Conversion Funnel Optimization

A large collection of tips for conversion funnel optimization

A list of conversion funnel optimization techniques


 What are funnels?
A simple statistical analysis on visitors from leadgen through checkout: the whole lifecycle. An e-commerce example:
-Visitors arriving to your website via eg. Adwords: 100%
-Visitor adding an item to their basket: 20%
-Proceeding to the checkout screen: 10%
-Entering their credit card no. / shipping address: 1%


 Conversion rates: When to start funnel optimization?
Funnel optimization is a fine-tuning process best applied after product/market fit, and validated unit economics.


 What's product/market fit?
Qualitatively, if you ask whether you have product/market fit, you probably don't;
Quantitatively, you can determine it by observing organic growth (cover/hide all drivers of growth from your traffic logs, that is attributable to in-house marketing; what remains is your organic growth. Is it growing? That's your product/market fitness.
Best indicator for long-term success of a newly launched product? Organic growth after the initial hype curve).


 If you're pre-pruduct/market fit:Customer developmentwill consistently get you better results.


 Metrics, and KPIs
You need to know the levers attached to your money flows directly. Canonically this is a mix of conversion, and retention; the techniques below aim primarily to improve conversion.


Cog: an industrial-strength PHP framework  A/B testing, and Multivariant testing
A/B testing is one of the simplest way for conversion funnel optimization. Basically, it's showing two alternative versions of the same page to different visitors, and using your conversion metric, figuring out which version "clicks" with your statistically significant target audience. Usual subjects are Call to Action, Title, "Purchase" button (and presentations thereof).
The difference between A/B testing, and Multivariant testing is the number of tests used: while A/B testing tests 2 alternative versions, Multivariant testing uses an inventory of arbitrary number of variants.


 Statistical significance
To figure out whether the difference between 2, or more variants of the same content is statistically significant, or just a random fluctuation, it's useful to calculatestatistical significance.

Most people shy away from doing conversion optimization due to low traffic. There are two aspects of this problem:

  • Low traffic is usually an indicator, that you haven't hit product/market fit yet. If that's the case, stop development, and start talking with your existing / target user base.
  • With that said, statistical significance availability is a function of data *only*; you can get this by just waiting a bit longer (datapoint: with ~100 uniques per day, you can determine an A/B test within a week).


 A/B testing: Server side VS Client side
The difference between doing conversion tests server side, or client side (eg via external apps: Optimizely, Google Website Optimizer, etc) is three-fold: setup, presentation, and URLs.
Client-side analytics is usually easy to set up: just add an extra parameter to the URL to display different versions of the same page. This makes them fragile: once the test is shut down, users who bookmarked the same page would return to a non-existing site.


 ROI analysis (expectation management)
ROI: 2 weeks / 1% improvement on average; follows asigmoid function; constrained at top usually by lead quality, market size, and leadgen efforts.


 Core equation for funnel optimization:
Incoming Visitors X conversion rate = no. of customers


Systematically dashboard every step on the funnel, target & improve the weakest (highest friction) step


 Vanity metrics
Every metric you're measuring should be Actionable, Auditable, and Accessible [ref]. Furthermore, pay close attention to first-order proxies of revenue (eg. conversion, lifetime value of Customer, bounce rate, etc), and find ways to exploit, and elevate these.



Optimization techniques:


 Product placement (what the user sees before they click):
check referers for userbase segmentation:
* where are they coming from?
* How does each of the leadgen sources compare from a full lifecycle perspective?
* What's the average time between landing on first page, and registering; and registration & subscription (or order / checkout? )


 -Thetag is heavily under-utilized by most webmasters; however, it is handled as first-class citizen by Google. If you have a single site-wide description tag, the search result's snippets are auto-generated based on the first unique paragraph on the page; however, if you set unique description tags for content-pages, that will be displayed on Google's SERP.
This has fundamental implications on SERP click-through rates -you can think about SERP as the shelf of products, and the other sites as the basis of differentiation. (SERP click-through rates, and the "long click" -user staying on your site without return to Google- might also be used as quality control signal by Google, which might have implications on your ranking)
In short: have a unique description tags for controlling SERP clickthrough rates; and have a uniquely relevant product placement in all referrer keywords that leads to you.


