Access Control Audit Trail Setup in Houston, TX 77055

What an access event actually proves

A line in an access log says one narrow thing: a credential number was presented at a named reader at a named second, and the controller unlocked. Everything assumed beyond that has to be built in deliberately, and in most systems we inherit along Long Point it never was.

Three gaps separate that line from a fact:

• Identity. The panel authenticated a card, not a face. Share one fob between two people, or copy a low-frequency 125 kHz prox card at a kiosk, and the log confidently names the wrong person.
• Physical reality. Unless a contact is wired into the door, “granted” only means the lock released — the person may have walked away, or the door may already have been open.
• Count. One grant can mean four people entered. Nothing in the electronics objects to that.

Audit trail setup closes those three gaps until the record and the building agree. On a small Spring Branch shop that is usually half a day of hardware and half a day of configuration, not a new system.

No door contact, no real audit trail

The most common thing missing on a retrofit here is a door position switch — a recessed magnetic contact in the frame telling the controller whether the leaf is open or shut. Without it the panel is blind between unlock commands and three useful event types never appear:

• Door forced open — the door opened with no grant and no exit request. This is the event that catches a jimmied strike or a taped latch.
• Door held open — the door stayed open past a set number of seconds. In August, on a shop with a hot back room, this is the event you will see most.

The exit side matters too. A request-to-exit sensor or push-bar switch tells the panel an inside egress was legitimate, so no forced alarm fires — but a REX event records that somebody left, never which somebody. If knowing who left matters, the exit leaf needs its own reader, decided before the conduit goes in.

Roll-up doors in the flex units off Hempstead have their own quirk: a standard frame contact cannot survive a sectional door’s travel and vibration and will false-alarm within a week. Those need overhead-door contacts with the correct gap rating, mounted to the track rather than the panel.

One man-door, five tenants, one keypad code

Much of 77055’s small-business stock is multi-tenant: a row of suites behind one common man-door, a shared corridor, a landlord-installed keypad from a previous decade. That layout defeats an audit trail faster than any technical fault.

A shared numeric code produces log lines reading “code accepted, 6:12 a.m.” forever. No person is attached, nobody can revoke it without telling four other businesses, and the day you need to know who was in the corridor at six, the answer is “one of about forty people.” If the common door must stay on a code, give each person a code rather than each business one code — and put your own credentialed reader on your suite door.

We also settle who owns the log: common-door events usually stay with the landlord, suite events with you. Written down at setup, that prevents the case where something happened at the shared door and each party assumed the other was recording it.

The clock is the part that fails quietly

Access panels keep time on a small battery-backed real-time clock, and those drift — a minute here, several over a year, faster in a hot electrical room. Your camera recorder drifts on its own schedule. Neither device warns you and both keep working, so the fault stays invisible until the afternoon you need the two records side by side.

Then it reads like this: the log grants the back door at 4:47:12, the footage shows it opening at 4:51:40, and you are eyeballing four minutes of video wondering whether this is even the right person. A record you have to argue with is not evidence.

The fix is to point every device at the same time source and confirm it held:

• Network time on the controller, the management software and the recorder — one source, all three.
• Time zone and daylight-saving rules per device. Central time with automatic DST is the setting people miss, and it shows up as a system exactly one hour wrong for half the year.
• A written verification — present a credential, note the second, open the clip, confirm the frames line up.

Where your events live, and how long they survive

There are usually two copies of your history and they do not last the same length of time.

The controller holds a fixed buffer of recent events — a set number of lines, typically thousands, overwritten oldest-first once full. It exists so the door keeps deciding and recording while the network is down. It is not archival storage; a two-shift shop can roll a small buffer in weeks.

The second copy is the database — a local workstation or a hosted service — where reporting and long-term history live, and the copy lost when the office PC is replaced and nobody exported first. We have opened cabinets here where the panel was healthy at eight years old and the entire searchable history was three weeks, the buffer being the only copy left.

What we set at commissioning:

• Polling frequency, so events move off the panel into the database instead of pooling until something overwrites them.
• A retention target chosen on purpose — long enough to cover how you actually find problems, which for most shops is a season, not a week.
• On hosted systems, a standing export into your own storage, so a lapsed subscription cannot walk off with your history.

The second log: changes to the system itself

Door traffic is half the record. The other half is administrative: who added a credential, who deleted one, who changed a schedule, who pushed a remote unlock from a phone, who logged in at all. Most platforms keep that history separately and switch it on by default — then it is defeated on day one because everyone signs in as the same “admin” account.

With one shared login every administrative line names the same user — the shared keypad problem again, moved into the software. We build named operator accounts, scoped so a supervisor can run a report and re-issue a fob but cannot widen a schedule or purge history, and leave a written list of who holds what.

Remote unlock earns its own mention: useful for letting a parts delivery in at seven, and the event most often misread later, because it shows a door opening with no credential at all.

What we do on an audit trail call in 77055

1. Read what you have, then walk every opening. A week of existing events shows what the system can record versus what it is recording. Then: position switch present? Exit hardware correct? Existing contact aligned, or shimmed into permanent “closed”?
2. Fix the hardware gaps. Add contacts, correct REX behavior, retire shared codes, and where the head-end supports it move readers off unsupervised Wiegand to a supervised OSDP run, so a cut reader line raises its own event.
3. Set time and retention. Network time on everything, DST confirmed, buffer and polling sized to the door’s traffic, backup configured.
4. Run a deliberate test. We generate each event on purpose — valid entry, denied credential, door held past its timer, forced open — and confirm each lands with the right label and the right second.
5. Prove the export and hand over the sheet. The report you would need in a real incident, produced from your account on your screen before we leave, plus one page covering where the log lives, how to filter it, how long it keeps and who to call.

Five things that quietly ruin an audit trail

• Recycling card numbers. Handing a returned fob to the next hire makes every historical line for that number ambiguous. Retire numbers; do not reissue them.
• Deleting instead of disabling. On some platforms removing a person removes or orphans their history. Disabling keeps the record and still ends the access.
• A contact that never reports open. A magnet shimmed to stop a nuisance alarm kills forced-open and held-open events permanently, and looks identical to a quiet door.
• Nobody owning the report. A log nobody reads is discovered during an incident, which is the worst possible time to learn it was misconfigured.
• An unsecured panel. A controller in an unlocked shared closet on factory credentials, reset button in reach, is a record anyone in the building can edit.

What moves the price on a Spring Branch job

We quote itemized after seeing the doors, never over the phone, because the distance between “change six settings” and “pull cable through a 1962 masonry wall” is enormous. What moves it:

• How many openings need a position switch added, and whether the frame is hollow metal, wood or masonry.
• Whether a pathway exists. Older slab buildings along Long Point often have no accessible attic over the back half, meaning surface raceway rather than a fished wall.
• Whether the existing head-end can do what you need or is a keypad with no reporting at all.
• Reader wiring — keeping existing Wiegand runs versus converting to supervised OSDP.
• Local software, a hosted service, or both — and how many operators need training.

The estimate is free, on site, and itemized before anything is committed.

Related services

Frequently asked questions