Online Marketers And Predicting The Future
These days, I make a living by using Multichannel Forensics to predict the future.
In essence, I present five or six versions of "the future" to a business leader. The business leader picks the version that meets or exceeds the expectations of her board of directors, or her ownership team. With luck, the leadership team executes to that scenario, and will experience success over the next three to five years.
The feedback I receive from business leaders is usually positive. Increasingly, leaders tell me they like the strategic debate about "the future" a lot more than rehashing "what worked and what didn't work" in recent marketing campaigns.
So when Jim brings up a parallel topic about "prediction" in online marketing, I stop and think ... What is it about predicting the future that holds back our new generation of bright, talented online marketing experts?
There are fundamental truths about predicting the future.
- We are guaranteed to be wrong all of the time.
- Our prediction is likely to be more accurate if we incorporate historical results into our prediction.
- Our prediction is likely to be more accurate if we use better tools, better math, to make the prediction.
- Our prediction is likely to be more accurate if we have enough business experience to have "seen everything" before, allowing us to blend numbers/facts with gut instinct, vision, and competitive intelligence.
- Our prediction is likely to be perceived as being more likely to be accurate if we state it with confidence.
- Our prediction is likely to be perceived as being more likely to be accurate if we've been reasonably accurate in the past.
Maybe this makes sense. Catalog, retail or finance individuals tend to have more experience, tend to have more history to fall back upon, tend to have "seen everything" already.
Online marketers tend to have better tools, in my opinion.
Read any blog or trade journal about web analytics, e-mail marketing, or search engine marketing, and you won't see a lot of talk about "prediction". You'll see a lot of talk about "optimization". You'll see a lot of talk about "targeting". You'll see a lot of talk about "operations". You'll see a lot of talk about metrics, "KPIs" as they are popularly called these days.
This is my opinion, so I invite you to disagree with me in the comments section of this post. I perceive that the online marketing folks have spent a decade building an infrastructure for this new marketing thing called "the internet". Spend all your time focusing on executing campaigns and developing measurement techniques for said campaigns, and you get really good at those things!
As leaders, it is our job to mentor today's brilliant online marketing managers and directors, because these folks will be the CEOs and EVPs of the direct marketing ecosystem in the not-so-distant future.
Conversely, if you're a catalog expert looking for a niche in a world increasingly dominated by online marketing, spend your time becoming great at predicting the future of direct marketing --- be it the future of tools, techniques, or best of all, the future sales of the brands we support. Today's CEOs are looking for this type of leadership, and are not getting enough of it from the current generation of marketing experts.
Labels: Multichannel Forensics, online marketing
7 Comments:
Great blog, I have also created a lens in a same niche. This is my first time , hope u guys like it. Here’s a brief intro: …..Are you fed up with work? Maybe your boss is a grade-A pain in the behind. Well, don't fret about your current situation; it's time to look for an alternative route to income. That company office cubicle isn't the only way to earn a living. These days there are opportunities opening up all the time. With the World-Wide-Web in full swing, many individuals are turning to the Internet in search of a money making business. Have you ever considered this new-age road of opportunity?
Kevin, are you saying that predictive models don't work?
Perhaps we are talking about different kinds of prediction, but I'm not aware of any predictive models that "guarantee to be wrong all of the time"...
Or are you saying that even if you use a predictive model and it gives you a prediction, you should ignore it and wait for the results?
For example, I'm pretty sure you used the RFM model or some variant of it in circulation planning to avoid mailing catalogs to folks who were very unlikely to respond. Did you do the circ planning and then mail catalogs to the "unlikely to respond" folks anyway?
The online version of the same scenario is this: Given 3 online campaigns, and the ability to predict which will be most profitable within a week or so of starting the campaigns using a basic Recency model, should I just go ahead and continue spending on the weak campaigns anyway?
If you are advocating this approach, that's exactly what doesn't make any sense to me - isn't that a waste of money, isn't that the opposite of "optimization"?
As a circ planner, wouldn't you get fired for intentionally mailing catalogs to folks you knew were unlikely to respond?
I'm thinking you misunderstood the thrust of my post, and if it was not clear, perhaps now it is...
I'm actually fully supportive of predictive modeling, fully supportive!
What I failed to communicate is that predictive modeling is great, in spite of the fact that you're always wrong.
For instance, if you use RFM, ou might predict that the best segment will spend $10.00.
That segment will never perform at $10.00 ... the segment might do $9.49 or $10.63.
In that sense, you're always wrong. And yet, you will make your business a fortune using RFM.
So what I failed to say is that it is 100% ok to always be wrong. It really is ok to be wrong all the time!
This is what I see that is missing from some of the online marketing folks I work with. I see an unwillingness to stick one's neck out, because there is a high probability of being wrong.
What I was trying to communicate is that it's good to be wrong, and we have to mentor our online marketing partners to be wrong more often.
I was in 100% agreement with what you wrote!
I figured we were talking about the same thing in different ways...perhaps the difference between the perspective of an analyst and a marketer?
Anyway, as a marketer, I would never be looking for a prediction of $10, what I would be looking for is a relative comparison of the strength or weakness of an approach.
If I can predict which campaign will generate more profit, it doesn't really matter if that profit is $10 or $9.49, if I'm comparing to $6 and $4 (or $5.76 and $4.32). If there is not enough of a "spread" to make a call on which campaign is better, then I don't yet have a model that works and there is no prediction.
A predictive model is simply math, and since most web analytics folks have a background in IT, I would expect they don't have an issue with the math.
Rather, from what I have seen, it comes down to this: web analytics folks have been so successful with the "let's test it and see what the facts say" pitch to management that going in with a "prediction" is culturally problematic for them and perceived as a risk.
Hopefully we can change that.
Your last paragraph sums it up well!
Stumbled across your post this morning and you are absolutely correct. We (online marketers) have had such a difficult time convincing brands to spend money online commensurate with eyeballs and time spent. To build advertisng-based predictive models was what caused the first bubble to burst as everyone jumped in with misguided expectations. The idea has everyone jumpy.
We've got some legs to stand on now, and aren't afraid that brands will turn completely away. We need to be doing more modeling as you've suggested to rationalize the outcomes of one approach vs. another.
Great post!
Thanks for adding to the comments, I appreciate it!
Online marketers will make progress, it will happen.
Post a Comment
Links to this post:
Create a Link
<< Home