The tools that have impacted my Analytics life in 2016 the most?
ProductsUp, a powerful lever for marketing your product data; Analysis Workspace, an Analytics interface that turns data tables into a playground; Spotwatch, a tool that helps you watch your and your competitors’ TV spots and act upon them; Klipfolio, a dashboarding tool with data connectors to kill for; and Funnel.io which imports your cost data into Analytics while you’re sleeping …
This year, I got into contact with several technologies to enhance our Analytics setup at siroop.ch, the Swiss online marketplace. Initially I wanted to write a longer post about each of them, but then time just flies and there is so much to do. I still want to share them for inspiration in this post. This is not a review of new technologies, some of them have been out there for a while, but those I mention here have had the greatest impact on my work this year.
ProductsUp – The Lever for your Product Data
ProductsUp has become that central TMS-like hub when it comes to generating product data for any destination, e.g. marketing tools such as Google Shopping, Facebook’s product-based ads, or Switzerland’s popular price comparison site “toppreise”. But we also use ProductsUp to generate dozens of additional product dimensions (via classifications) in Adobe Analytics. That way, we can leverage anything that is in the product database without needing yet another data layer variable on the site (which increases maintenance, costs developer work, makes it easier for unfriendly bots to steal your product data by crawling the site, slows down the tracking, increases the request sizes and thus increases the risk of truncation or rejection especially by mobile browsers etc…). This for example allows us to analyze products by “product gender” (products for men vs. products for women), availability, by price range, create useful combo-dimensions like “Product SKU: Product Name”, and many more) without adding a single line of code to our website or TMS.
It would be impossible for our developer team to create and maintain all these individual product data feeds. So with ProductsUp, they only have to maintain one feed – a single raw “Source” product feed from our shop database that is sent to ProductsUp several times a day. The rest is drag and drop in the ProductsUp interface and some support from the ProductsUp team, no ETL developer needed. ProductsUp then automatically pushes your data wherever you need it. ProductsUp can even join one data feed, e.g. the raw shop product feed, with other data feeds, e.g. a Google Spreadsheet or a file from an FTP server (e.g. a daily Adobe Analytics FTP Export with product performance data). This way you can for example rank the products that should be shown in Display ads so that well-performing products according to Analytics show more often than those that do not sell.
What Could Be Better?
For being such an important middleware for scalable e-commerce online marketing, ProductsUp lacks important security features. The BETA version now finally includes some form of version control – that is important since one false click can destroy your entire setup. And even if only a little thing is broken, it is hard to find out what was broken by whom and when. Two-factor authentication for login should be another one, and a more detailed rights management (most importantly allow people to edit some product feeds, but not all of them (e.g. I want to edit the Adobe Analytics feed, but I don’t need the rights to screw up the feed for Google Merchant Center). And we were not able to use it reliably for Google Analytics’s Product Dimension Widening, but that might have also been due to some problems of Google Analytics itself. Maybe a generic built-in Google Analytics feed (like the generic Adobe feed they already have) would be a good idea here.
Analysis Workspace (part of Adobe Analytics) – Where a Table Becomes a Playground
In the past, in order to get fast and flexible analysis out of Adobe Analytics you had to run the Java-Applet-based Desktop program “Ad hoc Analysis” aka “Discover”. So the best part for an Analyst was hidden in a powerful, but a bit-geeky-looking program that you first had to download.
Nowadays, Ad hoc Analysis is still very useful for many use cases (lots of rows, pre-configured tables with lots of columns, FTP exports etc.), but for most cases, the browser-based “Analysis Workspace” has taken over. Of all tools (not only Adobe tools) I use, I spend most of my time in Analysis Workspace.
Analysis Workspace looks and feels good, offers nice visualizations (unlike Discover) and allows you to stack your data in any way you like. It is mostly fast (as long as you know what you shouldn’t do with it), it is all drag-and-drop – in short, the best and most user-friendly interface for doing deeper analysis I have ever seen.
I call it the tool where a table becomes a playground, because you usually start out with a simple table and then keep adding breakdowns, Segments, Filters, Time Ranges (Rolling (e.g. “the last x days”) or Fixed (Jan1 – Jan3)), Calculated Metrics, and of course you create lots of those elements on the fly. And there are other useful features like the Segment Comparison, built-in Anomaly Detection and more… So while Adobe’s great strengths for me had been the Segment Engine and the Calculated Metric Builder, Analysis Workspace has given me a way to put those strengths to work in a fast and effective manner.
