She had a lot of customer-facing experience and, hence, that flavored a lot of her slides. Her slides each had a title caption of sorts which I am recreating here verbatim while I toss in my two cents on what I felt the explanation of as much was. Do note that there was an admission at one point that all of your analytics can be defeated by the HiPPO (highest paid person's opinion).
Get the data that matters. If customers are deceased, should you close off their credit card account? Executives attempting to prevent credit card fraud assumed so in one scenario only to be embarrassed by the reality that 95% of cards used after death were used by widows of deceased husbands. Instead it would have been better to target large purchases or another detail as a more focused indicator of potential fraud.
Everyone has enough data. There is a law of diminishing returns thing going on in thinking more data gives you a better picture.
Art and science. Take a dataset and see what jumps out.
Get directionally correction, not perfect. You can narrow in on scope too much. If you want to find a way to solve the problem of too many returned purchases the thing not to do is to drill into all customers who made returns in ninety days time where two things were returned in that window which were sporting equipment-related in nature and wherein the households of the returns held baseball aged children, etc. The Tammylynne Jonas example held even more of a drill in that this, but this is all I bothered to write in my notes. You get the idea, right? Don't filter away too far.
Don't build it yourself. Try to use good tools. Building everything yourself is expensive.
Data scientists are unicorns. They are hard to find and expensive to keep. Their compensation may not be worth the ROI (return on investment). Unicorns like to frolic with other unicorns, so a data scientist may not stay at your company long if there are not other data scientists there.
Not all data applies. You could drop the emphasis on how weather patterns affect sales and instead focus elsewhere. If you are trying to give people ads that trigger impulse purchases such as ice cream on a hot day, well, maybe that's fine for Dairy Queen soft serve sales but might be the wrong carryover to another space. If people are stuck in traffic (a bigger concern here than the weather) a duck into Dairy Queen for some Dairy Queen "ice cream" might entice, but that won't help you sell pants at Kohl's.
Should be actionable. Try to piece together what leads to what in lieu of just thinking: "Oh, no we have a problem." In the retail space 70% of returns are because something is too small.
Context patterns. In beacon technology we don't have to have GPS accuracy within twelve feet. We need to know what you are looking at/for. What does a longer than a sixty second delay in a department store tell you? Is the party comparison shopping, shopping with a friend, checking Instagram, unable to find the desired color and size, or confused in general? If you could fill in a story maybe you could help the hung up.
Sometimes there is no data. On autonomous driving cars, the artificial intelligence is like a toddler in that the more feedback you give it the smarter it gets. There are no autonomous driving cars that can handle snow or icy conditions however because the feedback hasn't been provided.
Get more than what your competitors have (but no less). You can't be late to the game. You cannot wait to have more perfect information in some scenarios.
Collect more data at the sources. Millennials and Gen Z are excited to give information on themselves. Make BuzzFeed Quizes and give them to Millennials and Gen Z. Which type of vegetable are you? etc.
Monetize what you can. At Coca-Cola you want to know how much coke you sell in a similar space as it compares to Pringles and twenty ounce Mountain Dew and their wallet shares.
Common sense still needed. Email campaigns and direct mail campaigns won't add up to more sales if the inventory is zero. (This is a scenario that actually happened that was spoken to.) Your analytics don't tell you about an inventory problem or, for that matter, a URL being advertised incorrectly. Sanity check to ensure a use of the head (do you rationally believe it?), heart (overcome feelings), and hands (can you get your target clientele to interact with it?).
Data doesn't solve everything. Sometimes trial and error gets you there.
Be compliant... not crazy. Don't worry too hard about compliance with the CCPA (California Consumer Privacy Act) or the GDPR. There could be fewer than ten requests a month telling your coffee company to drop clientele data. While "customer old preferences" have to be dropped when requested per the CCPA, that is subject to interpretation.
No comments:
Post a Comment