← Back to Blog

A practical Boomp guide

101-Post Social Media Experiment: What It Did—and Didn't—Prove

Publishing 101 times can prove that a system produced output.

It does not automatically prove that:

  • an algorithm rewarded consistency;
  • one format was a universal winner;
  • automated writing sounded human;
  • the audience trusted the business;
  • the posts changed buying behavior; or
  • the product created customers.

The original version of this article made several of those leaps. This August 2026 revision keeps the useful operational observations, exposes the missing evidence, and replaces confident conclusions with a test another person could actually evaluate.

That is a better use of an experiment than pretending the first story was research.

Why this page changed

In March 2026, I published a first-person account titled “10 Things I Learned From Letting Boomp Post 101 Times.” It described an early Boomp workflow operating on my own social presence.

The article had a useful instinct: use the product yourself before asking somebody else to trust it.

It had an evidence problem.

One account, one operator, mixed material, changing conditions, and platform dashboard totals can produce observations. They do not become causal conclusions merely because the post count is large.

The repository also does not retain the raw export or post-level evidence needed to audit the original claims now.

So this page is no longer presented as proof that “the platform works.” It is a transparent record of what the retained source can—and cannot—support.

The source record that remains

This is the complete evidence currently available in the repository:

Historical note in the original articleWhat is retained nowLimitation
101 posts were publishedThe original article, slug, title history, and a historical promotional row repeat the countNo post log or platform export is retained
9,253 total impressionsThe number appears in the original articleNo platform, date, post-level, or metric-definition breakdown
181 total engagementsThe number appears in the original articleThe listed components do not reconcile to the total
140 likes, 29 comments, 2 sharesThe components appear in the original articleThey sum to 171, leaving a ten-engagement difference
One post had a 25% engagement rateThe percentage appears without the underlying numerator and denominatorThe rate cannot be reproduced
One post had 81 “impressions” or “reach”The original wording changed units within the storyIt cannot be reconciled with the total without platform-level definitions
Carousels, misconception hooks, and questions performed bestSeveral selected examples were describedNo complete comparison set, sample sizes, or denominators remain

This is not a dataset another analyst can reproduce. It is an unaudited historical note with a useful warning attached.

The warning is simple:

Keep the evidence before writing the victory story.

What the experiment can support

If we accept the contemporaneous account as an honest record of the founder's experience, it supports several limited observations.

1. A system can reduce the blank-page decision

The early workflow allowed a long run of material to move without writing every caption from scratch on the day it published.

That matters operationally. It does not tell us whether every item was useful, accurate, distinctive, or commercially valuable.

2. Published posts create response work

The original account recorded comments and explicitly noted that a person still needed to respond.

Automation can prepare or publish material. It does not own the customer relationship caused by that material.

3. More attempts can reduce emotional pressure on one post

The founder reported feeling less attached to whether an individual item “won.” That is a valid first-person observation about the experience of producing a larger sample.

It is not evidence that every expert should publish at high frequency. More volume can also create more review, monitoring, correction, and noise.

4. Attention metrics and business outcomes are different

The retained article recorded impressions and engagement. It did not preserve attributed website movement, qualified inquiries, purchases, retained customers, or a control period.

The experiment can describe attention. It cannot establish commercial success.

5. The test exposed what the next product needed to measure

The strongest result was not a carousel or an engagement rate. It was discovering that output alone was an incomplete definition of “working.”

That lesson now shapes the current Boomp funnel: useful source material, owner judgment, finished work, verified publishing, customer movement, and paid continuation are different stages.

What the experiment cannot support

The original version turned several observations into rules. Here is the corrected interpretation.