See the example above where we split one metric into 12 segments and one dimension filter created with the dimensional value picker (“Channel X”), and we compare all of this for three different date ranges (the rows), one rolling (since Nov 10 until today (and today is tomorrow when I open the same Workspace again tomorrow), 2 fixed (20 Oct – 11 Nov and 9 Sep – 5 Oct). We can most importantly view all of this info in one small compact table, and it takes just a couple of seconds to load all this. The table is showing that, for visits entering through the Homepage, the two channels on the right have started to perform much worse before October 20, especially for New Visitors, whereas the other channels (which we group together into “Segment 1”) have stayed about the same.
What Could Be Better?
Let me just pick two here:
- Bugs: As revolutionary as Analysis Workspace has been, you kind of expect some more bugs than usual because the tool offers so many ways to move data around that it is impossible to test them all beforehand. But even so, there have just been too many bugs, most notably when it comes to PDF exports and scheduling. Empty exports, graphs without labels, emails coming 50 times within 5 minutes, Workspaces self-destructing etc… we have had it all. Adobe’s QA has failed, especially in the Fall release, and I have spent way, way too much time with Customer Care tickets. So before adding too many more revolutionary features, my biggest wish to Adobe is to straighten out the current functionality and improve QA in a way that it catches more of these bugs before they hit the clients.
- Feature- and UX-wise, Adobe needs to finally stop all this extreme text truncation – I rather have wrapped text than rows or columns that no-one can read anymore because only 4 characters are visible, especially when sharing them in a PDF or a screenshot (which is what I have to do when sharing insights).
Spotwatch – Your TV Impact in Real-Time (plus competitive intelligence)
Early this year, we were looking for a tool to track our TV spots’ impact on website behaviour. Our choice fell on Spotwatch, a tool that takes its name from being built around a “spot-watching” infrastructure. Imagine someone watching all TV channels all the time, writing down in real-time which spot is running where and making that data accessible to others. There are some of such “spot-watching” technologies out there, the initial challenge was mostly to find one that also covers the Swiss channels. Here Spotwatch was the only one. But Spotwatch does more than that, and that is why we took it.
- calculate the estimated Website Visitor uplift of a TV spot, TV channel, show, time and so on…
- see what spots competitors or brands are running who advertise products that we are selling as well. That way we can optimize our assortment (the products with high ad exposure are the ones whose descriptions and pictures you want to have looking nicely :))
- trigger AdWords or DoubleClick Bid Manager campaigns based on when our spots are running or when spots by competitors/by certain brands/etc… are running
What Could Be Better?
The uplift calculation inside of Spotwatch works for Visitors only, would be nice to have it for Revenue or other metrics as well. Moreover, when having concurrent spots, the GRP of these spots should be taken into account when calculating a spot’s uplift (otherwise you sometimes have documentaries on tiny channels outperforming a football match just because they had their spot at the same time by accident). I was told this would be coming in 2017, so looking forward to that!
Klipfolio – Dashboards with So Many Data Connectors
Klipfolio does not only seem to have the most competent support team on the planet (important because the first visualizations can be hard and advanced stuff even harder), they most importantly offer a dashboarding solution that is cheap, easy to access for anyone (app or browser), offers nicely-looking visualizations and – the deal-breaker for us – has a ton of easy-no-tech-needed data connectors to so many other tools out there.
As I have learnt over the course of my Analytics life, the tech effort is usually the tightest bottleneck not only when it comes to getting some tracking scripts on a website, but also when you want to get data from many sources (i.e. Analytics APIs, internal Excel files, databases etc…) into that one place where it shall all be joined and shown around. And I said “getting data” – let’s not even talk about maintaining all those API connectors.
Have Tableau? Great! And the number of tools with sensible Tableau connectors is growing. But still – getting all the data into Tableau in the format Tableau requires can take so long that you usually end up sacrificing some data sources. Klipfolio is able to take just about any data source and turn it into something you can work with without having to write a single line of code. (I admit it is a bit “apples to pears” comparing Klipfolio to Tableau because Tableau can also serve as a limited ad-hoc analysis tool, whereas Klipfolio is almost purely for visualization).
