Animal Shelter Software Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does the hold clock keep getting built as a date field?

The scope conversation almost always goes the same way. Somebody asks how long a stray is held. The answer is five days, or seven, or whatever the ordinance says. That number goes into a specification, a developer adds a hold end date to the animal record, and the meeting moves on. Eight months later an animal seized as evidence in a cruelty case is adopted out, because the evidence hold was never a hold at all. It was a sentence in a free text note that the outcome screen never read.

This failure is specific to animal services because one animal routinely carries several obligations at once, each from a different authority. The stray clock comes from your ordinance. A bite quarantine comes from the county health department, runs the standard ten day observation period for a dog or cat that bit a person, and may apply to an owned animal held at home rather than in your building. An evidence hold comes from a prosecutor and has no end date until a court gives one. Owner surrenders follow different rules again. One field cannot hold three obligations, so staff keep the other two in their heads and on the kennel card.

The fix is to type the holds. Each becomes its own record carrying a source of authority, the counting rule for your jurisdiction including how holidays and closure days are treated, and an end condition. Any action clearing an animal for outcome checks for open holds and refuses unless a named person records an explicit override with a reason. Ask a developer to draw that on a whiteboard before you sign. The ones who have done this work separate the hold types unprompted.

What goes wrong when years of animal history get migrated?

Shelter data is among the messiest small datasets in local government, and migration is where these projects quietly lose a month. Outcome codes have usually been redefined two or three times over a decade, often informally, so the same word means different things depending on the year and the person who typed it. Medical history sits half in structured vaccine records and half in free text. Adopter records are duplicated because the same person has three email addresses. Microchip numbers arrive in four formats. Photographs live in a folder named by an identifier that stopped matching the database after a system change in 2019.

The consequence that stings is the live release rate. Once outcome codes are mapped correctly to the standard national definitions, your historical figure moves. If nobody prepared the board for that, the software project becomes a credibility problem rather than an operational one, and you spend a committee meeting defending a number instead of demonstrating a system.

Treat migration as a distinct workstream with its own budget line, not a step at the end. Bring full structure across for recent years and a searchable archive for older records. Map outcome codes deliberately, with a supervisor deciding each ambiguous case rather than a script guessing. Publish the restated historical figure before go live with the mapping attached, so the change is your announcement rather than somebody else's discovery.

Why do microchip, syndication and payment integrations break after launch?

Integrations in this sector are mostly one way and mostly unmonitored, which is a combination that fails quietly. Adoption listings syndicate to public sites. Microchip lookups reach registries that each present a slightly different interface. Licensing payments run through a processor whose credentials expire. Jurisdiction attribution may depend on an address or parcel lookup. None of these fail loudly. The feed simply stops, and you find out when a volunteer mentions the website still shows a dog adopted three weeks ago.

The second cause is that these connections get built against whatever the partner exposed on the day, with no contract test and no alerting. When the partner changes a field, the feed keeps returning a success response and drops the record. Nobody notices, because nobody is watching a number that would move.

The fix is cheap if it is in scope from the start and awkward to retrofit. Every outbound feed gets a visible last success timestamp on an operations dashboard a real person checks, plus an alert when the gap exceeds a threshold you set. Failed records land in a queue with an owner rather than a log file. A reconciliation job compares what you believe is listed against what is actually listed. Put this in the first release, not the second.

What happens when the controlled substances log and bite quarantine are not covered?

These two gaps appear in almost every shelter build scoped from the adoption side inward, because neither is visible if you start at the animal profile screen and work outward.

Euthanasia solution is a Schedule II controlled substance and your registration depends on an accurate log. If the build does not record administration at the moment it happens with witness confirmation, and reconcile against inventory on a schedule, your clinic keeps a paper book alongside the new system. That book is now the legal record, transcribed quarterly by somebody who was not present, and you have paid for software that made your compliance position worse.

Bite quarantine fails differently. It serves a public health purpose rather than an ownership one, and it often applies to an animal that is not in your building. If the system can only place holds on animals in kennels, home quarantines go into a spreadsheet, inspection checkpoints get missed, and release depends on somebody remembering. When the health department asks for the record, you hand over a spreadsheet.

Both are inexpensive to build and expensive to retrofit, because both change the shape of the hold model. Name them in the specification yourself. A developer who has never heard of controlled substance reconciliation is not disqualified, but they will not raise it in discovery, and silence there costs you a phase two.

Should you build custom or configure what you already own?

Plenty of organisations reading this should configure Shelterluv properly and stop. If you are a rescue or a humane society with no field operation, Shelterluv handles intake, medical records, foster communication and the adoption pipeline better than any first release you commission, and it connects you to adoption syndication and national reporting without effort. PetPoint is the pragmatic choice where breadth and national reporting matter more than workflow polish. Chameleon and ShelterBuddy both hold up in municipal settings. Spending six figures to arrive at a worse version of Shelterluv is a real risk and we have talked organisations out of it.

The honest test is whether your operation is one business or several. A shelter is one data model. A public veterinary clinic is another. A licensing authority is a third. A law enforcement function with officers, citations, evidence and court dates is a fourth. Packaged products do two or three of those well and the rest thinly, which is why municipal departments so often run a fully digital shelter beside a paper dispatch log.

The tipping point is the field officer. Once your operation includes people responding to calls in vehicles with legal authority, packaged products stop fitting and the workaround becomes paper nobody can report on. A common middle path keeps the packaged product for adoption and foster and builds the field, licensing and contract reporting layer around it. That targets the half that is actually broken.

How do hidden costs get into the quote?

Six things reliably arrive after the number is agreed. Contracted jurisdictions come first, because each city brings its own ordinance, fee schedule and report format, so a price built around one municipality does not scale linearly to six. Licensing is second: it sounds like a form and is actually renewals, delinquency notices, online payment, refunds and reconciliation. The clinic is third, if it serves the public, because that is veterinary practice management and a project in its own right.

Offline mobile is fourth. A field application that keeps working with no signal and syncs afterwards is genuinely harder than a responsive web page, and it is not optional, since the parts of your county with the worst coverage generate the most calls. Migration is fifth, and it is routinely quoted as a small fixed line by firms that have never opened a shelter export. Training and parallel running is sixth, and it lands as real staff overtime if the go live date was set by the project plan rather than by intake volume.

Protect yourself two ways. Make the developer scope migration separately, after examining an actual export. And ask for offline behaviour to be demonstrated rather than described. Ask what happens to a citation written in a dead zone when the officer drives back into coverage. A vague answer means the price is not real.

What separates a build that works from one that fails here?

The builds that work start inside the building. Staff pain lives at intake, in the kennel, in the clinic and in the vehicle, not on the public adoption page. Teams that begin with the public site produce something attractive that changes nothing, and by the time money is needed for the operational core the appetite has gone.

They treat the record as evidence rather than data. Animal history supports cruelty prosecutions, public health decisions and contract invoicing, and it will outlive several software vendors. That means an append only audit trail, overrides that capture who authorised what and why, and a bulk export you can run yourself without asking anyone.

They have one named internal owner with authority, usually a supervisor who still works intake shifts. Committees produce feature lists. One person who knows what Friday evening before a holiday weekend looks like produces a system.

And they settle ownership before kickoff. The department or the organisation should own the repository, the cloud accounts and the right to hire another firm. At Digital Heroes the client owns the code from the first commit. In animal services that carries extra weight, because your records are public and legal documents that must remain producible for decades regardless of who maintains the software.