# How to model the return on collections software > How to model the return on collections software: what a point of liquidation is worth, why placement share and avoided claims belong behind a zero default, and what the model does not capture. - Source: https://ottoforfirms.com/roi/ - Publisher: Otto — https://ottoforfirms.com - Updated: 2026-07-27 Last updated 26 July 2026. Every vendor in this category will show you a payback figure. Almost all of them are built the same way: three or four coefficients baked into the page, a slider or two on top, and an output that was decided before you arrived. This page is the model itself, written out, so you can check the arithmetic and disagree with the parts you should disagree with. The interactive version — where every coefficient is a field you can change — is at [the ROI calculator](/#/roi). Nothing here is a measured result. Otto has published no customer outcomes, and the numbers a model like this produces are arithmetic on assumptions you supply. ## The one number the whole thing turns on A collections firm's revenue is three multiplications: **face value placed × liquidation rate × contingency fee.** That is the entire chain. A firm working $250M of face at 9% liquidation on a 25% contingency earns $5.6M. Every operational improvement worth modelling has to express itself somewhere in those three terms, and for software the honest place is the middle one. So the unit is **a percentage point of liquidation**. One point on a $250M book at a 25% fee is $625,000 a year. That single figure is usually larger than the whole software cost, which is exactly why hiding it inside a hardcoded constant is such an effective way to make a calculator say whatever you want. If a vendor's model assumes a lift and does not show you the assumption, the assumption **is** the result. Whether software moves liquidation at all is a real question, and the answer is not the same at every firm. Cleared dials that would otherwise have been blocked, promises captured at the moment they are made rather than at end of shift, and a shorter gap between placement and first contact are the mechanisms. How much they are worth at your firm depends on how much of that you are already doing. ## The lines that should default to zero Two of the value lines in most vendor models describe things the buyer does not control. Both default to **zero** in ours, and turning them on should be a deliberate act. **Placement share.** Banks and debt buyers move volume toward the firms that score best on their own scorecards, and a firm that can evidence its controls tends to score better. But share is granted by the client, on the client's timetable, and it may never move at all. A model that assumes several points of additional placement is assuming someone else's decision. **Consumer claims avoided.** Controls change what is possible inside your shop. They do not change what a plaintiff's firm chooses to file. Avoided-claim value is also the line most often over-counted, because the FDCPA shifts fees — under 15 U.S.C. § 1692k(a)(3) a successful consumer recovers costs and a reasonable attorney's fee — so the cost of a claim is your own defence plus the other side's fees, and it is tempting to count the whole of that against software. Include both if you believe them. Include them knowingly. ## The lines a firm controls **Compliance hours.** Time spent assembling evidence that already exists somewhere in the system. The honest input is not "hours spent on compliance" but the share of those hours that is *retrieval* rather than *judgement* — the retrieval is what software removes. **Roles not backfilled.** Data and operations work that stops being necessary. The defensible version of this line is a fraction of a role, not a headcount plan, and it double-counts with compliance hours if the same person appears in both. ## Payback is the only line that is not a multiplication Value does not arrive on day one. Migration, configuration and the point at which people actually change how they work take months, while the fee runs from the first invoice. A payback figure that ignores the ramp is a payback figure that is too short. The model accrues value against a ramp you set, with the fee running from month one, and compares the two. Two outcomes are worth naming: - If annual value exceeds annual cost, there is a month where cumulative value overtakes cumulative fee. That month is the payback. - If it does not, **there is no payback month at all** — both flows recur, so the gap never closes. The right output there is to say so, not to round to the end of the year. A model that always produces a payback month is not a model. ## What this does not capture - **It is a model, not a measurement.** No line in it is a customer result. - **The enforcement engine ships four configured jurisdictions** — Florida, Georgia, Texas and New York. If your book is mostly elsewhere, the compliance lines are worth less than the model suggests until more are configured. The published [statute of limitations reference](/statute-of-limitations/) covers fifty-one jurisdictions; that is a dataset we publish, not the rule set inside the product, and the two should not be read as the same thing. - **Your own cutover cost is not in it.** Staff time during migration, parallel running, and the work of deciding what your rules actually are — all real, none modelled. - **No discounting.** Money next year is treated as money now. - **Double counting is possible.** Compliance hours and roles not backfilled can describe the same person. - **Controls are not a defence.** Enforcing a rule is not an assurance of compliance, an opinion on your obligations, or protection from a claim. Otto's terms disclaim compliance outcomes and this page holds the same line. ## Using it Open [the calculator](/#/roi), replace every assumption with one you would defend to your own partners, and read the result as what it is: arithmetic on your inputs. If the output only works with the optimistic lines switched on, that is itself the finding. --- Every figure this page describes is a projection from assumptions you supply, not a measured result and not a customer outcome. Otto has published no customer results. Change the assumptions to ones you would defend to your own partners, and treat the output as arithmetic on your inputs rather than as a forecast.