Discover more from From Data to Product
Data Product ROI: Don't Shy Away from Cost Discussions
“Customers who use our product see a 10x increase in number of insights generated.” “The query time drops in half if you use our optimized system.” If you haven’t seen these type of claims, I envy your inbox. As data continues to capture market share, companies and products are quick to tell you how much benefit you can gain from using their product or solution. In many cases, I believe them. But knowing the benefits does not answer the question of “is it worth it?” We need to know the costs.
I think the last 10 years of data craze have created a world in which we are comfortable talking about remarkable increases in speed, efficiency, insights and access to data. It no longer seems okay to be “average” (at least based on my LinkedIn feed over the last year), everyone wants to be the best. Having “the best” is always going to be expensive. As we find our way through the volatile public and private markets, expense items that were approved without question 6 months ago might be denied without question now.
If you are in the position of either building or buying a product, you’ll need to find a way to discuss costs. In some cases, you might actually need to do a financial analysis. In other cases, you might just need to explain why the current state is not sustainable. Whether producing a dollar figure or a summary of why “things are on fire”, the story you tell about the costs of current state are key. Why? Because you need to explain the increase or change in cost from right now to future state.
This gets down to a fundamental issue: The benefit gained from a new approach, new product, or new [insert data word I haven’t heard of yet] needs to justify the change in cost produced. That change in cost is the thing people don’t like to discuss. Why? There are many, many parts to it. Here are some things to consider as part of the change in cost.
Establish your current costs. This should include people time, system maintenance, system variable cost and (very likely) opportunity cost of working on this system versus another. It is not likely that you’ll be able to pin down every detail. However, the mere fact that you are looking into these things puts you far ahead of others who are making similar pitches for new investment.
Draw a clear path for each current cost to expected future cost. It is really tempting to give a blanket “cost now” to “cost in the future”. However, that rarely helps others trust the numbers you are sharing. Be clear about which costs increase, which costs decrease and which costs stay the same. In areas where you are guessing (and everyone guesses), make clear what orders of magnitude you are estimating.
Be clear about ranges. No one believes your cost estimate of $13,545. They will believe a credible range that includes that value. It is really helpful to create best and worst case scenarios, even if the eventual cost estimate ends up being the middle of those ranges. I know this doesn’t sound as fun as working on ML models or experiments, but to get the support and investment you need, this is an important skill to develop.
Be reasonable about forecast length. Projecting change in cost 5-10 years down the road is often not reasonable. Try to be as clear as you can about the next 2-3 years, as things are likely to change within that window anyway. If your company is making decisions around data based on estimates 10 years out, that means there is a lot of guessing happening.
Tie it all together in a “what are we investing?” section. Using the word cost repeatedly will a) put everyone to sleep or b) make it seem like you are out to just burn through money. It is a good reminder to explain the costs as an investment and to package it that way.
If at the end of this exercise you can explain why you are building, buying, or some combination of both with an explanation of investment and benefit, the likelihood that your requests are taken seriously will increase. On top of that, by taking this approach to investment, you build credibility internally for how you think about data product investment and its relationship to business growth.
I’m the last person to say cost discussions are fun, but I’ve learned over the years that they are a critical part of success.