 --navigation optimization: know your referers, and consider adding specific landing pages for specific SEO / SEM purposes;
-- KYC: Know Your Customers: visitor language / geographic data can be acquired via IP -> denying sales by KYC for profit maximization (rejected credit cards, and other scams are costly; removing eg. Nigeria from the allowed credit card users list drastically reduces chargeback rates)


 Landing page optimization:
the first step in the funnel is probably the most important (first impression); nail the first experience.


 Improving the first run of the user experience( By Patrick:Improving the First Run Experience of your software):
-Never start out with an "empty canvas" -prefill it with options (Microsoft Word VS Balsamiq Mockups)
-Preserve the scent throughout the ad copy -> landing page -> first run
-Use a tour to have user experience value during their first run of the app


Design is a signaling mechanism of how serious you are. Bad design impacts conversion rates extremely.
--other side: pretty design
--The power of context: simple visual / copytext clues can lead to incremental results (people are picking it up subconsciously)
--Above the fold / under the fold:
-> google fold checker: things under the fold will get ignored for all conversion purposes
--landing page redesign: single element testing VS full page redesign. Choose it wisely
--Identify with your target users: your app is uninteresting -your costumers should be the focus. Don't sell to features -sell to benefits, and stories; your copytext should represent what your customers would look like after they use your products. Smiling people, successful businesswoman, etc


--Use simple language: 90% of the world don't talk english as first language; lowering the mental barrier for entry helps everyone. Discover "in-words" of your target audience to gain confidence; generally, speak their language.
--TheMonomythnarrative is a useful scheme for Call to Action optimization
--Social proof: the impact of reviews / social integration


 Payment Processing
-E-commerce sites:"...Sites who use PayPal or another off-site merchant solution have lower purchase completion rates."[ref: "But how am I doing Compared to Other Companies? -by KISSmetrics]


1, Make it easy to verify the accuracy of the information on your site.
You can build web site credibility by providing third-party support (citations, references, source material) for information you present, especially if you link to this evidence. Even if people don’t follow these links, you’ve shown confidence in your material.
2, Show that there’s a real organization behind your site.
3, Highlight the expertise in your organization and in the content and services you provide.
4, Show that honest and trustworthy people stand behind your site.
[ref:Stanford Guide for Web Credibility]

Credibility signals:
* BBB registration, third-party news outlet

Use trust logos to improve checkout conversion rates [ref,source]


 User flow
--Decrease the number of steps: perfection is achieved not when there's nothing to add, but when there's nothing to remove
--Decrease the number of asked-for information by pre-filled forms:
-FB, twitter, other data sources for name, e-mail address
-Paypal has shipping/billing info & phone numbers
--Decrease obstacles, by eg. crazyegg -monitor clickflow on the target site, and see what doesn't fit
--Frog boiling: some market segments prefer to be only displayed the current step. Some prefer having all the knobs presented at once. Test yours


 Grow optimization organically

-Best hypotheses usually come from internal arguments between engineering, and design. If a good UX idea is invalidated by KPIs, or a KPI points at a suboptimal design change, that's a strong indicator for the need of a "third way".

-Grow it externally, by validated learnings about consumer behaviour: user interviews, hallway testing, usability testing
-Crowdsourced problem detection: feedback button every page / live chat[3] / support chat


 Tech side

---ajax VS site navigation: matters a lot (page loading is a resource-expensive operation -ajax isn't)
---site loading speedup impacts user bounce rate significantly


capture user contact information early & often to keep users within the funnel via retargeting[4]/remarketing[5], newletters, and personalized offers. Since you know what they were looking for before bounce, you can do personalization.


 Useful market segmentations

-The goal of market segmentation is to custom-tailor the message presented to the user, so that they will deterministically take action upon it
-Where the user comes from (referers) implies what they are looking for
-Site versioning: returning visitors to get different landing pages, than first-time users
-Use previously acquired customer info for showing different content:
* GeoIP for geographic area
* Facebook can tell geographic area & gender of the user
* Rapleaf[6] can resolve e-mail address to gender & other user-specific information

-Think in devices:
* Mobile platform is fundamentally different, than web
* iPad / tablet platform is richer, than mobile, but still more rigid, than desktop web experience




[7] Massive conversion / metrics data across industriesby KISSMetrics
[8]Puzzling outcomes in A/B testingby Microsoft


About the author: Joel runs a product inbucation, and consulting company in San Francisco. If you'd like to read more by him, subscribe to the product development mailing list.