(this list is truncated and goes on much longer…)
So what kind of connectors am I talking about? Google Analytics? Check (but everyone has that)! Adobe Analytics? Check! A CSV or Excel attachment of a normal email? Check (that’s actually huge because it allows to integrate just about anything as most tools offer some scheduled email export)! A Google or Excel Spreadsheet? Check! An SQL query on a database? Check! A file from an FTP server? Check! Any REST API? Check! Salesforce Marketing Cloud? Check! New Relic? Check! And I could go on forever…
What Could Be Better?
The type-in formula editor has been a huge improvement as it now allows you to work much faster, but it is still really hard in Klipfolio to do something flexible (e.g. not hinging on “the data in Column E”), so that we often resort to not-so-sustainable “works-only-for-this-use-case-but-is-so-much-easier-and-faster” approaches. I would also like a better support or documentation of the JSON-based data sources, I struggle to do more advanced stuff with those (and this is the one thing where the Klipfolio support has not helped that much). And most importantly, loading our main dashboards takes soooo long and sometimes freezes the browser for half a minute and longer. Switching between dashboards thus becomes something you don’t like to do because it takes so much time (and why can’t I right-click on the dashboard name and open it in a new tab?).
I already wrote about Funnel.io in detail here, so suffice to say that it has taken away a lot of the pain involved in getting Campaign Cost data automatically into Google Analytics. We use it now for Facebook, DBM/DCM, Bing Ads as well as some non-autotagged AdWords campaigns. To get the DBM/DCM integration into a reliable format was painful to say the least, and we still have days every other week where we have gaps between what DBM reports and what funnel imports into Google Analytics (and we don’t have the time to go into every one of those gaps every time), e.g. shortly before Christmas all data before October was gone suddenly – but other than that, the cost imports are running well and Funnel has improved considerably in just half a year of using it.
There are some other useful technologies tools that enriched my marketing analytics life, but I am too lazy to write about them in depth:
- AdWords Scripts: Controlling thousands of landing pages is impossible without automation. So have a bot sent to your site to check for some conditions in your HTML (e.g. check if the products you are offering on the landing page went out of stock in the meantime) and have your campaigns paused automatically
- Google Analytics User Explorer: Of all the new Google Analytics features this year, I used this one the most because it helps a lot in debugging strange user behaviour (like duplicate transactions) – and as usual, the Google Analytics interface displays the data in a readable and helpful way (and does not truncate text like Adobe’s Pathing reports). But why do I need to create a single-user segment to see only a specific user ID’s protocol?
- Adobe Analytics Transaction ID Recording: This feature is very old, but it is so powerful and completely underused. It allows you to join the Visitor data of a Transaction with any other data from outside, e.g. a return of a product. That way we cannot only analyze which merchants have the highest cancellation rates, but also which campaigns lead to lots of refunds, and thus we are able to calculate our Post-Refund ROAS.
- DBM/DCM-to-Google Analytics Connector: This impacted my Analytics life this year because we tried to understand very long how it works, but there is hardly anyone around who knows how it works when you do not want to use only autotagged links (don’t expect your Google Analytics Premium partner to know about that). When we finally connected GA and DBM/DCM, it did not work at all, and it took over a week to disconnect it again (because that knowledge seems to be scarce as well), destroying all display campaign data in Google Analytics during that time. The promise is great (think of the AdWords integration with Google Analytics, just for DoubleClick), so let’s try this again next year.
- Tealium Data Layer Enrichment a.k.a. “Hosted Data Layer”: This new Tealium feature could be quite a gamechanger, unfortunately it cannot take more than 300,000 product IDs, and our shop has over 400,000 products, so the project had to be stopped for now.
As mentioned above when talking about ProductsUp, adding new variables to your data layer can be cumbersome. So another option is to push the data from ProductsUp to Tealium’s “hosted data layer”. When a user then views a product on your website, all you need to have in the data layer is the product ID. Tealium then pings its hosted data layer and adds the other data for this ID (brand, category, price, gender, availability, marketing labels etc…) to the data layer, and you can then push that data to tools that do not offer classification features like Adobe Analytics.
Even in Adobe, classifications are not ideal when it is important to track the current state of a product, e.g. the current stock availability. Adobe product classifications have no history. That’s cool for things like product names where you don’t want a new line in the report everytime someone fixes a typo, but for things like in-stock availability, it does not help because it would always show the current availability and not the availability at the moment the user saw the product.
Now It Is Your Turn!
Tell us about your most impactful Marketing or Analytics technologies of 2016!