster shapes, and both of them cost something the scheduling software does not show you.

Cover the post with four officers and you get 160 straight‑time hours and 8 hours of overtime every single week, permanently. Nobody is sick, nobody is on vacation, nothing has gone wrong; the overtime is structural. At a $17.00 wage — roughly the Baltimore market rate for unarmed work in 2026 — that costs $2,924.00 a week, $152,048.00 a year, and the effective cost of an hour of coverage is $17.40 rather than $17.00.

Cover the same post with five and every hour is straight time. Each officer works 33.6 hours. The weekly wage bill falls to $2,856.00 and the annual to $148,512.00. The five‑guard roster is $68.00 a week — $3,536.00 a year — cheaper than the four‑guard roster for exactly the same coverage.

Now the part that makes this a Maryland article rather than an arithmetic exercise. Md. Business Occupations and Professions §19‑504, in force since 1 January 2025, requires a security guard agency or security guard employer that employs five or more individuals as security guards to carry commercial general liability insurance including errors and omissions and completed operations with a $1,000,000 total aggregate minimum, and one that employs fewer than five to carry the same cover with a $500,000 total aggregate minimum. Proof goes to the Secretary of State Police, and if the cover lapses the agency license is suspended automatically — no hearing, no notice period, no discretion — and stays suspended until proof is filed.

So the fifth officer saves $3,536 a year in wages and moves the company onto a different insurance requirement. We are not going to tell you which is cheaper, because we genuinely do not know: the premium difference between a $500,000 and a $1,000,000 aggregate depends on your carrier, your loss history, your client contracts and what your building owners demand anyway. Plenty of Baltimore property managers require a million‑dollar aggregate in the contract regardless of what the statute says, which makes the whole question moot for the companies that serve them.

The point is narrower and, we think, sharper. That decision is currently being made by a scheduler. Somebody adds a weekend relief officer to fill a gap, the roster balances, the overtime line goes green, and a statutory threshold moves without anybody in the building noticing that a threshold exists. The number that matters — how many distinct individuals are on the roster as security guards today — is not a number any guard management product puts in front of you, because none of them is counting for that reason.

And §19‑504 counts individuals. Not full‑time equivalents, not hours, not payroll dollars. A retiree who works one eight‑hour Saturday a month is one individual. The officer covering a 42‑hour week is one individual. On the axis the statute uses, they are the same size.

The vendors, and the five different things they count

We checked sixteen guard management products on 26 August 2026, reading each company’s own published pages rather than an aggregator, and recording what the server actually returned. It was an unusually revealing sweep, and the first finding is not about price at all.

Four of the best‑known brand names in this category are now one website. Request tracktik.com, trackforce.com, trackforcevaliant.com or silvertracsoftware.com and every one of them lands on www.trackforce.com. Ask any of those four domains for /pricing and you arrive at the same page: trackforce.com/products/tracktik/pricing/. That matters because Silvertrac in particular is still quoted at a specific flat monthly figure across a dozen comparison articles, and the brand no longer has a site of its own to check it against. A second consolidation runs alongside it: lighthouse.io now redirects to teamsoftware.com/lighthouse, putting Lighthouse and WinTeam under one roof.

The second finding is that the page you land on is titled “Pricing” and contains no price. What it does contain is the clearest statement of a meter we have read in any category: “Priced by sites, not seats”, followed by “Your quote scales with the sites you protect – grow your guard team without growing your bill.” Modules are the second axis: tours and reporting first, then billing, payroll, business intelligence and visitor management as you add them. It is an honest description of a pricing philosophy and it is also, precisely, the opposite of what the products that do publish numbers are doing.

Count the nouns in the second column. User accounts. Active users. Seats. Employees. Sites. Five different populations, sold into one trade, and no two of them contain the same set of people. A dispatcher is a user account and a seat but not a guard. A client’s facilities manager with portal access is a user account but not an employee. A subcontracted officer is a person standing on your site who may be nobody’s employee at all. And a site is not a person in any sense — it is the only meter in the list that measures the thing your customer actually buys.

We want to be fair about this, because the category is not being cynical. Every one of those meters is defensible from where the vendor sits. Pricing by site, as TrackTik does, is genuinely kinder to a company with high turnover, and TrackTik says so plainly. Pricing by user account, as THERMS does, is genuinely kinder to a company running many sites out of a small office. Both are honest. It is only when you put them side by side that you notice the trade has no agreed unit of account at all — which is a strange thing to be true of an industry that sells one product, an hour of presence in a fixed place.

One product runs directly against the grain of all of that. Patrol Points publishes two plans, Core and Premium, with no figure attached to either — but its own feature grid states that “unlimited guard/user accounts” are included in the premium package, with NFC checkpoints carried as an additional fee and setup, hardware and first subscription due thirty days after the start date. It is the only product in the sweep that says out loud that the number of people is not the meter, and it is also one of the ones that will not tell you what the meter is.

Two vendors deserve a note rather than a row. Guardso did not respond on either of its domains when we checked, which is a connection failure and not a hidden price; we mention it so the table is not read as a judgment. And Belfry’s /plans page, which carries no rate, does publish a fact worth knowing: the company announced a $12 million raise for physical security billing and payroll. In a category this fragmented, that is a signal about where the money thinks the problem is, and the answer is billing.

