We’re not aiming to gain followers with our Instagram ads (we’re measuring their effect via conversions/ROAS), but it’s a positive side-effect.

It does make sense when you think about it — if you found an ad interesting (especially if it was from an inspirational account that wasn’t trying to sell too hard to you), you’re more likely to follow the account. From here we can try to predict how many followers we might gain per 1000 ad impressions, and how much each follower might cost us.

Combined with the fact that 20% of our customers have engaged with our Instagram account, Instagram is seeming like it might be a really good way to nurture leads, and will hopefully get better as new marketing features are released.

So how can we find out if followers become customers?

The trick is to use Custom Audience Overlap

To get the answer, you’ll need two Custom Audiences – one of recent customers, and another of the people who have engaged with your Instagram account. Here is a good guide on setting up custom audiences.

Creating an audience of recent customers can be done by uploading a list of your customers, or by using the “website activity” as below. You’ll need to adjust the time period so there are at least 1000 customers in this audience.

To create an audience of your Instagram followers, create a Custom Audience using “Engagement”, then select “Instagram Business Profile”.

Bringing it all together

Now for the payoff! Select the two audiences then from the “Actions” menu choose “Show audience overlap”.

Voilà! You’ll be presented with a window showing you the crossover between people who have engaged with you on Instagram and your customers. In our case, 20% of our customers have engaged with us on Instagram.

Tracking over time

If you’re putting as much effort into your Instagram as we are, you might want to see the changes reflected over time. To do this, you could for instance create a 180 day audience that excludes the 90 day audience, to get two comparable time periods.

Happy Instagramming!

Thanks go to my amazing marketing team (Milly and Jess), and Jon Quinton for checking the draft of this post.

P.S. find out how we use Google Scripts to track our follower count over time.

When a customer buys a print from King & McGaw, they can choose to have it framed with one of dozens of frame options. Framed products have a much higher customer satisfaction rating than unframed, and also generate 5x revenue.

Last year, we ran a series of tests which made framed products the default option. A series of A/B tests eventually lead us to a 20% uplift in average order value, and a confidence to think of frames as our primary product. Changing defaults can have powerful effects.

You can’t add simplicity, you can only take away complexity

This year as part of a drive on simplification, King & McGaw removed thousands of products from the site. Opening our London store increased their confidence in selling a smaller, more opinionated range.

The site still had a cluttered product page with too many choices, and converted poorly on mobile. A fundamental overhaul was the only way we thought we could get to a better situation.

Enter the Design Sprint

We took the best parts of the Design Sprint process and condensed it into three days (here are the actual notes). Some of it was squeezed into a train journey to the King & McGaw factory in Newhaven.

Our goal was to “build a high-quality, simple to use way of buying framed art that is curated and guided, and converts equally well on all targeted devices.”

After three days of interviewing, sketching, debating, voting and countless sticky notes, Nick Cotton produced this prototype, and Alannah Wood used it to start collecting feedback from customers in our London store.

The key change was displacing the upfront choices behind the “customise” button. We knew this would have an effect on a customer’s propensity to buy the “default” choice, so we ran an A/B test (~50k samples in each branch) to find out how the new page performed against the old.

A/B test results

Our analysis found that the new page had the following effects (to > 95% statistical significance):

• increased add-to-cart conversions by 13%
• increased purchases of the default option by 14%

The improvement in add-to-cart conversions was a welcome result. But confirmation that removing upfront choices caused more customers to buy the default option wasn’t necessarily good, as we default to the smallest size.

The shift to smaller products eroded some of the value brought by the change. So our next test will change the default size of our products to encourage customers to buy larger sizes. We hope to do this without losing our add-to-cart gains by losing customers to sticker shock.

Hit the follow button or add a comment if you found this useful!

Over the last few years we’ve experimented with pop-up shops as a way of drawing attention to our brand. It seemed inevitable that we would open a permanent store eventually, so when Heal’s invited us to open a concession in their flagship London store we jumped at the opportunity.

Within 3 months, we’d hired staff, built the store and opened the doors to our first shop. Here are some of the things we’ve learned so far.

We can collect insights from real customers 7 days a week

We put a lot of time and effort into understanding our customers – through surveys, user testing, market research and customer support insights. With the store, we have feedback from real customers 7 days a week.

During the first few months, the E-Commerce team was rotated on shifts to train the new shop assistants, and iron out any kinks. We were learning so much that it became obvious the shop staff should write daily reports so we could share insights across the company. Many of these have lead directly to changes on kingandmcgaw.com.

Having a finite space forced us to reduce our range by 99%

Most of the 19,000 products we sell on kingandmcgaw.com are manufactured on demand, so we’re not limited by factory space. The downside is we’ve avoided hard decisions about which products we should keep or kill.