Original conclusionWhy the evidence is insufficientDefensible statement
“The algorithm rewards consistency more than perfection”No baseline, control, quality definition, or platform-specific distribution evidencePublishing creates more opportunities for distribution; this test does not identify an algorithmic reward
“Carousels are the cheat code”One selected item and no complete format comparisonA carousel was remembered as a strong item; no general format advantage was established
Challenging misconceptions drove commentsTopic, format, audience, platform, and timing were not separatedSome selected examples used tension; causation is unknown
Direct questions increased comments significantlyNo before-and-after counts or denominators remainQuestions can invite responses, but this test did not quantify the effect
Nobody cared that software wrote itNo public AI accusation was reportedSilence cannot reveal what every viewer noticed, thought, or did
Automation instantly saved hoursNo task-time baseline or review-time record remainsThe founder felt relief; the amount of work moved was not measured
Frequency “flushes out” bad ideasNo explicit quality or learning rule was documentedA larger sample can create more observations when the evidence is retained
Publishing proved the business was openNo visitor study or customer-path evidence remainsRecent accurate material can be one inspection signal; it does not prove every visitor's conclusion
“The platform works”No success criterion or customer outcome was definedThe workflow published; broader product value remained unproven

The complete algorithm and inactivity boundary belongs in what happens when a business does not post regularly. The complete failure diagnosis belongs in why social media is not working.

Ten lessons worth keeping

The original list was too certain. These are the ten lessons the available evidence can responsibly support.

1. Output is an input

A count of published posts tells us that material went out. It does not tell us whether the right person saw it, moved, bought, or stayed.

2. Preserve the raw source before interpreting it

Export the post-level data, record the date and account, preserve metric definitions, and keep the file with the analysis. A screenshot total is not enough for a lasting research claim.

3. Make the arithmetic reconcile

If a total says 181 and the listed parts sum to 171, stop. Resolve the difference before publishing a conclusion.

4. Do not silently switch metrics

Impressions, reach, views, profile visits, and engagement are not interchangeable. Platform definitions can differ too. Label the source and unit for every number.

5. Keep numerators and denominators

A 25% rate cannot be checked without both values and the definition of an engagement. Preserve counts alongside percentages—especially in a small sample.

6. Separate platform, format, topic, and timing

If a carousel on one platform covers a stronger topic than a static post elsewhere, format alone did not create the difference. Tag every variable you may later compare.

7. Record the human work

Track source gathering, preparation, corrections, approval, delivery failures, comments, messages, and measurement. Automation that moves typing but creates substantial oversight should not be called zero work.

8. Treat absence of criticism as unknown

Nobody saying “this sounds like AI” does not prove that the material sounded faithful. Inspect the source, the business-specific substance, owner corrections, and what suitable customers did next.

9. Join attention to the customer path

Impressions and engagement can diagnose attention. Add attributed website visits, preview starts, qualified inquiries, purchases, and retained customers before making a business-value claim.

10. Define success before the next test

Choose the primary question, time window, population, metrics, and failure conditions first. Otherwise every dashboard contains a number that can be turned into a success story afterward.

What this means for a reluctant expert

The expert's problem was never simply “not enough posts.”

A system can publish 101 items and still fail to surface the question, distinction, safeguard, or explanation that makes the work valuable.

Good experts are often too close to their own evidence. What feels routine from inside the frame may be exactly what an outsider needs to understand. Solving that problem requires:

  1. a reliable source or a good question;
  2. outside attention that notices the useful part;
  3. preparation that turns one idea into something concrete;
  4. expert review for truth, nuance, privacy, and fit; and
  5. measurement that follows the right person beyond the post.

Volume cannot replace those jobs.

The full recognition argument is in how to promote good work without feeling sleazy. The complete website mechanism is in how to turn a website into social media posts.

A better next experiment

Before publishing the first item:

  1. name the account, platform, start date, end date, and intended customer job;
  2. export a baseline period;
  3. assign each item a stable ID;
  4. retain the source page, question, or proof behind it;
  5. label format, topic, destination, and approval changes;
  6. record preparation, review, correction, and response time;
  7. verify delivery rather than assuming scheduled means published;
  8. preserve post-level impressions or reach with the platform's definition;
  9. use attributed links for movement to an owned destination;
  10. count qualified inquiries, purchases, and retained customers; and
  11. publish the raw counts, missing data, and limitations with the interpretation.

