Se ha denunciado esta presentación.
Utilizamos tu perfil de LinkedIn y tus datos de actividad para personalizar los anuncios y mostrarte publicidad más relevante. Puedes cambiar tus preferencias de publicidad en cualquier momento.

Measurecamp - Improving e commerce tracking with universal analytics

My MeasureCamp presentation covering how I've used the beta version of Universal Analytics and Measurement Protocol to extend the e-commerce analytics functionality of GA.

  • Inicia sesión para ver los comentarios

Measurecamp - Improving e commerce tracking with universal analytics

  1. 1. Improving e-commerce analytics with Universal Analytics Matt Clarke @techpad, 17 February 13
  2. 2. I love Google Analytics, but as an e-commerce director, I encounter questions every day that it can’t answer. @techpadSunday, 17 February 13
  3. 3. What’s the conversion rate for product X? @techpadSunday, 17 February 13
  4. 4. What’s the conversion rate for product X? You can’t easily determine the conversion rate for products, categories or brands to monitor trends or optimise them.You might be able to do it externally with API hacks, but it’s not easy... @techpadSunday, 17 February 13
  5. 5. How much stock of brand Y did we sell? @techpadSunday, 17 February 13
  6. 6. How much stock of brand Y did we sell? Oddly, brand isn’t a standard dimension. Reporting on brand performance requires manual custom reports or dashboards. @techpadSunday, 17 February 13
  7. 7. How profitable is that Adwords campaign? @techpadSunday, 17 February 13
  8. 8. How profitable is that Adwords campaign? Google Analytics doesn’t consider Cost Of Goods Sold (COGS), so to calculate profit, you need to do it all externally. There’s no cost import option for product costs yet. @techpadSunday, 17 February 13
  9. 9. What’s the average margin for that category? @techpadSunday, 17 February 13
  10. 10. What’s the average margin for that category? Similarly, because there are no cost data for product costs, you can’t easily monitor margin trends... @techpadSunday, 17 February 13
  11. 11. When we increase the price of product X, what proportion of customers buy, defect or switch brands? @techpadSunday, 17 February 13
  12. 12. When we increase the price of product X, what proportion of customers buy, defect or switch brands? Price Elasticity of Demand analysis would be nice, as would metrics on latency to indicate which of the customers we acquired are still customers. @techpadSunday, 17 February 13
  13. 13. Google could fix some of this for us now... • If you could import product cost data, you could see product profit and margin, and the actual ROI/profitability of ads. • If you could connect Google Merchant Center data to GA you could get product conversion rate, brand performance data and new variables to use as custom segments. (You can do it with the API.) @techpadSunday, 17 February 13
  14. 14. That’s unlikely. But Universal Analytics might help... • You can create custom dimensions for product, brand, manufacturer and supplier to improve reports and aid analysis. • You can create custom metrics for product cost, profit, margin and shipping costs. • It’s in beta and not everything works properly (or at all) yet. It has limitations, but it’s still a big step forward. @techpadSunday, 17 February 13
  15. 15. Here’s how I’ve used it @techpadSunday, 17 February 13
  16. 16. Measurement Protocol and Universal Analytics • I opted to use a combination of the client- side Universal Analytics (analytics.js) and the server-side Measurement Protocol. • Using Measurement Protocol allows me to bypass the event limits of Universal Analytics and to send data (such as profit and product cost) that I don’t want to appear in the client-side JavaScript code. @techpadSunday, 17 February 13
  17. 17. Server-side data layer • To ease data access, I created a server-side data layer containing data objects for page, product, customer and transaction. • I push data to the client-side code from the (private) server-side data layer. • If I later decide to move to Google Tag Manager, or even switch analytics vendors, it’s a relatively easy task. @techpadSunday, 17 February 13
  18. 18. How it looks... Web analytics Database extension Page object Customer object Order object Product object Measurement Data layer Protocol and UA @techpadSunday, 17 February 13
  19. 19. How it works Web analytics Data layer Page type? extension Order page E-commerce tracking, Product page Other page plus product, supplier, manufacturer custom dimensions. Plus... Sensitive stuff I don’t want to be visible in the The usual, plus product, Universal Measurement client-side source code, supplier, manufacturer Analytics Protocol ie. profit and margin. custom dimensions to connect at order page... @techpadSunday, 17 February 13
