Die Cost vs. Die Price; Prediction vs. Reality

(disclosure I work as the Product Manager for a Sheet Metal Stamping Die Cost calculation product that is commercially available)

I have been having some really enlightening discussions with people lately about the applicability of this application to there business and what our product can/can’t do. It goes a little something like this.

act 1:

Prospect and I run through a sample part (similar to but not from the prospects parts). Software allows for import of the CAD data and with the use of either a predicted process or user input process determines resources and costs for those resources.

• me: so you see if I import the part data and define the stamping process, we can calculate a cost. Cost is based on a predicted number of hours, consumables, and mass of casting, etc and the rates for those items
• prospect: Wow, that’s cool…. But the cost here seems high. I don’t think I could sell the die for that cost
• me: Don’t worry. The database is fully editable, we can input in your hourly rates for Engineering, die makers, machine utilization, consumables, etc. Then we would recalculate and get a more realistic cost for you.
• prospect: Cost of machine utilization? Per hour? For like presses and things?
• me: yes.
• Prospect: but I don’t have those costs
• me: then how do you compute your costs today?
• prospect: we use an Excel spreadsheet, where the quoter puts in the estimate for number of engineering hours, number of stations in die/line, and number of weeks for construction and tryout. This generates the cost.
• me: that sounds like a price, not a cost
• prospect: that is  the cost. And yours still seems high
• me: don’t worry we can calibrate it still if you can tell me how many hours in each category for engineering, labor, mass of casting, or number of consumables were needed to complete each project. And from there you can calibrate the costing program to predict your usage, and therfore we can arrive at costs.
• prospect: sorry we don’t seem to have that kind of data
• me: so how DID you decide that my calculations were too high?
• prospect: I know that we can’t sell a ________ die for that price
• me: using your current system last year, were you profitable?
• prospect: uh nope, but again no body was profitable last year

act 2:

A paid customization/calibration project is undertaken. Several parts from the Customer (former prospect) are used to create data set for customization. Several parts of the same category but different projects are used. Customer also digs up the historical costs.

• Prospect: Here take a look at part A. Our cost for part A was 600,000 and you predicted 905,000. That’s too high even after we adjusted the hourly rates to match ours.
• me: Wow. That is pretty far off. How are we for the reported hours and mass of tool?
• Prospect: we don’t have that data.
• me: hmmmm. Okay we’ll look at that later. What else do we have?
• Prospect: Well now this is really interesting. We tried it for Part B and that cost was 1,200,000 and yor program predicted only 897,000
• me: wow again? But in the other direction. That is a big delta again. Any chance we know the labor hours or material for that job?
• prospect: nope.
• me: So we were really high on one job and really low on the other?
• prospect: yep. And to top it all off they were both inner wheel houses. Same customer, just different model variations. And you grossly over predicted one and underpredicted the other. I don’t think your program works at all.
• me: What? they were both wheel houses. How similar were they?
• prospect: Nearly the same size. just one is for the coupe and the other is the sedan.
• me: So for two wheel houses, with similar features, similar sizes we calculated each at around 900K or roughly 1.8 million for the two. And you built two dies for a combined 1.8 million.
• prospect: more or less….
• me: so what is the problem?
• prospect: you would not have properly predicted these cost?
• me: could anybody? What were your predicted costs?
• prospect: we quoted these jobs out at 750,000 and 800,000 respectively
• me: so not only did you lose your shirts, but you are not sure why?
• Prospect: look if you can get your program to quote +/- 2% then we can use it
• me: HUH?  but +/- 2% compared to what. Your not so good quotes, or your unpredicted losses
• Me: SILENTLY TO SELF. Basically, I know my costs are correct if they don’t match yours then right?.
• Me:Apparently you  accepted variation in your process in this case of nearly 100% accuracy for quote to real. I think we matched that…..
• Prospect: Sorry your costs are not close enough to reality for us to trust.

So yes. If you build a better mouse trap the world beats a path to your door. Except, they can’t admit that they have a mouse problem.

And if you do show them that you are catching mice. Those can’t be their mice. That mouse in the trap is dead, ours was alive and eating our cereal. (?????)

OK that analogy sucked. But really I can’t find any parallel for this at all. We can demonstrate through even the simplest definition that we provided benefit and that the benefit is real, and needed. But because the outputs don’t match the results they expect (results that they admit are subjective and perhaps even flawed, if not totally wrong) they can’t get on board with the software.