The meter is the user account, and it is billed at high water

THERMS is the most transparent pricing page in the category and it deserves to be read closely, partly because it is admirable and partly because the mechanic underneath it is one very few buyers will have thought about.

The ladder has ten bands. One to five user accounts is $45 a month; six to fifteen is $105; sixteen to thirty is $155; and so on up to three hundred and one to four hundred at $1,120. Above four hundred you write to them. It is flat‑rate within each band, which means the price per user account is not a price at all — it is a sawtooth.

Two things fall out of that table. The first is that the largest proportional step on the whole ladder is the very first one: going from five user accounts to six adds $60 a month for one person, a 133% increase. The second is that the per‑account cost is not monotonic. Read the cost of the first account in each band as you climb: $45.00, $17.50, $9.69, $6.94, $5.98, $5.13 — a steady fall — and then $5.15 at a hundred and one. Crossing into that one band leaves each account fractionally dearer than it was in the band below, which happens nowhere else on the ladder. It is also a long way above the $3.90 an account you were paying at a hundred, one person earlier.

Then there is the billing rule, which the vendor states plainly in its own FAQ: “Billing is calculated automatically based on the maximum amount of users your organization had during the billing period.” Not the average. Not the count on the invoice date. The high‑water mark.

That is a perfectly reasonable rule and it is also a rule with a shape. A special event that puts eight extra officers on for one Saturday prices the whole thirty days. A terminated officer whose account nobody deactivated until the following Friday prices the whole thirty days. A client portal invitation sent to four facilities managers during a bid prices the whole thirty days. In a trade whose defining characteristic is a roster that moves constantly, a high‑water meter charges you for your worst week and calls it your month.

Md. §19‑504 changes what you must insure between your fourth guard and your fifth. THERMS changes what you pay between your fifth account and your sixth. Two systems that have never heard of each other put their steepest step within one person of each other, in exactly the size range where a third of Baltimore’s guard companies live.

And the two steps count different populations, which is what makes the coincidence more than a curiosity. The insurance line counts individuals employed as security guards: your dispatcher does not move it, your office manager does not move it, your client’s portal login certainly does not. The software line counts active user accounts: every one of those people moves it. So a company at four guards and one dispatcher is at five user accounts and four individuals. Hire one more officer and you have crossed both lines at once. Hire an office administrator instead and you have crossed only the software one. Nothing in either system tells you which line you just moved.

The part no national platform models

This is the section we invest in every month, and it is the section that decides whether a build is worth doing. What follows is not a list of missing features. It is a list of facts about Maryland that a general‑purpose guard product cannot know, and that a company in Baltimore has to be right about anyway.

Three thresholds on one roster, and three ways of counting

A Baltimore guard company of a dozen people is sitting on three separate headcount thresholds. Each of them changes a legal or financial obligation. Each of them counts a different set of people. And each of them is measured on a different clock.

The fifteen‑employee test is the one that catches people, because it is retrospective. §3‑1304(a)(2)(i) says the number “shall be determined by calculating the average monthly number of employees employed by the employer during the immediately preceding year”, and (ii) insists that every employee counts “without regard to” status or eligibility. So a company that ran fourteen people for ten months of 2026 and twenty‑one for the two months of a summer stadium contract has an average of 15.17, and every hour of earned sick and safe leave taken in 2027 is paid leave. The decision was made in July. Nobody found out until January.

The relief guard who accrues nothing and moves two lines

Take one person: a retired postal worker who covers one eight‑hour Saturday every other week at a Fells Point office building. Two shifts a month. Sixteen hours. In the language the leave statute uses, eight hours in a two‑week pay period.

Under §3‑1304(c)(5)(i) that person accrues no earned sick and safe leave at all, because the statute excuses accrual in “a 2‑week pay period in which the employee worked fewer than 24 hours total”. Under §3‑1304(a)(2)(ii) that same person counts as one whole employee in the average that decides whether everybody else’s leave is paid. Under §19‑504 that same person is one of the five individuals that doubles the company’s required liability aggregate. And in THERMS they are one user account, which at the wrong moment in the ladder adds $60 a month — more than the first five accounts cost put together.

One person, working sixteen hours a month, accrues nothing under the leave statute, counts fully under the leave statute’s own threshold, moves the insurance requirement for the entire company, and can add more to the software bill than the four full‑timers beside them cost put together. There is no field in any scheduling product that says any of this.

Meanwhile the full‑timer on the four‑guard roster has a different quiet fact attached to them. Accrual under §3‑1304(b) runs at one hour for every thirty worked, and §3‑1304(c)(1) caps earning at forty hours a year. Forty times thirty is 1,200 hours, so an officer on a 42‑hour week stops earning leave in week 29 and works the remaining twenty‑three weeks of the year accruing nothing. An officer on the five‑guard roster at 33.6 hours reaches the same cap in week 36. The same policy, applied to the same post, produces two different accrual curves depending purely on how you sliced the roster — and the balance an employee sees in an app in October is a function of a scheduling decision made in January.

Two duties that turn on what the company knows

Maryland has written two obligations into this trade that are not about facts. They are about knowledge, which is a different and much harder thing for software to hold.

