Recommending the right product to the right customer sounds simple until a store tries to do it at scale. Most pet stores already collect some kind of data about customer purchase behavior, whether it’s order history, viewed items, or abandoned carts, but few manage to turn that volume of information into recommendations that actually help someone decide their next purchase. The result, in practice, tends to be a generic suggestion list that any automated system would generate, regardless of who is on the other side of the screen.
Hugo Galvao de Franca Filho, founder and director of Enjoy Pets, sees most pet stores getting this wrong not for lack of data but for using the wrong data or using the right data the wrong way. The difference between a recommendation system that works and one that just takes up screen space rarely comes down to how much information is available; it comes down to how that information gets interpreted before turning into a suggestion for the customer. This piece rounds up the most common mistakes in that process and what to do instead.
Treating every purchase as an equally strong signal
The most common mistake is giving every customer action the same weight: a product view counts as much as a repeat purchase of that same item. That produces generic recommendations because the system cannot tell passing curiosity apart from a real consumption pattern and ends up suggesting a product the customer already ruled out months ago, simply because they clicked once on that category during some unrelated search.
Hugo Galvao treats repeat purchases as the most reliable signal available, well above an isolated click or an item added to the cart and abandoned. A customer who reorders the same food every month reveals more about the right next product for them than any random browsing through the catalog, because that kind of repetition describes an actual consumption habit, not just an interest logged once in passing.
Personalizing without accounting for the pet’s life stage
Puppy food and senior pet food should not show up on the same recommendation list, yet that is exactly what happens when personalization looks only at product category, ignoring the animal’s life stage. The result is a suggestion that is technically related and, in practice, useless to whoever receives it, because it skips a basic detail any pet owner already factors into their decision.
This is, in the view of Hugo Galvao de Franca Filho, where purchase data needs to intersect with basic information about the pet, age, size, and known health condition to become a recommendation that actually makes sense. Without that intersection, personalization becomes just a nicer name for reshuffled generic recommendations, and customers notice the difference quickly, even without being able to name exactly what feels off about the suggestion.
Overdoing contact frequency
Having data about a customer is not an invitation to use it every week. A store that sends personalized recommendations too often teaches the customer to ignore the message before even opening it, which erodes the usefulness of the personalization system itself over time, even when each individual message is technically well targeted.
The standard used at Enjoy Pets, according to Hugo Galvao, follows the product’s actual consumption cycle: if a bag of food lasts thirty days, the restock recommendation lands around day twenty-five, not every week just because the data is available. Respecting that rhythm keeps the message relevant when it arrives, instead of becoming one more ignored notification in the customer’s inbox.
Ignoring what a customer never bought
Excessive focus on what a customer already bought hides an equally large opportunity: understanding what they never bought but likely need. A pet owner who has purchased food for two years and never bought a grooming item, for example, reveals an unmet need, not just a product preference the system should simply keep reinforcing.
Hugo Galvao de Franca Filho considers this kind of gap more valuable to the business than reinforcing recommendations for a product the customer would buy anyway. Closing that gap raises average order value without relying on discounts, because it addresses a real need that remains unmet, something no price promotion can replace over the medium term.
Fixing these four points does not require complex technology; it requires discipline in how the data already available gets interpreted and used before it turns into a recommendation. An example of a catalog structured around this logic is available at www.enjoypets.com.br.