Predictive Marketing Analysis

Predictive Marketing Analysis

Predictive Marketing Analysis

Know market nexus in advance, operationalize insights, create and engage customers well before your competitors.

4Ps of Marketing Strategy

If you begin with Product, Price, Place, Promotion,… you’re in trouble. That’s Tactical Marketing mix. Same bloke (Kotler, 2000), but not so common:

  • Partition
  • Probe
  • Prioritize
  • Position

Looks pure Analytics ready language… Certainly Predictive Marketing Analysis utilises all the other P’s, too: Public Relations, Perception, Process, People [whatever it means], Price, Promotion, Place, Product and add a handful of them. Now, that we have just multiplied the territory of Predictive Marketing Analytics, let’s start then!

What is Predictive Marketing

Predictive marketing determines patterns and predicts marketing related future outcomes and trends. Predictive marketing analysis extracts information from existing datasets to create data models through predictive technologies. For example predictive marketing can help you:

  • forecast what accounts look like your best customers
  • identify what accounts make the most sense to target
  • prioritize the accounts and leads intelligently
  • master audience targeting: which content, when, which channels, …
  • native advertising parameters
  • and hundreds of additional beauties

DnA Predictive marketing analytics uses data science magic to provide you with a meaningful outcome based on big data and powerful, proven models – even for the 4Ps of Marketing Strategy.

Predictive marketing analysis ROI

Marketing Analyses give results – pretty fast – Case Study

1,2 m USD in 94 minutes: gems buried into your current marketing processes

400 000 m³ rock (about four-and-a-half times the volume of Royal Albert Hall) to wash to unearth Gold worth the same – We are better than gold miners

Symptom: “… coupon redemption rate is critical … the current predictive marketing system does not work again …”

14:20 Facts: Recurring issue: people do not redeem their marketing promotion coupons (generated by current Predictive Marketing Process)

Client: “… so we want to move from SAS Marketing to Python / R/ KNIME / RapidMiner or another advanced analytics platform…”

15:10 DnA Team’s predictive marketing analysis almost immediately (within 50 minutes) recognized, that

  • The algorithm is good, the data structure behind have not changed and everything is running fine, the outcome is accurate
  • People are receiving their coupons – according to the web / mailing / mobile / distro channel data
  • Real time sensor data shows that the promoted products are in the stores across the country
  • The redemption rate is critically low

So the predictive marketing algorithm has to have some issues [coupons to wrong audience], or … What do you think?

  • They do not like the product? [According to research the product shines in the top 10 most wanted – you’re wrong]
  • Timing: customers do not want to buy it now [Oh, they WANT to buy it, now and anytime]
  • Data has changed? [No, everything is fine, we run DnA redemption algos in KNIME within an hour and the results were similar]
  • Shops were closed, strikes, earthquake … [No, not really :)]

1.2 million USD – Let us count:

  • Marketing promotion loss only in this case – if situation remains the same – will be around 0.4 m USD
  • The profit loss on this campaign – due to inefficient promotion: 0.8 m USD
  • According to ignited original plan migrating from SAS to RapidMiner, regenerating algos, testing, training, platform, etc: 0.8 m USD at ground Zero.
  • No estimation on Brand value erosion either for the Product or the Retailer

15:15 DnA Big Data Social Media Snippet Algo vacuum cleaned up some recent activity then applied Advanced Text Analytics (approx. 20 minutes) and some outcomes:

  • Recently emerging negative sentiment towards both Product Brand and Retailer
  • Advanced text analytics shows tactical (so recently emerging) correlation between the words “no”, “[Brand name]” and “[Product name]”

Advanced Text Analytics shows, that customers blame the store / brand / product. DnA keeps close eyes on predictive marketing platforms, algorithms and solutions and the “algo went wrong” path was not really believable. Maybe some data input issue, or similar…

DnA Team put among others the streaming product sensor data into KNIME & Tableau and cast a glance at it:

  • spatial analytics – at shop granularity at a bigger shop we recognized distance errors: all promoted products sensor data were coming from the stock behind the shop not from the shop area (so not from the shelves)…
  • immediately recognized, that almost in all shops product shelves looks empty countrywide

Conclusion: the retailer shops are not filling the shelves regularly [it turned out later, that it had a reason] so the product seemed out of stock.

15:54 Presentation to the CMO

  • It is a logistic issue – necessary actions have been taken place already [“Refill the shelves!!!”]
  • Current predictive marketing system is fully operational
  • Recommended sensor monitoring and alert system details presented to avoid such issues
  • 1.2+ m USD saved / won for the company (unnecessary platform migration and saved revenue, + …)
  • Customer communication recommendations how to save Brand [beyond Predictive Marketing, DnA’s speciality is the Broadcast / Media / Entertainment so we know how to communicate with people effectively]

18:25 Checkpoint: Social sentiment recovered, streaming data shows increased / expected turnover realised

20:50 I caught a 11+ (lbs) Walleye (=Pike Perch or Zander) on my fave lake …

Side note: due to a retail communication issue, it was considered, that the whole – perfectly working – predictive marketing system went wrong. Weakest link. This is why we at DnA stress continuous “gentle and effective“ training to C- and senior level officers to understand Big Data, the opportunities and analytics power behind these technologies. Mind the gap!

PS: SAS to RapidMiner / KNIME / Anaconda / R migration evaluation has started, but for different reasons…

Data & Analytics Predictive Marketing Analysis Team 2015-2017.