The supplier's offer was easy to like when presented as a unit rate. Ten million image-recognition credits cost ₹1.8 crore upfront, while usage-based charging was ₹2.10 per image. Dividing the package price by its full volume produced ₹1.80 per credit. The apparent discount was real if the company used the quantity it had bought under comparable terms. Our reviewed annual requirement, however, was six million images. At that volume the usage-based illustration was ₹1.26 crore, below the nonrefundable minimum by ₹54 lakh before tax, integration and other differences.
The proposal therefore asked us to buy more than a lower rate. It asked us to commit to a quantity and payment timing. Unused credits would not be refunded. The commercial decision depended on whether the additional volume was credible and useful, not simply whether the supplier had offered a lower price for each unit in the larger package.
Price the quantity the company is likely to use
Procurement colleagues were understandably interested in the discount. It was visible, negotiated and easy to compare with the ordinary rate. They also believed new clients might use the balance. I wanted that possibility separated from the requirement supported by the current plan. A growth ambition can justify examining a larger commitment, but it should remain a forecast with assumptions rather than become established consumption because a supplier's package makes the higher volume attractive.
At six million images, the package's effective cost per used image would differ from the headline rate because four million credits remained unused in the illustration. That does not make the supplier's pricing deceptive. It makes the buyer responsible for understanding the commitment. The package and usage option answer different commercial needs. One can be suitable for predictable scale while the other is more expensive per unit and more appropriate for uncertain demand.
I asked the team to identify the basis of the six-million requirement and the circumstances that might move it. Which client work was contracted? Which volumes were estimates? What technical or timing conditions would affect consumption? We did not need to pretend the forecast was exact. We needed enough evidence to decide whether the company should prepay the higher quantity, with the risk of unused credits explicit rather than hidden inside a favourable rate comparison.
The payment timing also mattered. A minimum paid upfront uses cash before the work consumes it, while the actual terms of the usage option determine a different payment pattern. The simple ₹54 lakh illustration did not include every financing or contractual effect. I wanted the comparison to state that limitation instead of allowing a useful first calculation to become the entire investment account. Finance can make a decision clearer without claiming that one arithmetic line has captured all its consequences.
Locate the point where the other choice becomes better
At ₹2.10 per image, the simple price crossover against ₹1.8 crore is approximately 8.57 million images. That calculation assumes the relevant service and other terms are comparable. It gives the team a useful threshold to examine, not a prediction that demand will reach it. Above that quantity, the package could become cheaper on that narrow price basis; below it, usage charging could remain lower. I wanted the recommendation to acknowledge the conditions under which it would favour the alternative.
That approach made the disagreement more productive. Procurement no longer had to defend the discount as inherently good, and delivery no longer had to argue that flexibility was valuable without saying when. The discussion could focus on the credibility of the volume, the cost of being wrong and the terms attached to each option. If new evidence later supported a different level of demand, revisiting the decision would be sensible rather than an admission that the earlier calculation had failed.
We selected usage-based charging for the reviewed period. The choice gave up the possibility of the lower effective rate at sufficiently high consumption in exchange for avoiding the immediate minimum. I did not claim the company had already saved ₹54 lakh for the year. That was a difference in an illustration at six million images. Actual charges would depend on actual use and the agreed terms. Calling the model a realised saving would remove the uncertainty that had been central to the decision.
I also did not present flexibility as free. A higher unit rate can become expensive if demand exceeds the supported forecast. Management should understand that consequence before approving the choice. The recommendation was that the available evidence did not justify the prepaid quantity then, not that minimum commitments were always poor procurement. A responsible account can explain why a decision was reasonable while preserving the scenario in which another option would have cost less.
Keep technical quality outside a price-only conclusion
The financial comparison did not determine whether the recognition service was technically suitable. Quality, review needs, integration, rights and operating conditions required their respective owners. Recognised text is also only one part of a wider digital-preservation workflow. A lower processing bill does not establish that a collection remains accessible or that the resulting record meets the client's requirements. Those distinctions belong beside the cost account, not disappear because finance has found an attractive number.
We therefore kept technical assessment separate from the claim about price. That separation did not mean the two teams worked without each other. If one option required substantially more review or integration, the financial comparison would need that evidence. The point was to avoid inferring quality from cost or treating the supplier's billing unit as a complete measure of useful customer work. A processed image and an accepted outcome may involve different effort.
For directors, the useful question is what the company is committing to buy and why the expected use supports it. Does the paper show the minimum, the credible volume, the crossover and the consequences of lower or higher demand? Which assumptions come from signed work and which from hoped-for growth? A board can examine those points without negotiating every supplier rate. It can also ask whether management reports modelled differences with the same restraint it uses when describing actual results.
The experience reinforced a habit I use in other financial choices: name the commitment before celebrating the percentage. A discount, an advance receipt or an early settlement can be valuable, but each has a quantity, period and set of conditions. The calculation becomes useful when those are explicit. The company should choose the arrangement its evidence supports, and keep enough of the alternative in view to know when the next decision ought to change.