Sewer Inspection Software Problems: The 5 That Cost Real Money, and How to Avoid Them

Why does trying to replace the inspection software sink so many of these projects?

The pitch usually starts as one system for everything. Within a month the scope has absorbed in truck capture, live coding, camera control and video encoding, and the project is now competing with WinCan, ITpipes, POSM and CUES Granite XP on the one thing they are genuinely good at. Those products were built for a crew in a van with a reel and a laptop, they handle the hardware, and rebuilding them returns nothing.

The reason this scope creep is so common here is that utilities experience the pain as one continuous mess. The contractor delivers in an inspection format, the engineer cannot use it, so the obvious conclusion is that the inspection software is the problem. It is not. The problem is that nothing owns the pipeline from a raw deliverable, through validation against your asset register, through consistent scoring, to a rehabilitation decision you can defend in front of a council.

Draw the boundary explicitly and write it into the statement of work. Inspection software is for the truck. Your platform is the utility's condition assessment system: it receives deliverables, validates them, stores raw observations, computes scoring, links video, ranks work and reports. Geographic information system stays authoritative for network geometry and the asset register. Your work order platform keeps managing crews. Everything else is yours. That single page prevents the most expensive version of this project, which is a multi year municipal software replacement nobody asked for.

What goes wrong when inspections have to be matched to your asset register?

This is where the schedule actually goes. The contractor inspects what they could access. Your register describes what you believe you own. They disagree constantly: a segment split when a manhole was added in 1996, a run inspected upstream to downstream with the nodes recorded the other way around, a lateral coded as a mainline, an identifier typed with a leading zero this season and without one last season, and pipe in the field that appears nowhere in the register.

Historic data is worse, because a decade of inspections came from three contractors on three packages, each using its own naming convention. Teams budget for parsing those formats, which is real but tractable work, and do not budget for the fact that a meaningful share of historical inspections cannot be attached to any current segment by any method.

The fix has two halves. Validate at intake rather than at analysis: parse the deliverable, match every inspection to a segment by identifier, by geometry and by upstream and downstream node pairing, and route failures to an exception queue with the specific reason attached. Flag length discrepancies beyond a tolerance automatically, because a run inspected at 340 feet against a register length of 290 is either a bad record or the wrong pipe and both matter. Send that exception list to the contractor while the crew is still in town.

The second half is accepting that some historical inspections stay unmatched permanently. Design for that. Unmatched inspections should be visible, counted and reportable rather than silently dropped, because a scoring model that quietly excludes a third of your history is producing a ranking nobody can defend.

Why do the integrations that matter here break after launch?

Four integrations carry this system and each fails differently. The geographic information system is the one that breaks most quietly. Segments get split, merged, renumbered or corrected during ordinary asset work, and a platform that took a one time copy of the register starts producing matches against geometry that no longer exists. Six months later the exception queue fills up and nobody can explain why.

Inspection exchange formats are the second. Vendors revise their export schemas at version upgrades, and a contractor who updated their software between campaigns will deliver something your parser was not written for. The parser has to be yours, with a version detection step and a clear failure that names the format rather than importing partial data.

Video storage is the third and it fails on cost rather than on function. Object storage bills accumulate quietly, and a retention policy that was never actually decided becomes an invoice conversation two years in. Decide it in design: full video for a defined period, keyframes at every observation retained permanently. That decision changes storage cost by an order of magnitude and preserves the evidence you reuse.

The work order platform is the fourth. Pushing rehabilitation packages into a system such as Cityworks is straightforward until somebody closes a job there and your platform never learns. Build the return path or accept that your completion reporting is manual.

The fix across all four is the same discipline: a scheduled reconciliation with the source of truth and an exception queue somebody owns, rather than a one time import and hope.

What happens when consent decree reporting and evidence retention are not covered?

Under a Clean Water Act consent decree your inspection data stops being an engineering resource and becomes evidence. The decree specifies inspection mileage, defect response times and rehabilitation commitments, with deadlines and stipulated penalties attached. Teams scope the analytics and treat reporting as an export they will add later, which is precisely backwards.