When we opened the store, we had to to distill the range down to 150 products. Reducing the range by more than 99% was hard, but now we have a litmus test for new products – “would this go in the store?”.

It’s also been eye-opening to see the effects of merchandising first-hand. This is much more obvious and immediate in a store with a narrow range than it is online.

Discrepancies between online and offline is confusing

Some products and offers available in the store that weren’t available online (and vice versa). This was confusing to customers and embarrassing for us.

Can I use vouchers in the store? Can I have my items delivered to the store? Can you accept returns in the store? Why isn’t this product available online?

No amount of planning could prepare us for every eventuality so we had to make decisions very quickly. We’re now working to ensure our range and messaging is consistent across channels.

Omni-channel attribution is difficult

Understanding and attributing the indirect value the store brings has been difficult. Though the website is driving traffic to the store, and some of the in-store leads convert online, the full picture remains blurry.

Have you ever opened a store for an E-Commerce brand? I’d love to hear from you in the comments!

Hacker Noon is how hackers start their afternoons. We’re a part of the @AMI family. We are now accepting submissions and happy to discuss advertising & sponsorship opportunities.

If you enjoyed this story, we recommend reading our latest tech stories and trending tech stories. Until next time, don’t take the realities of the world for granted!

E-commerce lessons from opening a bricks and mortar store was originally published in HackerNoon.com on Medium, where people are continuing the conversation by highlighting and responding to this story.

In January this year I focussed a lot of effort on redefining KPIs for kingandmcgaw.com, and building dashboards to make that data transparent to the teams responsible for delivering them.

I decided to build these dashboards in Google Sheets, using the Supermetrics plugin to automate the import of data from various sources (Google Analytics, Bing, Mailchimp, Facebook, Instagram and our back-office). Though there are many other powerful tools available, we’re already using Sheets extensively, so familiarity was an important factor.

What we’ll cover

• Automating data collection
• Querying the data from summary sheets
• Aggregating, charting and distributing

But first, examples

Here is an example showing campaign KPIs:

Here is another, showing trading KPIs

Not exactly eye candy, but they serve their purpose well. So how do they work? Let me show you my secret weapons – Supermetrics, and the Google Sheets QUERY function.

Automating data collection with Supermetrics

We use Supermetrics to automate the collection of data from various sources. There are many subtleties to this (it probably warrants its own post), but the key is to create one sheet per data source, with date in the first column so we can query it for different time periods. Here is what our Instagram follower count looks like (I wrote about how to automate this in my previous post).

Using this technique, we end up with a lot of sheets within our document that are specific to each thing we’re tracking. Here is a snapshot of some of ours from the campaigns dashboard.

We set Supermetrics up to import all the data each night, which means it’s always kept up to date with each full day’s activities. It would also be possible to copy/paste these values in from other sources, for a semi-manual version.

Querying the data from summary sheets

Once we have the raw data, we need to set up summary sheets. I’ve found that last week (LW), last month (LM) and month-to-date (MTD) fit well with our needs.

Each of these sheets has a hidden section where I hide the formulas I use to establish the date range. Here’s an example which uses various formulas to determine last month’s dates, and the equivalent period in the previous year.

Notice the “query” section? We can pass that into Google Sheets’ QUERY function to to pull in relevant data from our data sheets. Because the data we’re querying always has a date in column A, we’re free to pass the same query into everything on our summary sheets by referencing the cell it lives in.

If you want to see a trend for the time period in question (useful to know if performance is steady or spiky), just change SUM to SPARKLINE et voila!

Distributing

These reports are sent weekly and monthly by email as PDFs. We use Supermetrics to do this, but you could just as easily do it with a Google Apps script, or download them and send them as PDFs. We also bring these up on a screen during relevant team meetings.

Pulling it all together with an example

I’ve put together an example you can view and copy. The data is randomised and the date ranges are not dynamic, but hopefully it’s enough to demonstrate some of the techniques you’ll need to make your own.

View the example dashboard.

In summary

• Collect data for each source in its own sheet, using date as the first column. Automate this if possible.
• Configure a summary sheet for each time period you’re interested in

I hope you found this article useful. Please recommend if you enjoyed it, and add a comment if you have any questions!

Hacker Noon is how hackers start their afternoons. We’re a part of the @AMI family. We are now accepting submissions and happy to discuss advertising & sponsorship opportunities.

If you enjoyed this story, we recommend reading our latest tech stories and trending tech stories. Until next time, don’t take the realities of the world for granted!

Building an automated dashboard with Google Sheets (with example) was originally published in HackerNoon.com on Medium, where people are continuing the conversation by highlighting and responding to this story.

