Spectrum License Management Software Problems: The 5 That Cost Real Money, and How to Avoid Them

Why does trying to build a coverage planning tool sink these projects?

The obligation is a coverage number, so the first specification tends to include coverage prediction. Within a month somebody is asking about terrain databases, clutter classification and diffraction models, and a project that should have shipped in eight to twelve weeks has become an attempt to rebuild propagation engineering.

This is specific to spectrum because the person feeling the pain is usually in regulatory or legal, not in radio engineering, and from that seat the coverage number and the obligation look like one problem. They are two. LS telcom and ATDI represent decades of physics and validation work, and there is no version of this where rebuilding that pays.

Write the boundary before quoting. The planning environment predicts coverage. Your platform holds the licence, the authorisation area as real geography, the obligation with its rule, threshold and deadline, the evidence record with its parameters, and the owner. Coverage arrives as an import, not as a calculation you perform.

The same discipline applies at the other end. Comsearch and the other coordinators do coordination properly and there is no reason to take that in house, and a spectrum access system administrator such as Federated Wireless occupies a regulated role you cannot replicate. Your platform consumes what those parties produce and holds your side of the file. Say that in the statement of work and the scope stops drifting.

What goes wrong when a holdings spreadsheet and acquired filing history are migrated?

The spreadsheet imports in an afternoon and is the smaller half of the problem. It carries a column called notes containing conditions, exceptions and half remembered obligations in prose, and prose cannot be tracked. Somebody has to read every row and decide what is a deadline, what is a condition on an asset and what is a comment.

Acquired portfolios are the real cost. A company that bought a regional operator inherits call signs, coordination correspondence, conditions accepted during coordination and the specific orders that granted the licences, and that material frequently exists as boxes and as inboxes belonging to people who have left. There is no import for it. Somebody reads it.

The second failure is subtler. Entity names change through acquisitions, so a licence granted to a company that was later merged sits in the regulator's database under a name nobody internally uses. Migrating your spreadsheet without reconciling entity names produces a register that looks complete and omits exactly the holdings most at risk of being forgotten.

Sequence it to expose the gap early. Export your current holdings sheet and pull the regulator's record for every entity name you have ever operated under on the same day, then compare. That takes an afternoon and produces both your business case and your migration scope. Start with the services carrying buildout or performance risk and leave stable site licences on the existing process for a phase. Budget the historical file reading as a named line with a named reader rather than hiding it inside development.

Why do regulator data feeds and network inventory integrations break after launch?

The regulator's public data is authoritative and it is not tidy. Entity naming varies, call sign history includes assignments and administrative changes, and pending applications appear and disappear as they are processed. A reconciliation built against clean assumptions produces a wave of false exceptions in month two, the team stops reading the exception report, and the feature is dead while still technically running.

Network inventory breaks the other way. Linking licences to the sites and equipment actually operating under them is what surfaces a decommissioned site, and that link depends on identifiers maintained by the network team for their own purposes. Sites get renamed during consolidation, a new inventory system arrives, and the join quietly stops matching. Nothing errors, because a licence with no linked sites looks the same as a licence whose sites are unmatched.

Fix both with the same approach. Tune the reconciliation to your entity name list, including historic names, and treat a persistently noisy exception class as a defect to fix rather than a report to ignore. Report unmatched licences and unmatched sites as their own metric, so the join quality is visible rather than assumed. And where a spectrum access system governs assignments in shared spectrum, consume its records so your view of what you are authorised to transmit matches what is actually being granted rather than what you believe you configured.

What happens when coordination conditions and discontinuance are not covered?

Coordination is how interference protection is established, and the value of the process is entirely in the record it produces. Terrestrial microwave coordination follows the recognised industry bulletin process, and if a neighbouring licensee later claims harmful interference, your defence is the notice you sent and their lack of objection within the window. When that correspondence lives in an engineer's inbox, it leaves when the engineer does.

