How to Improve Your Marketing ROI with Predictive Analytics

Over the past few months, you’ve probably heard a lot about the benefits of predictive analytics. One question that keeps surfacing is “I know the benefits, but how can I use it?”

 

Predictive analytics is a critical tool your team can leverage to increase your profitability. But what is it and how does it work?

 

Predictive analytic models break down into three main categories.

  • Clustering models segment your data into smaller groups.
  • Propensity models predict the probability of something happening.
  • Collaborative filtering provides you with situational recommendations.

 

Think of it as tools in a toolbox. Each model serves a specific purpose.

 

Clustering Models

 

Clustering algorithms segment large datasets down into more manageable pieces.

 

This is highly valuable when segmenting your customers down into smaller niche groups.

 

One benefit to working with algorithms is their ability to be triggered. When you add a new customer to your database, they can be automatically segmented into their group.

 

Leveraging the power of algorithms, we can also get much more granular with how we choose to view our data.

 

The most common types of clustering algorithms are built on behavioral or categorical groupings.

 

Behavioral Clustering

 

Behavioral clustering segments users on how they interact with mediums such as your website. When you track your user’s actions, you can start to answer more complex behavioral questions. How does this segment engage with discounts? How frequently does this group book a room? What time of year do they buy? How much money do they spend per visit?

 

This is important when you’re aligning your campaign’s message with their actions.

 

For instance, let’s look at a frequent purchaser who only stays on the weekends. Slightly adjusting your offer might be enough to incentivize a longer stay. Maybe a “stay three nights, get the fourth night free” might achieve the desired outcome.

 

Categorical Clustering

 

Categorical clustering segments customers based on their buyer behavior. These groupings help segment people who just buy room nights from people who are high targets of on-property upsells such as spa deals, or food & beverage.

 

One area you can leverage categorical clustering is when you are planning email campaigns. Tailor your messaging to speak to the behavior of each group.

 

This personalization can have dramatic impacts on your conversion rates. Everything from the images you display to the copy all impact the bottom line and the campaign’s ROI.

 

Propensity Models

 

Propensity models help you identify and predict future behavior based on historical trends.

 

Life Time Value

 

Looking at demographic and psychographic data, you can estimate a customer’s lifetime value.

 

This becomes critical when you want to rank segments and maximize your marketing ROI.

 

Not all your customer segments are created equal.

 

Think about marketing as water flowing from one bucket to another. To maximize your investments, you will want to fill up your most profitable buckets first. When those segments are full, you can then begin to spill over into the less profitable ones.

 

Share of Wallet

 

When a person spends money, those expenses break down into expense categories.

 

For this, let’s focus on room nights being a “travel expense.”

 

Share of wallet looks at the total annual spend in a category verse your share of the pie. How much of that spend was spent engaging with your property?

 

For example, say someone spends $100 at your hotel. Over the course of the year, that person spends a total of $1,000 at hotels. Your hotel is capturing a 10% “share of wallet.”

 

These metrics help you identify future revenue potential of individual customer segments.

 

Think about it. If you are only capturing a 10% share, that person has an additional $900 to spend on travel this year.

 

This number will give you a good idea on where you might have future revenue possibilities.

 

It is significantly cheaper to sell to an existing customer than to attract a brand new one.

 

Propensity to Engage

 

When you send out an email, what is the likelihood the person receiving that email will engage with it?

 

Using these models you can segment your email lists by the probability of engagement.

 

This reduces the cost of marketing by removing people who have a low chance of engagement. In return, a lower spend leads to a higher campaign ROI.

 

Segmenting your customers in this way focuses your campaigns on outcomes, which lead to significantly higher engagement rates.

 

Propensity to Unsubscribe

 

These models help to predict when someone is likely to unsubscribe. Using this information you can optimize the frequency of your email campaigns.

 

Optimizing your email spacing can mean that you vary the email frequency from to one segment to another. This speaks to the individual buyer’s journey leading to higher levels of continued engagement.

 

Propensity to Buy

 

Implementing a propensity to buy algorithm is another way to maximize your ROI.

 

This type of segmentation looks at your buyers and identifies where they most likely are in their buyer’s journey.

 

The further along someone is in the buyer’s journey the fewer incentives they will need to commit and buy. The longer someone engages with your company, the more invested they feel in the outcome.

 

Where this can impact your ROI is by pairing your offers and discounts to the various stages in your buyer’s journey.

 

The further someone is along the buyer’s journey the less of a discount you need to give them to get the booking.

 

Remove extra discounts and incentives to people who don’t need them. Lowering discounts closer to the conversion line helps to maximize revenue.

 

Propensity to Churn

 

These campaigns help to win back potential customers who may be considering switching to one of your competitors.

 

By identifying individuals who might be looking at making a shift, you can target those customers and win back their business before they sign or book with someone else.

 

Collaborative Filtering

 

Collaborative filtering looks at behavioral and demographic data to provide recommendations for future actions.

 

This help to identify customers who might be interested in upsell opportunities.

 

Upsell Opportunities

 

When someone checks into your hotel, what is the likelihood that person is interested in a room upgrade?

Why not analyze prior booking behavior and upgrade history to identify potential future upsell opportunities.

 

Why wait until someone reaches your hotel to mention an upgrade? A targeted remarketing campaign before arrival can have the same effect.

 

Cross-sell Opportunities

 

Cross-sell opportunities are useful in recapturing lost revenue. This is an excellent way to re-engage less profitable guests.

 

Use this information to promote dining or spa offers that add to the profitability of the hotel.

 

What’s nice about these models is you’re engaging with people who have already purchased your product. They just identify those who might be interested in buying something else you have to offer.

 

Use these models to identify opportunities to maximize revenue across your hotel.

 

Next Sell Opportunities

 

What might your customer buy next? This is easy if you are in retail because you can suggest future products they might be interested in. For hotels, it gets a little more complicated.

 

One way you could use this is by looking at the share of wallet numbers we looked at earlier. Run these against your propensity to engage numbers. This will give you a good idea of when a segment might be looking for their next vacation.

 

Promote reward stay or other incentives to attract them to come back.

 

Conclusion

 

Predictive analytic models can be extremely useful when integrated into your hotel’s marketing strategy.

At first, it can become overwhelming to dive in and try and decipher where to start. Pick one or two from the list and begin building them out.

 

Remember, this is not an exhaustive list. There are a variety of other models out there, and this is just an initial starting point.

 

The trick to getting the most of your models is to have them integrate directly into your existing tools. This way the data flows freely without interruption.

Zachary Maley
No Comments

Post a Comment

Comment
Name
Email
Website