§19‑408(c)(1): a licensed security guard agency or security guard employer “shall notify the Secretary within 48 hours if it knows that an individual who provides security guard services on its behalf or for it has been arrested for a crime of violence … or a disqualifying crime”. The clock does not start at the arrest. It starts when the company knows. A Friday night arrest that surfaces in a Monday morning phone call started its clock on Monday, and the company has to be able to say when Monday was.

§19‑404.1(g): at least ninety days before a guard’s certification expires, the agency or employer must submit to the Secretary “a declaration … under the penalties of perjury stating whether it knows that a certified security guard … no longer satisfies all of the criteria for certification”. A perjury‑backed statement about the company’s own state of knowledge, per person, on a rolling three‑year cycle, due ninety days before a date that was set by whoever employed that guard three years ago — which may not have been you.

Notice what that means for a credential that is genuinely portable. The certification card belongs to the individual and is good for three years; the application that produced it, under COMAR 29.04.01.02(D), was prepared by “the hiring agency”, and under §19‑402(a)(2) the applicant had to be an employee or an applicant for employment at the time. So the card is the person’s, the paperwork was somebody else’s, and the perjury declaration is whoever employs them ninety days before the expiry date. You sign, under oath, about a credential you did not obtain, on a clock you did not start.

There is a third one, and it is the most operationally awkward of all. §19‑401(b) allows a licensed agency to put an uncertified person to work if the application has been submitted and “the Secretary has not disapproved the application”. That is a lawful employment status defined entirely by the absence of an event. You cannot query it. There is no record that says yes. There is only a submitted application and a silence, and the silence is what makes the shift lawful — until a letter arrives, at which point every hour after the letter is a different thing from every hour before it.

Fifteen clocks, and not one of them is a shift

We counted the deadlines. Not the good practice, not the client‑contract obligations — only the ones written into Maryland statute and regulation for a company that employs certified guards and holds a special police commission or two.

Fifteen deadlines, with durations of immediately, 48 hours, five days, ten days, fifteen days, thirty days, ninety days and a hundred and eighty. Not one of them is a shift, a pay period, a billing cycle or a client contract term — the four rhythms that every product in this category is actually built around.

Look at the two use‑of‑force rows together, because they are the most interesting pair. The guard has 48 hours to file with the company; the company has 48 hours from receipt to file with the State. The outer limit is therefore 96 hours, but the company’s own deadline is set by a decision the guard makes. File at hour one and the company has until hour 49. File at hour 47 and the company has until hour 95. You cannot know your own deadline unless you timestamp the moment the paper arrived, which means the receipt event has to be a first‑class record and not an email in somebody’s inbox.

Then read the definition. §19‑414(a) makes “use of force” include any physical striking, any significant physical contact that restricts movement “including control techniques”, and detainment without consent — and excludes “mere presence, verbal commands, or escorting an individual with minimal resistance”. Minimal is not defined. So whether a 48‑hour statutory clock started at all turns on a judgment about the degree of resistance during an escort, made in a corridor at two in the morning by the person with the most incentive to file nothing. The clock may already be running and nobody has decided whether it is.

And the report the company eventually files has to carry, under §19‑414(c)(2), the type of encounter, the type of force, the location, whether the person was arrested and on what charges, whether either party requested or required medical care, and demographic information including race, ethnicity, gender and age for every person involved. Those last fields cannot be reconstructed a week later from a narrative incident report, which is what most guard products collect. They have to be captured as structured fields at the time, by somebody who has just been in an altercation, on a phone.

The authority stops at the property line. The liability does not.

Special police officers are the other half of this trade and they are governed by a completely separate scheme. A commission is issued by the Governor under Md. Public Safety §3‑302, and under §3‑306(b) the commission itself must state its term and “the property that the commission is intended to cover”. Under §3‑307 the officer may exercise the powers of a police officer only on that property.

So a special police officer’s authority is a fact about real estate. Move them from the Charles Street building to the Pratt Street building and the powers do not travel; a fresh application is required, and under §3‑304(a)(2) “a separate application is required for each individual applicant”. A guard company covering six commissioned sites with a pool of twelve officers is not managing twelve credentials. It is managing a matrix.

Now put §3‑308(b) beside that. The entity that requested the appointment is responsible for any wrongful action the officer commits in the course of their duties and for “any abuse of the powers granted by the commission, either on or off the premises”. The authority is bounded by a property line drawn in an application. The liability is expressly not. That asymmetry is the single most important sentence in the subtitle for anybody pricing this work, and it is why the property boundary in the application has to live in your system as a real object rather than as a line in a Word document in a folder.

A commission that survives the sale of the building

§3‑313(c) is the provision we did not expect. A commission does not terminate when the employer no longer needs the officer because it has transferred the property to somebody else, provided the new owner executes a form affirming that it will employ the officer to protect that property and will assume the original employer’s responsibilities. The commission follows the real estate. Everywhere else in employment law a credential attaches to a person or to an employer; here it attaches to a building, and it can be handed over at settlement like a boiler warranty.

