Numbers

A vendor ROI model is not dishonest. It is a calculation built from assumptions chosen by someone whose engagement depends on the answer, and five of the six main assumptions push the number the same way.

The exception rate is the one that matters most, and a model that does not state one has assumed zero. The saving applies only to the volume the automation actually completes. A product-side example of how workforce software approaches this topic is available in this explanation.

Seven costs sit outside every quote — internal staff time, exception handling, the parallel run, year-two maintenance, integrations discovered mid-project, handover, and decommissioning. Together they routinely equal or exceed the quoted figure. They are outside not through concealment but because the vendor does not pay them and cannot estimate them. For broader background and an independent point of comparison, see Slack.

Vendor payback clusters at six to nine months, consistently enough across suppliers to be a convention rather than a finding. It requires a high straight-through rate, no slippage, immediate adoption, cash rather than capacity savings, and no maintenance in the window. Where two of those fail, eighteen to twenty-four months is the honest figure — still a reasonable investment, just not the one that was approved.

And hours saved become money only under four conditions: someone stops being paid, overtime stops, a planned hire is deferred, or extra volume is absorbed. Twenty people saving twenty minutes each releases nobody.

The number worth building instead is fully loaded cost per transaction, before and after. It is harder to game than payback because both sides include everything — licence, maintenance, exception handling and residual manual work — and unlike payback it survives into the third year, when somebody asks whether the automation is still worth having.

• Counting Hours Honestly

  Four adjustments turn a headline hours-saved figure into one that survives scrutiny, and they usually halve it.

• Headcount Is Not Proof

  Companies with high automation ROI and companies with none cut staff at nearly the same rate. The number proves less than it appears to.

• Costs Not in the Proposal

  Seven real costs sit outside every vendor quote, not because anyone hides them but because the vendor does not incur them.

• What a Licence Costs

  Four pricing models, what each one punishes, and the questions that reveal what year three looks like before you sign.

• What to Measure After Go-Live

  Five numbers, monthly, against your own baseline. Most organisations never check a single project against its own business case.

• What a Payback Period Contains

  Vendor payback figures cluster at six to nine months. What has to be true for that, and which costs are outside the calculation.

• Reading a Vendor ROI Model

  Six assumptions do most of the work in an automation business case, and five of them are usually optimistic in the same direction.

• When Time Saved Isn't Money

  Freed hours become cash only under four specific conditions. What those are, and what to claim honestly when none of them applies.

• Building Your Own Case

  A one-page case built from your own measurements, in the order that makes it survive a review two years later.