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

Why does building a better form instead of a consistency layer happen so often?

Because the form is the visible part of the failure and the consistency layer is not.

After a difficult activation, the complaints an emergency manager hears are about collection: the paper got wet, the photographs were on phones, half the addresses were typed wrong, one team lost their sheets. Those are all true and all fixable with a mobile form, so the next procurement asks for a mobile form. The team gets one, collection improves, and the submission still comes back, because the reason it came back was never the paper.

The reason is that the app asks for a conclusion. It offers destroyed, major, minor and affected, and a team lead picks one. A lead who has seen worse events is generous. A lead who is a building inspector and knows what a repair costs is conservative. Neither is wrong, and the aggregate is unusable, because a reviewer comparing your counts to your photographs can see the disagreement immediately.

The fix is to stop asking for the category and start deriving it. The application asks about observable conditions: roof structure condition, wall integrity, water depth inside the structure, foundation displacement, whether utilities are cut, whether the structure is accessible. The category comes out of rules you control, applied identically to every team. Write the acceptance test into the scope: two assessors with different experience walk the same five structures independently and the system returns the same categories. If a proposal cannot describe how that happens, you are buying a form and your rework will not move.

The second benefit matters more than people expect. Because the category is derived rather than typed, you can recalibrate mid event. If the state signals your major counts are reading high, you adjust the rule and reprocess, instead of sending fourteen teams back into the field.

What goes wrong when parcel and assessor data are joined?

The parcel join is the highest value half day in the whole project and it is where the schedule usually slips, because assessor exports are idiosyncratic and nobody looks at them until the build has started.

What you want is every structure in the impact area pre loaded with an address, assessed value, year built, construction type and occupancy, so teams confirm rather than type. What you actually get from an assessor is a file keyed on a parcel identifier, where a parcel may hold several structures, where a mobile home park is a single parcel with forty households, where address formatting differs from your address point layer, and where a meaningful slice of impacted structures have no parcel record at all because they are new, informal or on tribal or federal land.

Three failures follow from ignoring that:

• Unmatched structures are dropped from the pre load, so teams type them in the field under pressure and they arrive with inconsistent addresses that create duplicates at aggregation.
• One parcel equals one structure is assumed, and a mobile home park counts as a single damaged property. That single assumption can materially understate household impact, which is the part of the picture that drives individual assistance.
• Assessed value is treated as loss value with no documented relationship between the two, so the valuation method varies by whoever built the spreadsheet.

The fix is to pull a real assessor extract before the build starts, not a sample, and profile it. Decide explicitly how multi structure parcels are handled, how unmatched structures enter the system as first class records rather than exceptions, and what the documented relationship is between assessed value and estimated loss. Then run the join against a past event and count what fails, because the ten percent that does not match is exactly where a mobile home park disappears from a count.

Why do offline sync and photo linkage break after the first real activation?

They break because they were tested in a car park with intermittent signal rather than in a county with no towers for a full day.

Offline in this domain means the device holds the assigned grid, the parcel layer, the address points and the imagery basemap before deployment, and captures everything locally including full resolution photographs for a whole shift. Builds that treat offline as caching the last screen fail on the first day, because the basemap was never downloaded and a team without imagery cannot confirm which structure they are standing in front of.

Photographs are the specific trap. If the application stores photographs in the device gallery and matches them to structures later by timestamp, the evidence quietly disconnects from the record. Phones get replaced mid event, clocks drift, a team member photographs a structure and the assessment is entered by their partner on a different device, and a week later nobody can prove which house is in which picture. Photographs have to carry structure identity at the moment of capture.

Sync then has to be conflict safe. Two teams will assess the same corner property, because grid boundaries are imperfect and everyone is tired. A system that keeps the last write silently discards one assessment and no reviewer will ever know. The correct behaviour is to detect the collision at sync and route both records to a reviewer with the photographs attached.

Test all three deliberately before you accept the build. Put a device in airplane mode for eight hours with a hundred structures and a few hundred photographs, then sync. Have two devices assess the same structure and confirm a reviewer sees both.