The termination sequence is stranger still, and it is three clocks on one event. When you relieve a commissioned officer, §3‑310(c)(1) gives them 48 hours to surrender the badge and identification to you; §3‑310(c)(2) gives you 10 days to return the commission card to the Secretary; and §3‑313(a)(4) says the suspension or termination “may not take effect until 5 days after notice is sent” to both the officer and the employer. Read those together and there is a window — roughly the third, fourth and fifth days — in which a person has been relieved, has surrendered the badge, and still holds a live commission. Meanwhile §19‑507(b) gives that same person 10 days to return a badge bearing the agency’s name. Same locker, same Tuesday, two different deadlines depending on which piece of metal it is.

Eighty hours, and twelve

The two credentials require wildly different amounts of training, which is worth knowing before you promise a client an officer with arrest powers by the end of the month.

Twelve hours against eighty is a factor of 6.7, and the two people can end up standing at the same door on the same night. The twelve‑hour requirement is itself new: it arrived with Senate Bill 760 of 2024 and took effect on 1 January 2025, along with the “security guard employer” category and the §19‑504 insurance floors. Before that, the statute asked for a criminal history records check and no training hours at all. Twenty months is not long enough for the operational habits to have caught up, which is precisely why this is a good moment to build the record properly.

§19‑412(b) tells you what those hours contain, and it is not what the trade’s reputation would suggest: relevant criminal law, the appropriate use of force, the use of alternatives to force, de‑escalation techniques, and appropriate interactions with people under eighteen, people with behavioral health or other disabilities, and people in crisis. There is a real training record to keep here, with real content, and it renews on a per‑person three‑year cycle that is not aligned to anything else in the business.

One more cross‑statute detail worth a paragraph, because it is the kind of thing that only shows up when you read two books at once. Md. Lab. & Empl. §3‑413(d) lets an employer pay 85% of the State minimum wage — $12.75 against $15.00 — to an employee under the age of eighteen. §19‑402(a)(6) requires a certified security guard to be at least eighteen. So the youth wage exists in Maryland law and this trade can never lawfully reach it. That same statute, at §3‑413(a)(3), still defines a “small employer” as one with fourteen or fewer employees, a distinction that mattered when the two minimum wage tracks were different and has been a dead letter since both reached $15.00 on 1 January 2024.

The badge register that Maryland already designed for you

Our favorite find of the month is a regulation that is, without meaning to be, a database schema.

COMAR 29.04.01.08 governs badges. Written approval must be obtained from the State Police Licensing Division before any badge is issued, requested in writing on agency letterhead specifying the total number of badges required. The badges themselves “shall be numbered consecutively, within the agency, with no duplication, from number 1”. They may be gold finished for principals and corporate officers and shall be silver for everybody else. They must carry the full agency name and the word “Investigator” or “Security Guard”. A badge is issued only after the employee has received their certification card. The agency is responsible for the proper use of any badge issued. And loss, or a leaver’s failure to return one, is reported to the Licensing Division immediately.

A unique integer key starting at 1 with no gaps and no duplicates. A foreign key to the certification record that has to exist before the row can be created. An attribute whose value encodes the org chart, in metal. A column that changes for every row if the company is ever renamed. And a trigger with no stated latency. Maryland has specified the table. Almost nobody in the trade keeps it as one.

That last part is not a joke at anyone’s expense. Badge registers in small guard companies live in a spreadsheet, or in the locker, or in the head of whoever ordered the last batch. The regulation asks for consecutive numbering from 1 with no duplication for the life of the agency — which means the number is not reusable when somebody leaves, and a company on its third generation of owners is expected to know what badge 14 was. If you are building anything at all for a Maryland guard company, this is the cheapest table to get right and the one most likely to be wrong today.

Two provisions pointing at things that no longer exist

We look for these every month, because a rule that has outlived the thing it refers to is usually a rule nobody has re‑read, and the operational habits underneath it drift accordingly. This trade has two, and one of them is specific to Baltimore.

The first is the definition that decides whether you need a license at all. §19‑101(n) defines “security guard services” and then excludes three things: a marine guard or ship watchman; an unarmed employee of a bar, tavern or restaurant; and “a special police officer appointed and while performing under Title 3, Subtitle 3 of the Public Safety Article or §16‑16 of the Code of Public Local Laws of Baltimore City”. Baltimore City is the only jurisdiction in Maryland that gets its own carve‑out in that sentence. Go and look at §16‑16. The City Law Department’s own compilation of the Public Local Laws, dated 20 December 2024, prints the whole run as “§§ 16‑1. to 16‑16. {Repealed}”, annotated “Repealed by Ch. 453, Acts of 2014; Repealed by Ord. 24‑311” — the ordinance enacted on 24 January 2024 that cleared out nearly all of Subtitle 16, alongside the companion charter amendment Baltimore voters ratified as Question E on 5 November 2024. The live State definition of who needs a security guard license in Maryland exempts a category of Baltimore officer created by a section that has been repealed twice.

The second is smaller and blunter. COMAR 29.04.01.05, headed “Authorization for Employment Verification Required”, forbids submitting a certification application for a person who is not a born or naturalized citizen “unless there is attached to the application a valid work authorization from the Immigration and Naturalization Service”. The Immigration and Naturalization Service was abolished on 1 March 2003 and its functions were split across three agencies inside the Department of Homeland Security. The document a Maryland guard company is instructed to attach in 2026 has not been issued by that name in twenty‑three years.

