You know how the further along we get in business, the more we talk in business-y language? We’ll start referencing holistic overviews and something-driven approaches and we’ll pencil in times to touch base about shifting bleeding-edge paradigms at scale.

Okay…I do use the term “at scale”. I apologize to those whom I’ve afflicted with my business jargon.

One word that you’ll hear very often in business is analytics. Since it’s often used as jargon, it’s very easy to assume that it’s just another valueless cliche like “leverage” or “robust”. This word though — analytics — is a diamond in the rough, a real rose among thorns (okay, I’ll stop). In all seriousness though, what are analytics? How can they demonstrably affect how I build my applications? When should I care about them? Let’s dive in and explore the answers together.

Analytics hone every area of the business

When you think about the word, analytics just refers to data that will be analyzed. You might then ask, what are we searching for in this analysis? Well, it completely depends! Well-crafted analytics serve as input data to lots of different processes, both on the business side and on the technical side.

Let’s run with an ecommerce example to explore both of these aspects for a bit. Imagine that you’ve been recording each time a product is clicked on your site, and each time a transaction occurs. That data alone is generic enough to be used for all sorts of different things, but it’s still applicable enough to be valuable in all of those situations.

For example, on the business side, you could ask yourselves questions such as:

• Do some products not get any clicks, let alone conversions? This would indicate that the product just isn’t attractive to your audience and/or that it’s not discoverable.
• Do some products generate a lot of clicks, but with very few conversions? This might hint at something being wrong with the product page or the way that the product is currently represented; it could also just be a factor of the product itself.
• Are some products almost always the only product in the conversion? Perhaps this reveals this product is more of an impulse buy, bought before the customer even added anything else to their cart.
• Can a large group of the conversions for a particular product be attributed to one marketing channel? The obvious effect here is that if so, we should do more of that marketing!
• What pages on our site are best at turning clicks into conversions? Well, then let’s find out what makes those pages conversion-machines and apply that logic elsewhere.
• Are certain products getting more seasonal interest than you’d expect? Then perhaps your marketing relating to this item should adapt more to the seasons than it currently does.
• How many products does a user click on before converting? If this number is high, it might indicate that your product pages aren’t converting well enough, or that maybe the structure of your application is making it difficult to find the items that the potential customer wants.
• Do repeat customers keep searching for the same items? Maybe it’d be a good idea to keep those items front and center so there’s no friction on their repeat purchases!
• How often are potential conversions abandoned? Again, if this number is high, it points to something about your application flow slowing down potential customers and dissuading their interest.

See how these questions have answers that aren’t just for business folks? When we track clicks and conversion events from day one in a flexible system like Algolia, we can then improve our business “holistically” — in other words, tracking analytics makes the business better all around.

Analytics can be input data for new features

AI is one of the biggest tools that we use nowadays to solve complex problems in ecommerce. For example, Algolia Recommend can (as the name suggests) recommend products, content, or categories to users based on some previous data. Maybe we’re suggesting products they’d like if they like the one they’re looking at now or products that are frequently bought alongside the one they’re viewing now. Either way, we’ll need input data to train the AI on our product catalog, and the way we do this is to send click and conversion events to Algolia!

Note that different features require different amounts of training data. For example, Algolia lets you A/B test different index settings, but that only makes sense if you have a way to measure the results. Enter analytics. When you send click and conversion events from the beginning, A/B testing can quantitatively measure the effects of slight tweaks to your search index configuration. That means that the more data we have available, the more of an impact we’ll have on our app. Personalization works this way too — if we don’t have any data on the users behavior, that’s alright! We just won’t personalize the results in any way. But the more data that we have on our users’ behaviors, the better we can tweak the search results to be more likely to reflect preferences.

On the other hand, how would you create a Frequently Bought Together section without at least some data? You can’t just display random results because it’ll be obvious that they’re random — for example, if you run a tech reseller, you’ll end up saying that iPhones are commonly bought in the same order as Galaxy Phones, whereas the much more appropriate suggestion would be a charger or a phone case. Algolia can handle the Frequently Bought Together section for us — by sharing at least 1,000 conversion events (checkouts, in our example) with multiple products over the last 30 days, we’re then good to go.

Related Products are a section that ecommerce sites generally should have. And thanks to Recommend’s hybrid engine, you can start making related product recommendations from day one based on information already in your records. But to truly reflect your users’ behavior, you’ll need event data — 10,000 clicks or conversions over the last 30 days is enough for Algolia to make accurate conclusions about what products generally relate to each other. If you start tracking click events from day one, though, it’ll be likely that by the time you choose to implement this Related Products section, you’ll already have the necessary data to train the model in Algolia, so that you can move ahead right away.

Analytics can save you from future headaches

Let’s be honest. If you’re the developer, your manager is going to ask you to implement some analytics system at some point. And if you’re the manager, you know that you’re going to ask for that. The problem that happens in many codebases though is that different analytics are tracked in different ways because they were implemented at different times and for different purposes. Often, the same events are duplicated in multiple analytics systems, an occurrence despised by everybody involved. How can we avoid this conundrum?

Here’s a simple answer: start tracking as much as possible from day one.

This simple guide can help, and here is the gist:

• Break down every little thing that could happen on a page stemming from user interactions.
• Be as detailed as possible, preferably including enough information in the event itself and could even recreate that event programatically.
• Build the system so that you could plausibly track a user’s every choice from the moment that they arrive on your site, to the moment that they leave.
• Don’t forget about tracking events in interaction points outside of your main site, like in emails, ads, push notifications, or across different domain/subdomain properties.
• You can’t have too many meaningful events! Definitely don’t send an event for every pixel the user scrolls, but if they linger on a certain section of that site, that has meaning! Event-ify it!

In the future, you’re going to end up implementing more features that need analytics, but you’ll already have that data on-hand. You can always filter out the extra data within your analytics dashboard that you don’t care about, but you can take comfort in knowing that data still exists for whenever you or somebody else does need it. You’ll never have to reimplement the concept of analytics, or divide your event data between different platforms. You’ll never have to worry about taking duplicate events into account when analyzing how your business is performing, and you’ll always have enough data to model your customer flows accurately.

What have we learned?

• While some folks talk about analytics like just another piece of cliche business jargon, it’s actually a super useful concept that should be implemented almost immediately in your application.
• Features that you’ll want to implement down the road may depend on having analytics data available to you already.
• You’ll probably end up implementing some metric-tracker a handful of times for different features and using different platforms — so you’ll save yourself a lot of work by planning ahead and storing everything now.
• Algolia is a great place to store everything! If you’d prefer to store your analytics data elsewhere, but want the benefits of Algolia’s dashboard and AI-driven search, there are plugins for GTM and Segment available too.

If you’ve made it this far, I’ll bet you’re ready to get started tracking clicks and conversions in your application. Here’s the link to sign up for Algolia and here’s the detailed instructional guide to help you implement this. And don’t worry if you haven’t rode down this path yet in existing applications — now’s your chance to course correct (and we won’t tell anyone 😉).