Social referrals to the site, from 2.0% to 5.8% in seventeen months

Seventeen months of social media work for a company running a subscription service in language learning and education, carried out as a partner to the web production company Nanago (株式会社ななごー). The referral rate from Instagram Stories to the client’s own site went from 2.0% in the first month of the engagement to 5.8% in the last, a rise of about 2.9 times.

What made the difference was moving the KPI off the follower count and onto the share of people who, having seen a post, went on to the site. Followers do not convert into subscriptions on their own; the metric was rebuilt around the stretch that social media actually covers, from first recognition to arrival on the site.

The problem: attention on social media was not reaching the site

The client’s business problem was subscriber growth. Social media is not where a subscription is completed, though. It covers the stretch from first recognition to the moment someone arrives on the site, and that stretch was where the loss was happening.

Rather than a surface figure such as follower count, the working KPI became the share of people who saw a post and then moved to the site. Three things were true at the start:

1. The rate of movement from Stories to the site was 2.0%. Posts were being seen; the visits were not following.
2. The feed save rate was 0.75%. The content was not something people expected to return to.
3. Hashtags were weighted towards terms carrying two million posts or more. Exposure was being buried under the volume.

Approach 1: define the audience by motive, not by activity

Message design starts from why someone would act, not from what they are currently doing. The audience here was defined as women in their twenties to forties with a professional skill, for whom marriage, children, or another change at home had made full-time hours difficult, and who wanted to work from that skill.

What such a reader wants from the service (Gain) and what she is putting up with now (Pain) were each written out under seven headings.

• Gain — flexibility of hours; freedom of location; earning potential; freedom over the shape of a career; independence and growth; variety of work; holding work and home together
• Pain — constrained hours; the difficulty of holding work and home together; terms of employment; unmet ambition; uncertainty about the future; unequal treatment at work; wanting to change field

The engagement turned on shifting the axis of the message from acquiring a skill to working without being tied to a place, and to holding the shape of one’s own career. The effect showed early: posts about where and how a person works took the top positions for both reach and saves, and by the second month the save rate had gone from 0.75% to 1.27%.

Approach 2: carry the same axis through to hashtag selection

A hashtag is chosen on whether the people searching that term match the intended reader, not on how much exposure the term might attract.

The written-out Gain and Pain went straight into hashtag selection, with the post count behind each candidate term measured to set the threshold. Terms carrying around two million posts bury a post, so the set was rebuilt around mid-sized terms in the tens of thousands that still matched the axis of the message.

The share of impressions coming from non-followers was set as the working measure of new reach, and moved from 12.8% to 29.1% in about two months.

Approach 3: review the content mix monthly and act on it the next month

Posts were sorted into six kinds — the case for subscribing, the case for the work itself, subject-matter columns, practical guidance, career accounts, and campaign notices — and the balance between them was planned month by month.

The monthly report analysed reach, save rate, and web taps post by post, not only as monthly totals, and the next month’s balance was set from that. This cycle ran for seventeen months. Some of the decisions it produced:

• reducing the share of posts aimed at readers with no experience or a side income in mind, in favour of practical guidance and career accounts
• reworking career accounts that readers were not reading as their own situation, moving them closer to the texture of daily life
• varying the wording of the call to action in Stories, to lift the web tap rate
• catching months where the mix had flattened out, and widening it with material on particular industries and roles

Each of these followed from the previous month’s measured figures. Setting the balance from data rather than from impression is what produced improvement across seventeen consecutive months.

Approach 4: open LinkedIn from zero and test by job function

Partway through the engagement the client’s LinkedIn page was opened from nothing and then run for nine months.

Because LinkedIn’s readership differs from Instagram’s, the testing was designed around job function. The target function was changed month by month — finance, legal, IT, trade, communications, corporate administration — and click-through rate identified which readers responded. Certain functional segments, together with content addressed to graduates of business postgraduate programmes, showed the strongest response, and the message design followed that finding from then on.

Results

Measure	At the start	Final month	Change
Referral rate from Stories to the site	2.0%	5.8%	about 2.9×
Feed save rate	0.75%	1.94% average across the period	about 2.6× on the period average
Followers	—	—	about 1.2× over 17 months
Reach per post	—	—	held while post volume was cut 25%
LinkedIn	not opened	9 months running	opened from zero

The referral rate is the result that matters. Movement from Stories to the site went from 2.0% in the first month to 5.8% in the last: the same number of impressions now producing about 2.9 times the visits. The first five months averaged 3.4% and the last five averaged 4.9%, a climb sustained across the whole engagement rather than a single good month.

Save rate went from 0.75% in the first month to a period average of 1.94%. The opening four months improved consecutively — 0.75%, 1.27%, 1.78%, 2.05% — after which the figure moved in a range of 1.2% to 3.7% depending on what each month’s posts covered. A save is a reader expecting to come back to something, which says more about the quality of a prospective subscriber than an impression count does.

Reach per post held while the number of posts fell. Monthly feed volume was cut by 25%, and average reach per post in the final month was close to that of the first. The result came from the precision of the content design rather than from volume.

LinkedIn was opened from zero. Impressions grew about 1.9 times over the nine months from launch. Click-through rate excluding advertising moved in a monthly range of 4% to 12%, and testing by job function identified which readers were responding.

What the engagement covered

The work ran from strategy through to analysis, not production support alone.

• Audience definition — analysing what the end reader needs, values, and does, and settling on how to approach her.
• Hypothesis work — structuring the reader’s problem and the solution she is looking for from both the Gain and the Pain side, as hypotheses that can be tested.
• Message design — writing the message against what the reader wants from the service.
• Hashtag and keyword selection — choosing terms along the axis of the message, on measured post counts.
• Production direction — holding content consistent across every channel under one message.
• Copywriting — writing posts a reader wants to react to or keep.
• Reporting — analysing the data post by post once work is under way, and returning the reading of it that sets the next decision.