Neither of these is a scandal and neither will get anybody fined. They are useful for a different reason: they tell you which parts of the scheme are being actively administered and which parts are on autopilot, and they are a good argument for keeping your own record of what you actually filed and when, because the paperwork of record here is older than a lot of the people carrying it.

The checkout Maryland wrote for you

Most of the trades we write about are not taxable services in Maryland, and the tax section ends up being about what you buy. This one is different, and it is the reason we think the storefront argument is stronger here than in almost any category we have looked at.

Six percent out, three percent in

Md. Tax‑General §11‑101(m) lists fifteen enumerated taxable services, and item (10) is “a security service, including: (i) a detective, guard, or armored car service; and (ii) a security systems service”. The Comptroller’s own published list of taxable property and services repeats it in the same words. There is no ambiguity to litigate and no dominant‑purpose test to argue about: a Maryland guard company charges 6% on what it invoices, full stop.

Items (14) and (15) of the same subsection are newer. Data and information technology services described under NAICS sectors 518, 519 and 5415, and system software and application software publishing under NAICS 5132, became taxable at 3% from 1 July 2025. That is the rate on the software you buy — and, we should say plainly, on the fixed‑price work we sell.

So this trade sits on both sides of one tax. You collect six on the hours you sell and pay three on the systems that schedule them. A twelve‑guard company on a $105 THERMS subscription is paying $3.15 a month in Maryland sales tax on its software, which is not a number worth restructuring anything over, but it is a number that should appear on a quote rather than as a surprise on an invoice. Ours does.

The 6% on the other side is a much bigger operational fact than it looks. It means your invoice is a tax document, that the rate applies to the taxable price of the service rather than to some subset of it, and that a resale exemption exists but has conditions. §11‑101(n)(3)(iii) excludes from “use” the exercise of a right over a taxable service acquired for sale “if the buyer intends to resell the taxable service in the form that the buyer receives” it. Guard companies subcontract constantly — a Baltimore firm covering an overflow shift in Frederick, a prime contractor filling a county‑line site with a partner’s officers. Whether those hours come to you exempt turns on whether you are reselling them in the form you received them, and a prime that receives straight‑line post hours and rebills a blended supervised rate should have a considered answer to that question rather than a habit. This is a genuinely fact‑specific area and it belongs with your own tax adviser, not with a blog; what belongs in your software is the evidence trail that lets the adviser answer it.

The same officer, the same night, two different answers

Two carve‑outs in the Comptroller’s own list make the point better than we could invent it.

Under the exemption for licensed caterers — which lets a caterer buy materials, equipment and supplies for resale as part of a catering contract — the Comptroller lists the items that qualify. Audio and visual equipment. Decorations. Generators, HVAC systems and portable restrooms. Lighting. Pipe and drape. Tents. And security services. So when a licensed caterer books your officers as part of a catering contract for a wedding at a Harbor East hotel, the Comptroller’s own framing treats that as a purchase for resale, and the tax follows the hours downstream onto the caterer’s invoice rather than sitting on yours.

Under the film production exemption, the same document lists the services that are exempt when sold to a certified film producer and used directly in production — editing, sound mixing, animation, photography on set, script and costume fabrication — and then lists the services that are not eligible. The first entry on that list is “Bodyguard or security services”.

Baltimore is a production city. On a closed street in Fells Point, the sound equipment is exempt, the video and digital editing is exempt, the costume fabrication is exempt — and the officer standing at the tape at the end of the block is taxed at six percent. Walk the same officer four blocks to a caterer’s wedding and the same hour is bought for resale. Same person, same badge number, same night. The tax answer is a property of the buyer, and nothing on a shift record captures the buyer’s status.

This is the argument for building the client record properly rather than as a name and an address. A client is a legal character — a licensed caterer, a certified film production company, a tax‑exempt organization with a Maryland exemption certificate, a federal agency, an ordinary commercial buyer — and the character decides the tax treatment of every hour you sell them. Most guard software stores clients so that shifts can be assigned to them. That is a scheduling model. The invoicing model needs a different shape.

The uniform that is taxed when you buy it and taxed again when you wash it

A small one, but it recurs monthly and it lands entirely on the cost side. Uniforms are ordinary tangible personal property and are taxed at 6% when purchased; the Comptroller’s caterer list is careful to say that costumes qualify for that exemption “but not standard uniforms for staff”, and the film list puts crew uniforms squarely among the items that are not eligible. Then §11‑101(m)(2) makes “commercial cleaning or laundering of textiles for a buyer who is engaged in a business that requires the recurring service” a taxable service in its own right. If you launder uniforms under a recurring commercial contract, the washing is taxed as well as the shirt. It is a small line and it should be a tracked line, because it is per‑officer, it recurs, and it is exactly the sort of cost that gets buried in an overhead bucket and never makes it into a post‑level margin.

What custom software actually costs

We publish our prices, which in this category makes us unusual enough to feel slightly self‑conscious about mentioning it.

Set that against the meters. A twelve‑guard Baltimore company on THERMS spends $1,260 a year, and that is a fair price for what it does. But the same company running one 24‑hour post with four officers instead of five is spending $3,536 a year on structural overtime for that post alone — nearly three times the subscription, on one contract, before anything has gone wrong. Run three such posts and the number is $10,608. The argument for building is not that the subscription is expensive. It is that the subscription is not measuring the expensive thing.

