Build vs Buy Interconnect Settlement Software: You Are Buying Evidence, Not Invoices

Where staying put is the correct commercial answer

A small operator running a handful of interconnect partners on a single softswitch, whose bundled wholesale billing rates those partners acceptably, does not have a settlement problem. It has a settlement chore. Reconciling a dozen invoices in a workbook each month is unpleasant and it is also cheaper than any alternative, and the money belongs in traffic acquisition rather than in a platform.

For a large operator, the packaged products are credible. TEOCO and Subex both bring mature interconnect and settlement capability with deep control libraries and long track records in this exact discipline. Comarch and Hansen approach it from full business support estates, which is an advantage if you are buying the whole stack and a disadvantage if you want settlement alone without touching billing. None of these are weak products, and dismissing them would be dishonest.

Keep buying the layers underneath as well. Mediation, switch platforms, number portability lookups and reference data feeds are commodity components with vendors who maintain them, and there is no version of this project where rebuilding mediation makes sense.

The condition that keeps buying right is whether your team works inside the system. If your wholesale manager opens the settlement tool and can see every partner position, you are fine. The moment a parallel spreadsheet appears alongside the platform, you own two versions of the truth and are paying for one of them, and that spreadsheet will become the version people trust.

The point where settlement stops being billing

Retail billing has an authority: you are it. Interconnect does not. Your counterparty has their own switches, their own mediation and their own rating, and both of you are correct about your own data. Settlement is the process of agreeing a figure between two parties who measured the same traffic with different instruments, under a contract that specifies how disagreements are resolved and by when. What you are building is evidence infrastructure with an invoicing function attached, and that framing changes what the system must hold.

Variance is normal and has ordinary causes worth enumerating before anyone accuses anyone. Billing increments differ, so a partner rating in sixty second increments and you rating per second produce different billed minutes from identical calls. Rounding conventions differ at call level and again at aggregate level. Calls answered by an announcement get treated differently. Short duration grace periods apply in one direction and not the other. Day boundaries depend on whose clock. Records go missing at a mediation feed on either side.

Build when the volume of agreements makes that unmanageable by hand. Past roughly fifteen active bilateral relationships, and earlier if one partner carries a large share of your traffic, the reconciliation work stops fitting into anyone's month. Build when rate sheets arrive as spreadsheets with thousands of rows of additions, changes and deletions effective in a week, because loading one incorrectly means rating a week of traffic wrong and finding out at settlement. Build when your netting position lands four weeks after close, because routing and credit decisions are then made on stale margin and the loss never appears as a line item anywhere.

Costs on both sides, including the ones that grow

Packaged licensing in this category typically scales with record volume, partner count or both, which means the commercial success you are working toward increases the bill. That is defensible pricing and it is also worth modelling three years out rather than one. The friction that matters more is onboarding: implementation is a project before the first reconciliation runs, and partner file format additions frequently route through the vendor's professional services at day rates and on their queue.

A build is capital plus maintenance. In Digital Heroes delivery experience, a first release covering the rate engine with effective dating, call record ingestion, partner invoice ingestion for your largest counterparties, bilateral reconciliation with variance decomposition and dispute cases runs $90,000 to $200,000 and ships in 14 to 20 weeks. A full platform adding netting statements, multi currency, access jurisdiction determination, roaming settlement, a routing margin feed and fraud alarms runs $250,000 to $600,000 phased over 9 to 15 months. Running costs land near 15 to 20 percent of build a year.

Price moves on specific things. Record volume, because rating a few million calls a day and a few hundred million are different engineering problems. The number of settlement models you carry, since voice interconnect, access and roaming are effectively three products sharing infrastructure. Jurisdiction determination for United States access, which brings reference data licensing and its own update cadence. Currency handling, which sounds minor until agreements specify different rate dates. And the number of distinct partner formats you must ingest on day one, which is the item most often underestimated when scoping.

The costs that appear after the first quarter

Partner format churn is permanent. Every counterparty invents their own invoice layout and changes it without notice, and a settlement system that cannot absorb a new format inside days will be worked around. This is the single strongest argument for a build over a vendor queue, and it is also a standing obligation you take on: someone in your own team has to own format mapping as a routine operational task rather than a project.

Storage and re-rating capacity is the second. Disputes run for quarters, so you must be able to re-rate a period at the rules in force when the traffic was carried, which means keeping call records and full rate history rather than current state. That is a deliberate retention decision with a cost attached, and it is cheaper than losing a dispute.

Reference data is the third. If you settle access in the United States, jurisdiction determination depends on licensed industry number data with its own subscription and refresh cycle, and stale reference data produces rating that is wrong in a way nobody notices until a counterparty raises it.

Fourth, currency and rate date conventions. Agreements specify which day's exchange rate applies, and they do not agree with each other. A single global setting will produce netting statements your partners dispute for reasons that have nothing to do with minutes, which is a frustrating way to lose credibility in a negotiation.

The test that funds the project

Take the last month a partner disputed you. Time how long it took to produce a defensible statement of exactly where the two views diverged, broken down by destination group and by day. If it took a week and ended in a compromise, price a month of engineering against the disputed amount. That comparison is the whole business case, and for most wholesale operators it resolves the argument immediately.

Then check the clock. Look at your last four disputes and mark which ones were resolved on the merits and which were conceded because the contractual window closed before evidence could be assembled. Money lost to a deadline rather than to an argument is pure recoverable revenue, and it is usually the line that gets a project approved.

Third, date your netting position. Work out how many days after period close you knew the net figure per partner. If the answer is beyond a week, your routing and credit decisions are being taken on margin information that is already historical, and knowing early that a partner's inbound volume collapsed while your outbound cost held steady is the difference between reacting and filing.

Fourth, count formats. List every distinct invoice and rate sheet layout you receive today. That count, not your partner count, is the real scoping input, and operators are consistently surprised by it.

How to sequence it and how to vet the builder

Build the rate engine first, with effective dating and full history, because everything above it depends on being able to re-rate any past period at the rules that applied. Then call record ingestion, then partner invoice ingestion for your largest counterparties, then reconciliation with variance decomposition and dispute cases. Netting, roaming and access jurisdiction belong in phase two, and attempting all three settlement models at once is the most common way these projects overrun.

Ask any candidate to explain how they will re-rate a call from two years ago at the rates in force on the day it was made. If the answer involves updating a current rate table, end the conversation, because retrofitting effective dated rating means rebuilding everything above it. Ask who onboards a new partner invoice format and how long it takes, and insist the answer puts your own operations team in control through a mapping interface.

Ask about variance classification rather than variance detection. Anyone can show that two numbers differ. The valuable behaviour is stating that the gap sits in one destination group across three days and is consistent with an increment convention mismatch, which turns a four week argument into a phone call.

Digital Heroes builds this category PRD first, with the rating and reconciliation rules agreed in writing before code. Indian, United States and United Kingdom entities are all available, so an operator can contract in the jurisdiction its interconnect agreements are already governed by. Fifty plus people, 2,000 plus delivered projects, and a channel of 2.5 million YouTube subscribers where the engineering is discussed in the open. Bring two artefacts to a first conversation: one partner agreement with its rate sheet, and one month where your number and theirs disagreed.

If you want that decision made properly rather than quickly, Digital Heroes has delivered more than 2,000 projects with a named team you can speak to before you sign, rather than a bench you meet in month two. You keep the specification either way.