What goes wrong is not usually a missed commitment. It is that nobody knows the position between reporting cycles. An engineer assembles the report from several systems, checks it twice because the consequences of an error are real, and burns weeks per cycle. If the numbers come out short, there is no time left to do anything about it, because the shortfall was invisible while it was accumulating.

The second gap is response clocks. A defect above a defined severity starts an obligation, and if that clock lives in an engineer's head it is not a clock. It is an intention.

Build commitments as tracked objects with progress computed continuously from the inspection and construction records, so the position is known on any given day. Start response clocks automatically when a qualifying defect is accepted, and escalate them. Generate the periodic report as a document with the underlying records one click behind each number. And design retention deliberately, since video and observations under a decree may be needed years after any vendor relationship ends. Confirm your specific obligations with counsel rather than with a software specification.

Should you build custom or configure what you already own?

Configure, if you run one inspection crew on one software package across a system of modest size. WinCan or ITpipes already holds your data in a consistent format, your operators code consistently because there is one of them, and a custom platform would solve a problem you do not have. Spend the money on inspection mileage instead. Autodesk Info360 Asset is a credible answer if your gap is risk modelling rather than deliverable chaos, and if your inspections already arrive clean and matched it will do real work without a build.

We also never recommend rebuilding capture and coding software. Those products represent years of hardware integration and field hardening, and you will not improve on them.

Build when two or more of these are true. Inspection is contracted to multiple vendors delivering in different formats. Manual reconciliation of deliverables consumes something close to a staff position. You are under a consent decree or expect to be. Your rehabilitation programme is a substantial recurring capital line and prioritisation is currently argued rather than computed. Or you hold years of historical inspections in incompatible formats that cannot be analysed together, which is the trigger we see most often.

How do hidden costs get into the quote?

Asset register cleanup is the big one and it is nearly always outside the quoted scope. If a meaningful share of your segments cannot be matched confidently, somebody has to resolve them, and that work is neither fast nor automatable. Measure it before signing: hand a developer last season's deliverable exactly as the contractor sent it and ask how many inspections cannot be matched to your register. That number is the honest project brief.

Format count is the second. Each inspection exchange format is a parser with its own quirks, so a quote covering deliverable ingestion should name the specific formats. The fourth one is not free.

Video infrastructure is the third and it is an operating cost rather than a build cost, which is why it hides. Ask for a monthly figure at your actual annual inspection volume under the retention policy being proposed, not a figure for the pilot basin.

Then the ones that surface later. Manhole and lateral inspections, which carry their own coding schemes and are frequently assumed to be included. Hydraulic model integration if you want inflow and infiltration analysis connected. And the parallel period where engineers use both the old spreadsheets and the new ranking, which is short but real.

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

The builds that work store raw observations rather than only rolled up segment grades. This one decision determines whether you can ever change your scoring logic without re-inspecting, and a platform that keeps only grades locks in whatever interpretation you had in year one for the life of the archive. Ask the question directly and treat the answer as disqualifying if it is wrong.

They also take coding consistency seriously. Certification means an operator was trained, not that two operators code the same pipe identically. Observation rate per hundred feet by pipe material and vintage, compared across operators and contractors, finds systematic under coding and over coding quickly. Without that check, your ranking quietly sorts pipes by which contractor inspected them, and almost every multi vendor system has this without knowing.

They separate condition from consequence and opportunity. A grade five defect in a small diameter line behind a garage is not the same problem as a grade four in a large interceptor under a hospital access road, and a paving moratorium can decide the year even when it does not decide the pipe. Weight those inputs separately, version the ranking, and keep every version so a segment moving down the list has a traceable reason.

Finally, settle ownership of the code, the cloud infrastructure and the video storage accounts in writing before kickoff. At Digital Heroes the client owns the code from the first commit. Under a decree that archive is legally significant evidence, and it has to outlive any vendor relationship, including ours.