What we would actually build

Everything above points at the same missing object: a system whose unit of account is the post, with people attached to it, rather than a list of shifts with names on them.

The first thing we would build is the coverage model of your own book. Six months of scheduled hours, actual clocked hours and billed hours, joined post by post. That produces something no vendor can sell you, because it is about your contracts and not about the category: the real cost of an hour of coverage at each site, including the overtime that the roster shape makes unavoidable, the relief hours nobody planned, and the callout hours that turned straight time into time and a half at nine on a Sunday evening. Almost every guard company we have spoken to knows its bill rates precisely and its post‑level margins approximately. That gap is where the money is, and it is a week of work to close it.

The second is the credential register, which is the thing that turns a roster into a compliance object. Every officer carries a certification with an expiry date, a training record with two different hour requirements on a three‑year cycle, possibly a commission tied to a named property, possibly a handgun permit that under §19‑413 requires three separate conditions before a single armed shift — employer approval, the guard’s own notification to the Secretary, and the Secretary’s confirmation — and a badge number that Maryland says must be unique and consecutive since the company’s first day. The register is not hard. What is hard, and what nothing in the category does, is making the schedule refuse to place a person on a post they are not currently credentialed for, and making the ninety‑day and forty‑eight‑hour queues appear on somebody’s Monday morning instead of in a folder.

The third is the threshold monitor, and it is embarrassingly cheap to build. Three numbers on a dashboard: how many individuals are currently employed as security guards, computed the way §19‑504 computes it; the trailing twelve‑month average monthly employee count, computed the way §3‑1304 computes it, counting everybody; and the maximum active user account count this billing period, computed the way your vendor computes it. Three numbers, three different populations, three different arithmetics, each with the threshold it is approaching. That is perhaps two days of work and we have never seen it in any product in this category, because no product is looking at all three books at once.

The fourth is the incident record built for §19‑414 rather than for a narrative. A timestamped receipt event when the officer’s form arrives, which is what starts your own 48‑hour clock. Structured fields for encounter type, force type, location, arrest and charges, medical care for both parties, and the demographic fields the statute names. A clear escalation for the escort‑with‑minimal‑resistance judgment, so that the decision not to file is itself a record with a name against it. Most incident modules in this category are a photo, a free‑text box and a timestamp. That is a good product for a client report and a poor one for a statutory filing.

And the fifth is the storefront, which is the part of this trade most under‑built relative to how well it would work. More on that below.

Build, buy, or both

We will say the same thing to you on a call that we say in every one of these articles, which is that most companies should keep buying most of their software.

If you run eight officers across three sites and your problem is that you cannot prove a patrol happened, buy something. THERMS at $105 a month with a full published ladder, or GuardsPro at $8 a user, will do guard tours, checkpoints, incident capture and a client portal better than anything we could build for $12,000, and they will do it on Monday. If you run four hundred officers across sixty sites with union agreements and a payroll integration, buy TrackTik or WinTeam and negotiate hard, because the integration surface alone is a multi‑year project and the module pricing is designed to be negotiated.

The case for building starts in a narrow band, and it is a band a lot of Baltimore companies are sitting in right now:

• You run somewhere between roughly ten and eighty officers — too big for a spreadsheet, too small to be interesting to an enterprise vendor.
• You have at least one 24‑hour post, and you could not say from memory whether it is covered by four people or five.
• You bill by the hour and you cannot produce a margin by post without exporting two systems into a spreadsheet.
• You hold at least one special police commission, so part of your roster is tied to a named property.
• Your certification expiries live in a calendar, and your badge numbers live in a drawer.
• More than half of your new work arrives as an inbound phone call that somebody has to price by hand.

If three or more of those are true, the honest recommendation is usually not a replacement. It is to keep the guard tour product you already have, and build the two things it will never do: the coverage and credential model, and the storefront. That is a Prototype Sprint at $3,500 and an Online Store from $6,000, and it leaves your existing subscription exactly where it is.

The store this trade should have and mostly does not

We think the online store is the strongest single argument in this category, and it rests on something specific to security work: a large and growing share of the demand is short‑notice, well‑defined and small — and it is currently answered by a phone call, if it is answered at all.

The industry sells itself on contracts: twelve months, a fixed post, a monthly invoice. That is the good business and it is not going anywhere. But sitting beside it is a stream of work that behaves like retail. A church needs two officers for a funeral on Saturday. A Canton bar needs door coverage for a one‑off event and does not want an unarmed employee doing it. A property manager needs vacant‑unit checks for six weeks between tenants. A construction site needs overnight coverage until the fence goes up. A production company needs three officers for two nights of a street shoot. A wedding venue needs somebody at the car park until one in the morning. Every one of those is a defined window, a defined post type, a defined location and a defined minimum — and every one of them is currently a voicemail on a Friday afternoon.

What a real store looks like here is not a shopping cart with guards in it. It is a booking tool that asks what the job is: what kind of site, what hours, armed or unarmed, one officer or three, indoors or on a public‑facing door. It quotes from your own rate card with your own minimum — four hours is the trade’s usual floor and it should be enforced by the software rather than argued about afterwards. It knows the difference between a Tuesday afternoon and a holiday overnight, because your cost of covering them is not the same. It applies the 6% correctly, and it asks the one question that decides it: who is buying, and are they a licensed caterer, a certified production company or an exempt organization with a certificate to upload. It generates the written agreement. It takes a deposit by bank transfer before it offers a card. And it tells the customer, honestly, whether you can actually staff it — because the credential register knows who is available and who is currently credentialed for that kind of post.

