Focus on the MVDP – The Minimum Viable Data Product #startups #lean #data

Since the release of “The Lean Startup” there’s three letters that have been preached over and over…MVP – Minimum Viable Product. The over copied mantra of “what do you need to do to get this thing out of the door?”.

And in part it’s valid, there’s method in the madness. In part I also believe that it’s flawed as customers will buy in to a complete product not iterations and then used as little experiments to see what happens when you tweak the model. Life though in startup land it is never as clean cut as some people believe the lean model is. My issue with MVP is that the focus is on the product but I don’t see that as the most valuable asset to the business. For most startups the app, the website or API is just a delivery method to what really matters, the data.

So instead of the Minimum Viable Product, why do I never hear about the Minimum Viable Data Product? Let me give you a simple example.

Disrupting Bacon! The Sequel!

You can read the original bacon disruption post here. Sandwich ordering for me is simple I go into a shop order what I want, wait a bit for it to be made, pay and leave again. If you really want to appify the process then fair play to you.

Some people will look at that concept and think, “Simple Jase!, have the menus on the app, select what you want and pay within the app and pick it up. Job done and I’ll take my 8% fees from the sandwich shop owner.”

I proved in the original post that was going to be very hard work, growth of willing retailers has to be at least 75% month on month and then there’s small matter of getting users to download the app in first place. Remember, Just Eat didn’t come wading in the UK with £10m in their back pocket for no reason, UK wide marketing costs a lot of money.

Shifting The Focus To The Minimum Viable Data Product

Think about it, the data collected by the likes of Nectar and Clubcard is not really there for the customer’s benefit, that’s merely the pay off and by product of being given enough permission to mine every basket and every item that passes through the point of sale. The real value in that data is once it’s sliced and diced and you have customer profiles, regional segments of buying behaviour and detailed time blocks of seasonal purchases of everyday items. At that point the main sell is to the suppliers of the products.

Designing The MVDP

So thinking about it for sandwich and bacon disruption alike. If the first focus was on the minimum viable data product, how would it look? There’s usually some form of sign up process with a minimum level of information.

Firstname
Lastname
Email address

If the user, assuming it’s an app, is giving you permission to track their location to find the nearest vendors then location is a wonder metric. With permission there’s no reason you can’t store the app ID along with the location data. Does that customer move around a lot or are they stuck within a certain radius every day, every hour? Could I push information of the most popular eateries where they are if they move around a lot?

Location
Radius movement

The frequency and value of the customer is highly important. Are they using the your service everyday (signups is NOT a good metric, active users IS), what’s the value of their order, can you predict long term customer value based on the purchase data? Can you split it down per supplier? No, well find someone who can…

Date/time of purchase.
Purchase value ($/£)

What did the customer purchase? Even Subway miss this by a long shot, so do Starbucks (and they know what I’m buying). So here’s your chance for glory. Does the customer order the same thing everyday, creatures of habit, or is the customer varied in their purchases? Do they use the same supplier day in day out or do they move around the different ones in the area?

Retailer ID/name
Item list of ordered items.

There’s a possibility that an app will already be picking all this data up but has no idea how to process it, and that’s fine. Just store it and when you get to a point of thinking you have enough data find someone who can help you understand it.

Defining The Questions

There’s going to be two types of questions, one’s that you’ll ask yourself:

“Which suppliers are generating us the most revenue?”
“Who are our top purchasing customers?”
“How many orders a day is our app fulfilling?”

Those aren’t unreasonable questions and ones that I’d be expecting the board and management team to be asking everytime you have a meeting.

The retailers, your suppliers, on the other hand will have their own questions.

“How many customers purchased via the app?”
“What’s the sales volume?”
“How much in fees did I incur to fulfil those sales?”
“What were the top performing items in inventory month on month?”

Even with a simple list, trust me there’s loads more, there’s some wins in there for the retailer, especially in pushing distressed stock or capacity planning for more perishable goods. No point holding too much cottage cheese when no one is buying it.

From The Department Of “How Are We Going To Make Money?”

From the top, data has value. Not everything in life has to be open data. I made this very clear to a website many years ago, “The value is in my website”, to my reply, “No it’s not, the value is in the data your users are entering into your website. Don’t give it away”.

So as a new business (I’m going off the word Startup by this point) the key questions I’d be asking are:

What is our minimum viable data model, what do we need to collect to make better decisions later?
Can we mine, slice, dice this data and sell it on to our suppliers?
What is the value of the data we will have? Do we sell it on as an extra or as part of the product package.
What is the collection method to acquire the data? Is it a website, an app, smart TV application or something else? Will the data model change if the collection method changes.

And Finally…..

Customer buy in is crucial and it happens right at the start, the moment they sign up. From that point on everything can be tracked to aid your commercial goals. What’s important though is to chop and change the data model around once you get going, you need to make time to focus on the MVDP before you go into launch mode. If you change course halfway through and start asking for gender information users might get a bit irked and wonder if there’s really a customer segment where ordering sandwiches is really a matter of whether someone is transgender, female or male.

Just be careful what you ask for and how you ask for it.