What happens when insurance status and re-inspection are left out?

These two are almost always cut to make a delivery date, and they are the two that shape the argument you are making.

Uninsured loss is what drives the individual assistance case. A hundred damaged homes with insurance and a hundred without are completely different submissions, and if the form has no field for it, the distinction is gone and cannot be recovered without going back out. Teams will not always get a clean answer at the door, so the design has to accept unknown as a real value that can be resolved in a follow up pass, rather than forcing a yes or no that a tired assessor will guess at. Flood coverage needs to be captured separately from general coverage, because they behave differently in the argument.

Re-inspection is the other gap. A property nobody could reach on day two because a road was closed has to come back into the queue, not fall out of it. Builds that model assessment as a single terminal event have no way to express not yet assessed, so those structures either vanish from the count or get recorded as no damage, which is worse.

Both fixes are small if they are designed in and awkward if they are not. Access status is a field with its own workflow. Insurance is captured as reported, with the source noted, and unknown is a legitimate state that appears on a follow up worklist. Neither is expensive. Retrofitting either after an activation means re-contacting properties you have already visited.

Should you build custom or configure what you already own?

For a large share of counties the honest answer is do not build, and we would say so before quoting.

If you activate once every few years with a small assessment cadre and no standing analyst, a custom system will be unfamiliar exactly when the clock is running. Crisis Track is purpose built for this, handles the federal side arithmetic well and reports in shapes reviewers recognise. If your practice fits its model, it is a strong and honest buy.

If you already have Esri Survey123 licensed, and most jurisdictions do, the cheapest real improvement available to you is not new software. It is a half day of calibration with your team leads before each deployment, plus rewriting your existing form to capture observable conditions instead of a category, and doing the category derivation in a spreadsheet afterwards. That is a configuration exercise you can complete in a week, and for a county that activates rarely it will outperform a custom build. Veoci is the reasonable answer if you want assessment configured alongside your other emergency management functions in one environment.

Build when you are a state agency trying to get consistent numbers out of many counties, when you are a large jurisdiction assessing more than once a year, when your submissions have been returned for category or valuation inconsistency, or when assessment data has to flow into recovery and cost tracking rather than being retyped.

How do hidden costs get into the quote?

Four items account for most of the overruns, and every one is knowable before kickoff.

State scope priced as county scope. Building for one county is one rule set. Building for a state means configurable categories per jurisdiction plus a reconciliation layer above them, and it is the largest single multiplier in this category. Decide which you are buying before anyone quotes.

Assessor data condition assumed. The usual schedule risk is not engineering, it is the extract. Profile a real file in week one, and if the parcel layer is stale or the address points disagree with it, that is a data project with its own budget.

Offline treated as a feature. Local storage of imagery, full resolution photograph handling and conflict safe sync are architecture decided at the start. Retrofitting them approaches a rewrite, and a quote that lists offline support as a bullet has not priced it.

Cost recovery handoff sold as an export. Assessment feeds recovery, and doing that properly means shared identifiers so the same structures, locations and photographs carry through. A comma separated file dropped into another system loses the links you need at audit.

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

Ask how they would derive a damage category. If the answer is a dropdown with four options, they have built a form and your consistency problem is untouched. The right answer captures observable conditions and derives the category through rules you can change and reprocess against mid event.

Ask what they have shipped that works with no connectivity for a full day, including full resolution photograph capture bound to structures, and ask specifically what happens when two teams assess the same property. Ask how they will handle the ten percent of impacted structures with no parcel record, because there always is such a set and dropping it is how a mobile home park disappears from a count.

Ask whether they have worked with assessor exports before, and have them profile yours before quoting rather than after.

Then settle ownership before kickoff, including the collected data and the photographs, in infrastructure you control. At Digital Heroes the jurisdiction owns the code and the data from the first commit. Assessment records feed declaration requests and later audits, so they cannot sit inside a subscription you might not renew. Start by pulling your last assessment, taking one grid, and checking whether you can still connect each photograph to the structure it documents. That answer usually decides the project on its own.