Then it sells the things that ride along on the same officer at proper margin: additional hours, holiday and overnight rates, a second officer, vehicle patrol passes, vacant‑property inspection routes, key‑holding and alarm response, escort‑to‑car services for a hospital or a late‑shift employer, temporary camera hire, and the recurring contract that turns one Saturday funeral into a parish that calls you every time. The one‑off job is not the revenue. The one‑off job is the acquisition channel for the contract, and at the moment most companies in this trade are letting it go to voicemail because nobody is in the office at five on a Friday.

That store is yours. It runs on your customer list, at your margin, on your own domain, and it is the one asset in this business that appreciates while the vehicles depreciate and the officers turn over. Our Online Store package starts at $6,000, fixed, and you own it outright.

Who we are

We are a small studio of ex‑startup founders in Baltimore. We build web apps, online stores and operations systems at fixed prices, we work directly with the people who write the code, and you own everything at the end — the repository, the data, the infrastructure accounts.

We are not a physical security specialist and we would be suspicious of anybody claiming to be one after a fortnight of reading. What we are good at is the part of this that is a data problem: taking a roster that three different rules count three different ways, fifteen statutory clocks that run on people rather than on shifts, and a set of contracts priced by the hour but costed by the person, and turning them into a handful of tables that agree with each other on a Monday morning. Everything in this article — the CBP extraction, the coverage arithmetic, the sixteen pricing URLs, the band ladder — is the same work we would do in the first week of a project, done in public.

Questions Baltimore security companies actually ask

How much does security guard management software cost in 2026?

Between about $45 and several thousand dollars a month, and most of the category will not tell you. Checked on 26 August 2026: THERMS publishes a complete ten‑band ladder from $45 a month for one to five user accounts up to $1,120 for 301 to 400, and states that billing is calculated on the maximum number of users your organization had during the billing period. GuardsPro publishes $5, $8 and $10 per user per month, or $4.25, $6.80 and $8.50 annually, with a three‑user minimum. ShiftFlow publishes $5.99 per seat per month or $60.00 per seat per year. Celayix publishes $400–$800 a month up to 100 employees and $1,000–$8,000 up to 1,000, with a 25‑license minimum on an annual contract. TrackTik — which now also serves the trackforce.com, trackforcevaliant.com and silvertracsoftware.com domains — serves a pricing page with no dollar figure that says it is priced by sites, not seats. Belfry says it charges per employee per month and publishes no rate; its /pricing URL returns a 404 while /plans returns 200. TEAM Software (WinTeam, Lighthouse) publishes nothing, and OfficerReports, GuardMetrics and Novagems all return 404 on their own /pricing URLs.

How many security guards can I employ before my insurance requirement changes in Maryland?

Four. Md. Bus. Occ. & Prof. §19‑504 requires an agency or employer that employs five or more individuals as security guards to carry commercial general liability insurance including errors and omissions and completed operations with a $1,000,000 total aggregate minimum, and one employing fewer than five to carry the same cover at $500,000. Proof goes to the Secretary of State Police. The test is a headcount of individuals — not full‑time equivalents, not hours — so a part‑time weekend relief officer counts exactly as much as a full‑timer. If the cover lapses, the agency license is suspended automatically and stays suspended until proof is filed; for a security guard employer, the Secretary may prohibit it from employing a certified guard at all. These requirements took effect on 1 January 2025 under Senate Bill 760 of 2024.

What training does a Maryland security guard need in 2026?

Twelve hours before certification and eight hours before each three‑year renewal, both approved by the Maryland Police Training and Standards Commission. §19‑412(b) requires those programs to cover relevant criminal law, the appropriate use of force, the use of alternatives to force, de‑escalation techniques, and appropriate interactions with people under 18, people with behavioral health or other disabilities, and people in crisis. Anyone employed as a police officer within the three years immediately preceding the application is excused the twelve hours on producing documentation. A special police officer is a different credential with a different number: an initial commission under Md. Pub. Safety §3‑303(c)(2) requires at least 80 hours of instruction covering criminal law, the constitutional requirements for search, seizure and arrest, and the appropriate use of force, with 12 hours of in‑service training before each three‑year renewal.

Do I need a security guard agency license in Maryland, or can I register as a security guard employer?

It depends on who your guards protect. §19‑101 defines a security guard agency as a person who conducts a business providing security guard services, and a security guard employer as a person who employs guards only to provide services to itself. Sell coverage to other businesses and you need the agency license; guard only your own property and you register as an employer. Either way the individual officers must be certified. The agency application fee is $200 for an individual and $375 for a firm; renewal is $200 and $400; the license runs three years and expires in January or July; and the Licensing Division states a decision within 90 days. Two requirements catch people out. The individual applicant, or for a firm the appointed representative member, must be at least 25 years old under §19‑302(d); and under §19‑303 that person must have five years as a full‑time police officer, certified private detective, fire investigator or correctional supervisor, or three years in an investigative capacity as a detective. The founder’s career history is, in Maryland, a licensing requirement.