  20. 20. Universal Analytics @techpadSunday, 17 February 13
  21. 21. The basics of UA • Uses analytics.js instead of ga.js and replaces all the old GA cookies with _ga. • Has a totally new syntax, but all the good stuff remains - including events. • Custom dimensions and metrics replace the old custom variables. • Will support user ID (but doesn’t yet). @techpadSunday, 17 February 13
  22. 22. How I’ve used it • On the product page, I’m firing custom dimensions for: product, brand, manufacturer, supplier, author and product cohort. • I’m hoping to connect these with data fired on the payment complete page (attached to transaction/item) to allow me to get data on stuff bog standard GA can’t measure. @techpadSunday, 17 February 13
  23. 23. An overview of custom dimensions and metrics • First, define your custom dimension/metric in the Google Analytics UI. • Send your CD/CM with another data type, such as a pageview or an event. • ga(‘send’,‘pageview’,{dimension1:‘Apple iPad 3’}); • You can combine several together. @techpadSunday, 17 February 13
  24. 24. Measurement Protocol @techpadSunday, 17 February 13
  25. 25. The basics of MP • A server-side way of sending GA data (like PHP-GA), based on a simple URL system. You send a payload to an HTTP endpoint. • Sending data with an HTTP POST returns a 200 OK. (Sadly, even if it’s malformed...) • There’s currently no client library, so you’ll need to code your own. However, this is quite straightforward with PHP and Curl. @techpadSunday, 17 February 13
  26. 26. How I’m using it • Used on payment page to send data I want to hide from anyone viewing the source. • I want to supplement the e-commerce tracking with extra custom dimensions and metrics, such as product, brand, supplier, manufacturer, profit and margin. * • I’m joining these to UA dimensions fired on the product pages attached to pageviews. @techpadSunday, 17 February 13
  27. 27. What do I send? • Latency - days since last purchase (CM) • Customer cohort (event) • Days as customer (event) • Lifetime spend (event) • Carrier (CD) • Newsletter preferences (CD) @techpadSunday, 17 February 13
  28. 28. What do I send? • Product (CD) • Brand (CD) • Manufacturer (CD) • Supplier (CD) • Product author (CD) • Product cohort (CD) @techpadSunday, 17 February 13
  29. 29. What do I send? • Product cost (CM) • Product margin (CM) • Product profit (CM) • Shipping cost (CM) • However, the CMs and CDs attached to the item and transaction don’t work yet. @techpadSunday, 17 February 13
  30. 30. Using Measurement Protocol alongside Universal Analytics @techpadSunday, 17 February 13
  31. 31. Using Measurement Protocol with Universal Analytics • Measurement Protocol and UA use a visitor ID called the client ID (or cid). • This isn’t to be confused with the user ID. (That doesn’t exist yet.) • If you want to use Measurement Protocol and UA together, you need to use the same client ID. Otherwise GA won’t connect the data to a given visitor. @techpadSunday, 17 February 13
  32. 32. 32-bit IDs vs UUID v4 • Universal Analytics (ga.js) cookies use the old-style 32-bit UUIDs for compatibility with the current version of GA. • The docs (and GA team) recommend using UUID v4 for client ID, but (undocumented) it can also support 32-bit IDs. • If you want to use both together, you need to use the _ga 32-bit ID, not UUID v4. @techpadSunday, 17 February 13
  33. 33. Obtaining the 32-bit ID • Parse the new _ga cookie to obtain a given visitor’s 32-bit ID to send with MP data. • Example: 1.2.123456789.9876543219876 • Extract the last two items from the cookie and use that as your 32-bit ID. • I’ve written some cookie parsing code. Get a copy from here: @techpadSunday, 17 February 13
  34. 34. To round up... @techpadSunday, 17 February 13
  35. 35. Measurement Protocol and Universal Analytics are amazing • They open up new possibilities for tracking. • They should help improve analysis. • They’re better integrated than CVs. • You get 20 CDs and 20 CMs. • The new user ID (when supported) should make even more cool stuff viable. @techpadSunday, 17 February 13
  36. 36. But there are a few issues... • You can’t send custom dimensions and metrics with either a transaction or a transaction item yet. However, the GA team is planning to add support, which is great. • At the moment, custom metrics can’t include a decimal point. Why? (Consider rounding and multiplying, extracting in API, then dividing to get the original value.) @techpadSunday, 17 February 13
  37. 37. Tips • Using a data layer means you can use data in Measurement Protocol and Universal Analytics easily and keeps things tidier. • Use Measurement Protocol to send stuff you don’t want visible in the JS code. • Parse the 32-bit ID from the _ga cookie and use it as the Measurement Protocol client ID, if using both together. @techpadSunday, 17 February 13