Property Assessment and CAMA Software Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does reproducibility get designed out of the system?

Not through carelessness. Through ordinary good practice applied to the wrong problem. Almost every database habit favours current state: update the record, keep it clean, avoid duplication. Cost tables get revised for the new year in place. A land schedule is corrected. A depreciation curve is tuned. A sale is reclassified as invalid and removed from the comparable pool. Each change is defensible on its own, and together they destroy the office's ability to answer the only question that matters at a hearing.

The requirement is unusual and it needs stating explicitly, because a developer who has not worked in assessment will not infer it. Every assessment year has to be an immutable snapshot. All valuation inputs effective dated. The model itself versioned. Every value carrying a reference to the model version and the input set that produced it. Recalculating a prior year should return the identical number without exception, forever, because appeals, corrections and litigation reach back years and the state oversight review will ask.

The reason this gets cut is that it looks like overhead in a demonstration. Nobody asks to see the 2024 roll in a sales meeting. They ask to see the valuation screen, and a system that stores current state produces that screen faster and cheaper.

The test to apply before a contract is signed: ask how they would reproduce a value from three years ago exactly. If the answer does not involve versioned models, effective dated tables and stored inputs, everything else in the proposal is decoration. This is the design decision worth arguing about hardest.

What goes wrong when you convert parcel, sales and exemption history?

Conversion is nearly always the largest single line in an assessment project and it is where the true condition of your historic data becomes visible for the first time. Offices are routinely surprised, and the surprise arrives against a statutory calendar that does not move.

The recurring defects are specific. Parcel history that goes quiet at the point of a split or merge, because the legacy system treated the parcel number as the primary key and the workaround was a new record with a note. Sales flagged invalid with no recorded reason, which is unusable in a ratio study and indefensible if challenged. Exemptions stored as flags rather than dated entitlements, so nothing in the data says when a senior exemption was granted, on what evidence, or when it should expire. Characteristics with no source, so nobody can say whether a finished basement came from an inspection, a permit or a phone call in 2009.

Do not attempt to reshape all of it. Profile first, then decide deliberately what converts clean, what converts marked as pre conversion with known limitations, and what stays in a readable archive rather than being forced into a new schema. Publish that decision to the appraisal staff, because they are the ones who will be asked about a 2011 characteristic at a hearing and they need to know what the system can and cannot tell them.

Model parcel lineage explicitly in the new system, with predecessor and successor relationships and dated events, so history survives every future split and merge. Doing that at conversion is cheap. Retrofitting it later means reconstructing splits from deeds and plats by hand.

Why do GIS, recorder, permit and tax billing integrations break after launch?

Because each one belongs to a different office with its own priorities and its own upgrade schedule, and none of them is obliged to warn you. The recorder changes a document type code. The permit system is upgraded and the address format shifts. The geographic information system layer is republished with new parcel identifiers after a reprojection. Tax billing changes its extract cadence at year end. Each of these is routine work for the owning department and an outage for the assessor.

The failures that follow are quiet rather than loud. A permit feed stops arriving and nobody notices for two months, so improvements go undiscovered on parcels that were being inspected on that trigger. A geographic identifier mismatch means a hundred parcels silently stop joining to their map geometry, and they only surface when a public portal user reports a blank map.

The pattern that survives is unglamorous. Alert on silence as well as on error, because a feed that stops looks exactly like a quiet week. Reconcile counts daily between the source and what landed. Version the parser per source rather than writing one importer that tries to cope with everything. And build the inter office relationship deliberately: a named contact in each department who knows to tell you before a change, which is a governance job rather than a technical one and is frequently the thing nobody assigns.

What happens when ratio study and state reporting are not covered?

You find out from someone else's study. Your state revenue department runs its own analysis, and learning from it that a neighbourhood is out of tolerance is the wrong time to learn it, because by then the roll is set and the remedy is next year.