Are security guard services subject to Maryland sales tax?

Yes. Md. Tax‑General §11‑101(m)(10) makes “a security service” a taxable service and names detective, guard and armored car services and security systems services expressly, so a Maryland guard company charges 6% on what it invoices. Two carve‑outs are worth knowing about. The Comptroller’s published list of taxable property and services includes security services among the items that qualify for the licensed caterer exemption, so a caterer can buy them for resale as part of a catering contract. The same document lists “Bodyguard or security services” among the services not eligible for the film production exemption, so an officer on a certified Baltimore production is taxed while much of the equipment beside them is not. Separately, §11‑101(m)(14) and (15) made data and IT services and software publishing taxable at 3% from 1 July 2025, which is the rate on the software you buy. Take your own position to a tax adviser rather than to an article.

What is a special police officer in Maryland, and how is it different from a security guard?

A special police officer holds a commission issued by the Governor under Md. Public Safety Title 3, Subtitle 3, and exercises the powers of a police officer only on the property described in the application. A security guard holds a certification card from the Secretary of State Police and has no police powers at all. The practical differences are large: the commission is applied for by the employer, one application per individual, requires at least 80 hours of initial instruction, costs $100 to apply for and $60 to renew, runs three years, and requires the officer to take the constitutional oath before a circuit court clerk within 30 days and before performing any duty. Under §3‑308(b) the entity that requested the appointment is also responsible for any abuse of the commission’s powers “either on or off the premises”, so the liability is not bounded by the property line even though the authority is. And under §3‑313(c) a commission does not terminate when the building is sold, provided the buyer executes a form assuming the original employer’s responsibilities — the commission can follow the real estate rather than the payroll.

How quickly does a Maryland security company have to report a use of force?

There are two clocks, in series. Under §19‑414(b)(1) the guard reports to the agency or employer within 48 hours on a form provided by the Secretary, unless seriously injured or disabled. Under §19‑414(c)(1) the company then reports to the Secretary within 48 hours after receiving that form — so the second clock starts on receipt, not on the incident, and your deadline is set by when your officer files. The report must carry the type of encounter, the type of force, the location, whether the person was arrested and on what charges, whether either party requested or required medical care, and demographic information including race, ethnicity, gender and age for everyone involved. “Use of force” is defined to include any physical striking, any significant physical contact restricting movement “including control techniques”, detainment without consent, and the discharge of a firearm, pepper mace or an electronic control device. It excludes “mere presence, verbal commands, or escorting an individual with minimal resistance” — and minimal is not defined anywhere in the statute.

Is custom software cheaper than a guard management platform for a security company?

Not as a straight replacement, and we will tell you so on the call. A twelve‑guard Baltimore company on THERMS spends about $1,260 a year, and nothing we build for $12,000 pays that back by replacing guard tours. The arithmetic changes beside the subscription rather than against it. The same company covering one 24‑hour post with four officers instead of five is paying $68 a week — $3,536 a year — in structural overtime on a $17.00 wage, for one contract, before anything has gone wrong; three such posts is $10,608. It is also carrying fifteen statutory clocks that run on individual people rather than on shifts, three headcount thresholds that count three different populations three different ways, and a badge register that Maryland regulation specifies down to the numbering scheme. Our fixed prices are a $3,500 Prototype Sprint, an online store from $6,000 and a custom app or operations system from $12,000, and the case is almost always made on coverage arithmetic, credential expiry and a bookable storefront rather than on the subscription line.

Method and sources. Vendor pricing was read directly from each company’s own published pages on 26 August 2026; HTTP status codes are what each server returned to a plain desktop request on that date, and where a URL returned a 404, a redirect or a page without a price, we have said so rather than quoting an aggregator. Prices change without notice, and the ladders, per‑account figures and step sizes derived from them are our arithmetic. Establishment, employment and payroll figures are from the US Census Bureau, County Business Patterns 2023 county file, NAICS 561612 and 561621, extracted and totaled by us; CBP counts only establishments with paid employees, so sole proprietors without staff are excluded and the real population of the trade is larger. Statutory citations are to the Annotated Code of Maryland: Business Occupations and Professions Title 19, Public Safety Title 3 Subtitle 3, Labor and Employment §§3‑413, 3‑415, 3‑420 and 3‑1304, and Tax‑General §11‑101; regulatory citations are to COMAR 29.04.01. Fee schedules, processing times and portal requirements are from the Maryland State Police Licensing Division’s own security guard agency license and security guard certification pages. The taxable and exempt lists are from the Comptroller of Maryland’s List of Tangible Personal Property and Services Subject to Sales and Use Tax. The repeal of Public Local Laws §§16‑1 to 16‑16 is as printed in the Baltimore City Law Department’s own compilation of the Code of Public Local Laws, dated 20 December 2024. Wage figures used in the coverage model are illustrative and drawn from published Baltimore market ranges for unarmed work in 2026; the coverage, overtime, accrual and threshold arithmetic built on them is entirely ours. The composite security company in this article is ours, built from the shapes of businesses we have talked to, and is illustrative rather than a real client. Nothing here is legal, tax, insurance or regulatory advice — verify your own position with the Maryland State Police Licensing Division, the Comptroller of Maryland, your insurance broker or your own advisers before relying on any of it.