Without a control or stronger design, describe associations and operational observations—not algorithmic causation.

How the current Boomp first look fits

The current first experience does not ask someone to trust a 101-post success claim.

It asks for one public website URL and returns up to six talking points Boomp can support from that site before email. The owner judges whether Boomp noticed something true and useful.

Email saves that first look and continues into post ideas. Finished work belongs to the current paid path, with new customers beginning in manual review.

The first look does not prove Boomp knows the owner's complete voice. It does not interview undocumented expertise, supply private evidence, verify changing facts, obtain customer permission, replace professional review, or guarantee attention or customers.

It gives the visitor a smaller piece of evidence to judge before making a larger decision.

How this test should produce customer evidence

Measure the path in raw counts:

  1. qualified search entrances to this evidence page;
  2. movement to the recognition or website mechanism;
  3. attributed clicks into the personalized first look;
  4. preview attempts and grounded first looks produced;
  5. email captures and continued post ideas;
  6. checkout progression and paid first months; and
  7. customers who reach the review experience.

At current volume, one customer can move a percentage dramatically. Keep the numerator and denominator visible, retain zero outcomes, and do not convert a small fluctuation into another victory story.

See what Boomp notices on your website.

Related: What happens if you stop posting regularly? · Why social media is not working · Why I built Boomp · Turn your website into social posts

Frequently asked questions

What did Boomp's 101-post experiment prove?

It supports a narrow operational claim: the early system was used to produce a large run of published material and created work that still needed human monitoring and response. The retained source does not prove that consistency caused greater reach, one format outperformed another, people could not recognize AI-assisted writing, or the posts produced customers.

What were the results of the 101-post experiment?

The original March 2026 article recorded 101 posts, 9,253 impressions, and 181 engagements, then listed 140 likes, 29 comments, and 2 shares. Those components total 171, not 181. No raw export or post-level dataset is retained in the repository, so the ten-engagement difference and the platform-level figures cannot be reconciled. Treat the numbers as an unaudited historical note, not a benchmark.

Does posting consistently make social media algorithms reward a business?

The experiment cannot establish that. Publishing creates new opportunities for distribution, but proving an algorithmic consistency reward would require platform-specific data, a defined baseline, comparable content, a time window, and a design that separates posting frequency from format, topic, audience, account history, and platform changes.

Did the experiment prove that carousels get the most engagement?

No. The original story described one carousel as the highest-performing item but retained no complete format counts, post-level denominators, platform breakdown, or comparison table. One highlighted item cannot establish that carousels are generally superior.

Did the experiment prove that people cannot tell when AI helped write a post?

No. The original article said nobody publicly called the content AI-generated. Silence is not evidence that nobody noticed, cared, disengaged, or formed an opinion. Voice quality requires source inspection, human review, and outcome evidence—not the absence of one kind of comment.

How should a business measure a social media publishing test?

Retain a post-level export with date, platform, format, source, approval changes, delivery status, impressions or reach, meaningful responses, attributed visits, qualified inquiries, and customer outcomes. Define each metric before the test, preserve raw numerators and denominators, and record the human work required.

What does Boomp's current first look create?

The current first look accepts a public website and shows up to six website-grounded talking points before email. They are not finished posts or proof that Boomp already knows the owner's voice. Email saves the first look and continues into post ideas; finished work belongs to the paid path, with new customers starting in manual review.

See what your website gives us to work with

Enter a public website to get up to six website-grounded talking points. Review them before email; email saves the first look and continues into post ideas. Nothing goes live without your approval.

Get the website-grounded first look

Save this to Pinterest

101-Post Social Media Experiment: What It Did—and Didn't—Prove
KC

Written by Kathleen Celmins

Founder of Boomp. Helping local businesses stay visible on social media without doing the work themselves.