You work for a food delivery company. Your CEO says let’s try an on-time guarantee product.


  • Yeah, do whatever.
  • Another feature added to your long list of features backlog.
  • Get devs to build. Not hard considering every team is overstaffed.
  • Ship it.
  • Mention it in your performance review.
  • Get some applause.
  • Nobody remembers after a year.
  • Feature gets shut down.

Now: (There will be questions, lots of it.)

  • Are we trying to win the reliability narrative with this feature?

  • Is it our attempt to win the faster delivery narrative?
  • Will it be a separate service with a higher price tag? And thus secure a higher margin?
Is it going to be an add-on, like the insurance products in the ride-hailing business?
  • Is it going to toggle the checkout process? When are we going to show the toggle?
  • Are we going to give it to every customer?
  • Are we choosing carefully?
  • Do we select by running some kind of ML model that predicts, on the basis of historical data, which deliveries we can promise will be on time? But running an ML model every time we show checkout is hugely expensive?
  • Do we do a deterministic “if else” thing, which is less costly, may be less accurate, but will save us the cost?
  • What if we don’t do the delivery on time? Then they get a voucher? Means there is the risk angle? How do we predict the risk?
  • Are we shipping this to verify whether customers are willing to pay more for on time delivery?
  • Why not just show the toggle and when they turn it on, say it’s coming soon? Nope, company is too big for such cheap experiments. -
How much are we willing to spend for the validation of our hypothesis? -
Okay, so we do a model of the whole thing?
  • If this is successful: Will it become part of the subscription product?
  • Will the cost of the on-time guarantee be included in the subscription? Or do we add it to the subscription product and charge more for it?
  • If our deliveries are on time 80% of the time anyway, then maybe it makes sense to give it away for free in order to win the narrative of being faster and more reliable than our competitor? Will people switching to us and the increase in MTUs be worth the cost of the on-time/voucher?
  • Once modelling is done, have to present to risk team.
Have to do a presentation to the regional ops team. Need to show our CFO. He is very aggressive in terms of cost reduction. Let’s see if he is sold on it.