The professional framework here is well established through guidance from the International Association of Assessing Officers on ratio studies, with measures such as the coefficient of dispersion and the price related differential used to test uniformity and vertical equity. The problem is not that offices disagree with the framework. It is that the study is run once a year by one analyst, after the fact, so it functions as a report card rather than a control.

Build it as a continuous measure instead. Run the ratio study against the current model by neighbourhood, class and value range, and flag drift while there is still time to correct the model. That single change converts an annual audit into a working tool.

The related trap is sales chasing, meaning selectively adjusting recently sold properties so the ratio study looks better without improving equity for anyone else. It is a well known failure mode and a system that makes it easy is a liability rather than a feature. Make sales validation an explicit, evidenced step with the reason for every exclusion recorded and reviewable, because an unexplained exclusion is the first thing a challenger or an oversight reviewer will find.

Should you build custom or configure what you already own?

For most offices the honest recommendation is to keep the CAMA system of record and build only the layer around it, and we would say that in a procurement meeting. Tyler iasWorld is genuinely deep and supports large complex jurisdictions. Vision Government Solutions, Vanguard Appraisals and Patriot Properties each serve their markets well and encode a great deal of statutory practice, particularly in the regions their products grew up in. Rebuilding that from nothing is a multi year programme with a statutory deadline attached, which is a bad combination.

What those products cover least well for any specific office tends to be the same three things: valuation modelling flexibility, appeal evidence assembly, and field data collection. That is why the pattern we see most often is not replacement. It is an office keeping its roll of record and building the layer where the pain actually is.

Build a full replacement only in narrow cases. A very large jurisdiction where the economics change. A state whose statutes no product fits without heavy localisation, where localisation delivered as configuration will stay fragile. Or a statewide programme serving many counties, where the build amortises across all of them. That last case is real and it is where custom mass appraisal work is most often justified.

How do hidden costs get into the quote?

• Data conversion. Nearly always the largest line and frequently the vaguest. Insist on a profiling phase priced separately, before the conversion itself is estimated.
• Exemption and abatement programmes. Each has its own eligibility, evidence and calculation. A quote that treats them as a count of record types has not read your statutes.
• Income approach depth. Commercial valuation with income models is materially more work than cost and sales for residential, and it is often assumed rather than scoped.
• Accessibility. Public facing pages for a government body carry standards that are not optional, and retrofitting them costs more than building to them.
• State reporting formats. Prescribed, and they change. Ask who absorbs the cost when they do.
• Public records and retention obligations. Records schedules and public records request handling are real constraints on the design, not policy documents to be signed afterwards.

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

The builds that work generate the appeal packet from the same data that produced the value, so the packet cannot contradict the roll. Subject characteristics with the source of each, sketch and photos, permit history, comparables with adjustments shown, the model derivation for the year under appeal, and prior appeal outcomes. Offices that assemble packets by hand lose most of appeal season to clerical work rather than appraisal judgement, and the packets they produce will not survive a competent agent's scrutiny because a summary written afterwards never quite matches what was actually done.

They track the appeal as a case with deadlines, evidence exchange, hearing outcome and the resulting value change written back with a reason code. That is what lets the office answer the questions that improve next year's roll: which classes and neighbourhoods generate appeals, which arguments succeed, and which agents file at volume.

They use automation for prioritisation and not for the record. Comparing aerial or street level imagery across years to flag likely change is a genuine efficiency gain for an office that cannot inspect everything on cycle. Changing an assessed characteristic because a model saw a roof shape is not defensible at a hearing, and the office will be asked, so imagery findings route to a human inspection rather than into the record.

They plan backwards from the roll certification date with real contingency, because statutory calendars do not move for technical delays and a slipped conversion in an assessment office is a public event.

And they settle ownership before award. The office should hold the repository, the infrastructure accounts and the unrestricted right to engage another firm. At Digital Heroes the client owns the code from the first commit. For a public body this is basic stewardship: the assessment roll is a public function, and the software producing it should not depend on any single vendor's continued goodwill or pricing.