The failure that costs most is not the missing file. It is the condition inside it. An azimuth restriction or a power limit accepted during coordination is a constraint on that path forever, and a future engineer with no visibility of it will change an antenna and quietly break an agreement made years earlier. That is not a document management problem. It is a constraint that has to be enforceable at the point of change.

Discontinuance is the second uncovered risk and it is the quietest one in this category. Permanent discontinuance of operation can terminate an authorisation, and a site removed during a consolidation and never rebuilt puts a licence at risk without anyone filing anything or noticing anything. The spreadsheet has no view of operational status because it was never joined to the network.

Attach correspondence to the asset rather than to a person, with response windows tracked as obligations carrying a clock rather than as emails awaiting somebody's memory. Record accepted conditions as constraints on the path or site so a change request meets them. And link licences to operating sites so a decommission raises a flag against every authorisation depending on it.

Should you build custom or configure what you already own?

Keep the calendar and the shared folder if you hold a dozen site authorisations with no performance conditions, no leases and one regulator. That is a genuinely defensible answer at that size and we would rather you spent the money on the network. Adding a shared reminder for each renewal and a single owner for the folder closes most of the practical risk.

Keep buying where the product is the specialism. LS telcom and ATDI are your planning environment and should stay so. Comsearch runs coordination and there is no case for replicating it. Federated Wireless and the other spectrum access system administrators occupy a regulated role. None of that is a gap in your operation.

Build when two or more hold. You carry performance or buildout obligations that require evidence rather than a filing. Your holdings span more than one service type or more than one acquired entity. Coordination correspondence is your interference defence and it lives in individual inboxes. You lease spectrum in or out and carry obligations sitting in agreements rather than on the licence. Or one person is the only reliable index of what you hold and when it is due. The tipping point is that planning tools model propagation and coordinators manage notices, while the obligation register that sits above both is made of your holdings, your acquisitions and your specific grants, so nobody sells it.

How do hidden costs get into the quote?

Historical file reading is the largest and it is almost never in a development quote, because it is not development. Someone has to read boxes and inboxes from acquired markets and turn prose into obligations, conditions and evidence records. Price it as a named line with a named person and an estimated volume, or it will be discovered as a delay rather than a cost.

Service type count is the second. A broadcast authorisation, a mobile market licence, a site licence and a microwave path are four different obligation shapes with four different evidence requirements, and a quote written around one of them will be reopened when the second arrives.

Coverage evidence is the third and it is the heaviest engineering component. Importing a geospatial export with its model parameters and computing a covered population against a licensed area is real work, and a quote that prices it as attaching a map has priced a filing cabinet.

Then the quieter ones. Multiple regulators if you operate across borders, since each publishes data differently and some publish very little. Ongoing reconciliation tuning, because entity naming edge cases keep arriving. And the internal time to agree who owns each obligation, which sounds administrative and is the decision the whole system depends on.

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

Ask a prospective developer to model an obligation on a whiteboard. The right picture has a licence, an authorisation area as real geography, an obligation with a rule, a threshold and a deadline, an evidence record carrying its parameters, and a named owner. Anyone who draws a task list with due dates has built a project tracker, and it will fail at the first coverage showing.

Ask how coverage results come in from your planning environment. If the answer involves a screenshot or a PDF, the evidence will not be reproducible at the only moment it matters. The workable answer handles a geospatial export with the propagation model, terrain and clutter settings recorded as data alongside the result and a timestamp.

Ask whether they have worked with public regulatory data feeds. Reconciling against a licensing database sounds trivial and is not, because entity naming, call sign history and pending applications all have edge cases that only appear at volume. A developer who has done it will ask about your acquired entity names before quoting rather than after.

Settle ownership of the code, the repository and the hosting accounts in writing before kickoff. At Digital Heroes it is yours from the first commit. This is a small, fast build relative to the asset it protects, which is unusual in telecommunications and worth saying plainly, and the register it produces should outlive every vendor relationship including ours.