Draft for Tamir's review. Not published.
Test the conditions against your operation
Take a month of real orders, routes or operating events and check which fall outside the guarantee's conditions. For uptime, match each recent interruption to covered or excluded. Exclusions can consume most of a headline commitment.
Ask each vendor to compute its promised saving on your actual data, with the calculation shown. A percentage with none of your data behind it is a brochure. Separate simulated or modeled results from performance observed under agreed conditions.
Own the baseline and the measurement
A savings share can be fair when the measurement is yours. Build the baseline from a full year of your own data. Normalize it for factors outside the vendor's control, such as energy prices, volume and season. Make your systems the measurement source, with a dispute process.
Check that two suppliers are not claiming the same benefit. Cap savings-based fees and set a conversion to a license after a period. Tie payment to capabilities you can use rather than installation, and final payment to an acceptance test on your own data.
Depending on your seat
If you're on the board, ask whether the results the contract rewards are observed or simulated. Ask who must prove the cause of a failure, and whether the remedies match the business consequences.
If you run operations or IT, check the responsibilities left between the vendor and your own teams, such as integration with your ERP or plant systems. Ask what changes in the guarantee when you add products or change how you operate. Pilot at one site and measure the result against your bill or your output.
What to check before you decide
- Run a month of your real orders, routes or events against the guarantee's conditions and mark what falls outside.
- Match each recent interruption or failure to covered or excluded.
- Ask for the promised saving computed on your own data, with the calculation.
- Set the baseline from a full year of your data, normalized for factors outside the vendor's control.
- Separate simulated or modeled results from performance observed in your operation.
- Check who measures, who must prove the cause of a failure, and what the remedy is.
- Tie final payment to an acceptance test on your own data, with exit terms if the guarantee is never met.
Questions people ask
Should a refinery AI guarantee count simulated results as earned performance?
Separate modelled potential from realised performance under agreed conditions. Approval depends on the baseline, attribution, exclusions, and commercial terms reviewed with appropriate advisers.
Two fleet charging vendors promise big savings from smart scheduling, how do I know which saving is real?
Ask each vendor to compute the saving on your actual routes, return times and tariff, and show the calculation; a promised percentage with no fleet data behind it is a brochure. Then prefer the vendor whose software works with chargers you could buy elsewhere. It depends on your tariff structure, your depot grid capacity and how predictable your vehicle returns are.
Route optimisation vendor wants a share of fuel savings with a baseline they define, how should the deal be structured?
Set the baseline yourself from a year of your data, normalise for fuel price, distance and load, and cap the fee. Define who measures and how disputes are settled. A savings share can be fair when the measurement is yours. It depends on how stable your routes are and on whether your telematics can measure what the contract needs.
Should we accept energy software savings claimed by more than one supplier?
Reconcile rights and value attribution across the agreements before signing. The decision depends on who can call the flexibility, which commitments conflict, and how benefits are measured.
Does this refinery software uptime guarantee cover the failures that hurt us?
Evaluate the guarantee against actual business failure scenarios. Its usefulness depends on covered dependencies, evidence requirements, remedies, and responsibilities reviewed with commercial and legal owners.
Should hospital software payments depend on installation or usable business outcomes?
Tie acceptance to defined capabilities the hospital can actually use. The terms depend on attributable deliverables, business prerequisites, and fair responsibility for dependencies reviewed by procurement and legal owners.
Warehouse robotics contract has a throughput guarantee with lots of conditions, what should I check before signing?
A throughput guarantee is worth what its conditions allow, so test them against your real order profile, item sizes and peak days. Make the acceptance test use your own data and tie the final payment to it. It depends on how far your order mix is from the vendor's reference and on how the ERP and warehouse systems will exchange orders.
How I can help with this decision
- Ask or talk (Free)
- I give my view on which conditions in the guarantee will bite your operation, and which baseline and acceptance test to insist on.
- Review (Pay if it was worth it)
- I write an independent review of the guarantee, its baseline and measurement, and the payment terms against your data. I recommend terms to change, a pilot structure, or rejecting the guarantee.
- Retain (When it makes sense)
- I stay close through the pilot and acceptance to review measured results against the baseline you agreed.