While some tools have sprung up that allow you to track follower count over time, none of them give you direct access to the data so you can’t add it to Google Sheets. Luckily, the problem was easily solved with a simple Google Apps Script, which you can have up and running in 2 minutes. Here’s how!

• Create a new Google Sheet, and create headings “followers”, “following”, “Media count”, “Likes”, “Comments”, “Engagement Ratio” in cells A1 to F1
• Open “Tools > Script Editor”
• Paste this script and change the variables on lines 2 and 4 to the name of your sheet, and the name of the Instagram account you’d like to target.

• Save the script (you’ll be prompted to name the project)
• Open “Run > insertFollowerCount” and when prompted to for authorization choose “Review Permissions” then allow permission on the dialog box

If you switch back to your spreadsheet you should have a new row added with today’s date and the number of followers.

Because Instagram’s API doesn’t give access to historical follower counts, you’ll need to run this every day. Thankfully we can automate this.

Running the script on a schedule

• Open “Edit > Current project’s triggers”
• Click “No triggers set up. Click here to add one now.” then choose “Time-driven” and “Day timer” and choose a time you’d like the script to run.

You’ll now get a new row added to your spreadsheet every night. Once you’ve collected a week or two worth of data, you’ll be able to make it useful.

Making it useful

For our KPI dashboards we track changes over a specific time period and visualise them in charts. Here is an example that shows last week’s metrics using a sparkline to see if changes are caused by a consistent trend, or a one-time spike.

If you were so inclined (we’re not!), you could also track your competitor’s accounts.

Like what you see? Read about how to tie this all together into automated dashboards.

Special thanks

The current version is provided by @JulianDev on Github. Damian Bast provided the previous version of the script based after mins stopped working when Instagram changed their API. Thanks both for your contributions!

1. If there is no calendar event, the meeting doesn’t exist

Uncertainty is inherently unproductive. When people are left wondering whether a meeting is happening or not it’s a waste of everyone’s time, so do yourself a favour and put it in everyone’s calendar.

2. Distribute an agenda

As chair, you must distribute an agenda ahead of time outlining the discussion points, and what you are hoping to achieve. Sometimes it’s useful to list some out-of-bounds topics that may distract from the meeting’s goal. Keep it as brief as it needs to be (often a few bullet points on the calendar invitation will suffice).

3. Avoid “the big reveal”

Share as many of your ideas as possible ahead of time, especially when there are complex or sensitive subjects involved. It becomes easier to focus on what matters by giving attendees time to digest your ideas and discuss them with you ahead of time.

4. If you called the meeting, you chair it

It’s your responsibility to keep the meeting on track based on the agenda you distributed. Beware of conflict-averse attendees who will agree with the meeting’s outcomes publicly but are likely to derail progress by disagreeing privately.

5. Give everyone something to point at

People seem to become much more aligned when everyone can point at the same thing. So use a single screen, projection, whiteboard or sketch to distributing handouts (and save paper too).

6. Capture actions, and ensure they have been accepted

At the end of the meeting, review the actions that have been created. Make sure you all agree and nothing was missed. This helps confirm everyone sees things the same way and accepts what they are accountable for. Where possible, create and assign tasks directly in your project management tool.

7. Take minutes and distribute them

Ideally these minutes are taken on a screen where everyone can see what is being written in real time (we use Google Docs or Basecamp). Keep them as short and informal as possible and put these in a place where they can be discovered.

After countless tests, and experimenting with different ways to marry the UX, prioritisation and testing processes, here is what I find works:

Let’s explore the key parts of this individually.

1. Generate a master sketch per funnel step

Achieving growth for visitors to our site (and just about any e-commerce site) is a matter of improving the performance of any of these steps:

Each step has a primary metric. Simple, right? The means of achieving this growth are not always as obvious, but luckily that’s the fun part.

Focussing on one particular funnel step at a time, we’ll pull together a cross-functional team into a workshop to agree on the most high-impact changes we could make. We’ll usually start by reviewing current usage, user research, customer feedback, Hotjar recordings, best-in-class examples and business objectives. This is followed by a number of rounds of rapid sketching, and we end up with a “master” sketch that represents the best of our ideas.

2. Break the design down into testable components

The master sketch will then be translated into a high-fidelity image or prototype to ensure ideas survive the transition from Sharpies to pixels.

The design will then be broken down into independently testable elements. In the example above, the category header and horizontal navigation represent two testable elements.

We’ll then generate a test plan for each component which covers the hypothesis, metrics to track and any relevant research or data. The primary metric we are tracking is almost always the main test metric for that funnel step – for example product pages measure adds to cart, not completed transactions.

3. Schedule tests to iterate your way towards the master sketch

Armed with a refined sketch and a number of test plans, we schedule the tests to iterate our way towards the master sketch. This slows the rate of change, but ensures that we are creating value and learning as much as possible with every experiment.

Because some of these depend on the success of the previous step, we need to make some assumptions here about what is likely to win, but also be flexible enough that it’s OK to be wrong. Here is an example of some tests in the queue for lister and product pages:

Because each experiment’s goal is getting to the next funnel step, we can run experiments on each funnel step concurrently without them interfering with each other (or at least that’s the theory).

Lister and product page tests generally run for two weeks, and we offset these so a new test launches every week, on a Tuesday. We’ve found that alternating between big and small tests gives us a higher throughput through our small development team. There are sometimes other items in here drawn from the backlog which are unrelated to the master sketch.

4. Build, user tests and QA

All experiments are reviewed internally using pull requests for tech QA and Heroku review apps for non techies. If the test introduces new behaviour, we will also run user tests (using WhatUsersDo) at this stage to check usability.

5. Implement winning tests and compile learnings

We keep a very close eye on which tests are likely to win, and often prepare the implementation started before the test has concluded, in the case of a likely win. Conversion experts rarely talk about the amount of work required to implement a winning test, but in my experience it can take as long to implement the variation as it took to build the test, once you’ve accounted for all the corners you’ve cut to validate your hypothesis.

If the test has won on the primary metric (or is not worse, or is improved but at low statistical confidence), and has not caused any negative side effects, we implement it and launch the next test before we have done deeper analysis. This bias to action gives us the right balance of speed and confidence.

Deep analysis can take anything from 30 minutes to many days, depending on how much there is to be learned. This usually consists of Google Sheets filled with Google Analytics data, but it can also include watching Hotjar sessions, analysing products sold, or any number of other data sources.

And repeat

Users are surprising. Often the side-effects are more interesting than the primary metric we are measuring, which leads to us generating more ideas back into the process.

By feeding the build-measure-learn cycle, we are continually increasing the quality of the hypotheses and generating higher stakes tests with higher degrees of confidence.

One of my favourite innovations to come out of Netflix’s engineering team is Chaos Monkey. It’s a tool which is designed to deliberately wreak havoc on a company’s servers in order to detect failures before they can happen on their own.

It does this by randomly shutting down live servers and ensuring more servers are created in their place. The stakes are high.

Small teams are forced to face a similar scenario when a member of staff takes holidays or sick leave. When an area of responsibility is wholly owned by only one person, things can fall apart when they are not around.

Even in small teams like ours, we need to ensure a certain amount of redundancy to avoid a bus factor of one.

Holidays are a team’s Chaos Monkey was originally published in Product management and technical leadership on Medium, where people are continuing the conversation by highlighting and responding to this story.

We’ve found this process to be super helpful for building things, but figuring out what to build is often more difficult.

“Everything is important”

• Difficult to prioritise business as usual (BAU) tasks against features when we have only one team
• Difficult to prioritise platform development (strategic impact) with features (customer impact)
• No procedure for researching the expected impact of a feature
• Difficult for development team to see what is coming down the pipeline

How we prioritise features now

At King & McGaw, anyone in the company is allowed to add to the backlog. Each week it ends up filled with tasks which range from visionary to chores.

At least once a fortnight, our Head of Development and I prioritise the features using this simple ranking system:

Impact — expected creation or protection of revenue or other strategic objective.

Ease — the cost in man-hours to get the job done.

The result is then multiplied by two and given a final percentage score (well, out of 98). Any task which is expected to have maximum ease (typically BAU tasks, less than 2 hours) can jump the queue and be scheduled for development at the discretion of the team. That’s it! Or is it?

What if we don’t have enough information to rank the feature?

If we can’t estimate the impact

• Data analysis! Extract the information you need from historical sales data, Analytics, server logs or whatever is appropriate
• Clarify objectives with stakeholders, especially if it’s an internal requirement
• Model the expected outcome (“lifting this metric by x % will have this effect”)

In any of these scenarios we will add a book icon (📚) to the Basecamp task until we have enough information to rank the feature.

If we can’t estimate the ease

• There are multiple ways to solve the problem, and perhaps a simple implementation will have less of an impact than a complex one so it also affects impact. Pick one, and rank based on that.
• The problem has not been defined well enough for the developers to estimate.

In any of these scenarios we will have a small workshop to uncover the challenges and propose the most likely outcome.

It’s not perfect

This simple system provides us with structure without being overbearing. It also means we need to be disciplined. If there is something we really believe in, but it doesn’t have a ranking which will allow it to be developed, perhaps we need to look at ways to make it easier, or to have a higher impact.

Originally published on the King & McGaw Developers blog.

The simplest way to prioritise your backlog was originally published in Product management and technical leadership on Medium, where people are continuing the conversation by highlighting and responding to this story.