We are not a do-it-for-you agency.We are a do-it-with-you practice.
We still do the work. The difference is that your team watches it happen, questions it, and can run it after we leave.
Wenstein is boutique by design: thirty years across ecommerce, search, AI, analytics and web technology, delivered by a small senior team rather than an account structure.
Every relationship starts the same way
One bounded piece of work. Not a retainer, not a platform, not a package. Something finite with a clear question behind it. Here is what that turned into once.
Year one, consultancy. A multi-brand distributor came to us with more than twenty websites, several content platforms, several ecommerce platforms and several agencies. Product data lived in spreadsheets. Systems integration had no owner. They hired us for one thing: audit it, and say what to do about it.
Thirty days. We delivered the ecosystem audit, a redesigned architecture, and a ballpark on what consolidation, reorganization and investment would cost and would save.
Then, consulting. They asked for help executing it. That meant selecting the platform and the agencies, managing the vendors, and hiring their own digital project manager. We recruited the person whose job was to need us less.
Before
After
Then they asked us to do the same for marketing. Search, paid, email, affiliate.
Nearly eight years on, advisory. We consult on ecommerce, digital marketing and AI, and we advise their leadership directly. The consulting did not stop when the advisory started. The stages add up, they do not hand over.
Those numbers belong to the client, not to us. We are not claiming we produced that revenue, and anyone who claims otherwise about a business this size is selling you something. They are here to answer one question: what size of operation this practice is used to working inside. The measured results further down are the part we will stand behind.
What the three stages actually mean
Consultancy
External, diagnostic, finite. You buy an answer, not a relationship. It ends with a plan you could hand to someone else, and sometimes that is exactly what happens.
Consulting
We help you execute what the plan said, alongside your team rather than instead of it. This is where knowledge transfers, because it transfers by working next to people, not by writing documentation at them.
Advisory
A standing relationship with the people making the decisions. Fewer deliverables, more judgment. It exists because the earlier work earned it.
You can stop at any stage, and stopping is a normal outcome rather than a failed one.
What we work on
Three kinds of work, one thing that runs underneath all of them, and one way to hand the whole thing over. Every one of them is a place a client has actually started.
Foundations
The things everything else depends on. Get these wrong and no amount of visibility work compensates.
eCommerce Platform Consulting Web Accessibility Cold Email Infrastructure
Being found, and being quoted
Search and AI visibility. The SEO page covers technical, on-page, audits, catalogs and the platform-specific work.
Demand you can trace
Buying attention rather than earning it, with every dollar attributable to something.
Underneath all three
Measurement
None of the three above means anything without a baseline frozen before the work starts and an instrument that can read the result.
Or have us own all of it
Fractional
When you would rather one senior person held the whole thing from inside your team, with a defined term and a handover.
How we work, and how to hold us to it
Three rules. They are the reason to choose us, and they are easy to check.
One exception worth naming. Everything described here is a do-it-with-you arrangement, where your team keeps ownership. In a fractional digital leadership role we take the seat and the accountability instead, under a defined term and a handover mandate.
Every claim carries its number, and every number carries what it is out of. A ratio without a denominator is decoration. “8.6x return” on 1.67 conversions is not a result, and we will say so.
Every recommendation states what would prove it wrong. We record that before the work starts, not after the report. If we cannot name it, it is an opinion and we label it as an opinion.
We report what did not work. On our last on-page program, 12 of 20 pages improved and 8 declined. You get both numbers.
What measured work looks like here
Three clients, three different questions. We have anonymized all of them.
A catalog nobody could see
A distributor came to us with 754 of more than six thousand pages indexed by Google. Twelve and a half percent. The rest of the catalog was not ranking badly, it was ineligible. Traffic was not flat either, it was falling: impressions down 42.9% against the same months a year earlier.
Pages Google had indexed when we arrived
Measurement was in worse shape than the content. The tag manager had never been installed on the site at all, ecommerce tracking was switched off on a site that sells, and Search Console was linked to nothing.
That property now runs over eight million search impressions a year. We are deliberately not turning that into a multiple. The baseline we hold from 2022 is a top-thousand-query extract rather than a whole-property total, so dividing the two numbers would flatter us. The direction is not in question. The exact factor is not ours to claim.
The same client, and the part we could not have arranged
The control group we did not ask for
Google suspended this client’s Merchant Center account for twenty-six months. A calculation fault in their ecommerce platform triggered it. The Shopping channel was gone, and it was gone during the period Google was feeding that same product data into organic results as well as paid.
The strongest month in the entire record is month twenty-five of that suspension.
We report this because of what it rules out. The usual objection to any result like ours is that the growth came from Shopping, not from the work. On this account that explanation is not available. Google removed it, for two years, and the search numbers moved anyway. We did not arrange this and we would never have been allowed to.
What it does not prove: that the suspension helped, or what the same work would have produced with the channel open. That was never observable, and we are not going to pretend otherwise.
The page-level work, including the parts that failed
117 pages across six batches, June to September 2026, at a different client. That one runs no paid search and no ads budget, so every result is organic and uncontaminated by spend.
Click-through rate change, nine weeks after applying
One batch, twenty pages
Catalog scale is a supply problem, not a writing problem. On a distributor running over nine hundred published products, we did not quote per page. We built the intake: manufacturers receive a templated sheet that converts what they already hold, specifications, materials, dimensions, certifications, into content formatted for direct upload. The bottleneck moves from writing to collection, and it keeps working after the pages we wrote are done.
We compare against pages we deliberately did not touch, not against a site average. Then we remove the batch’s biggest page and re-run the numbers. A headline that does not survive that test is not a result. That check has killed one of our own headline numbers.
Whether answer engines actually quote you
Back to the first client. Its pages carry a structured question-and-answer estate built years before anyone called this AEO, for a search feature Google has since largely withdrawn. The markup kept its value anyway, because answer engines read it. It is now the second most used source in its category, ahead of a national retailer you have heard of, and behind only a forum.
And the number that matters more: the same client is named in 5.8% of the answers it feeds. Cited constantly, credited almost never. That gap is an entity and attribution problem, and it is the one we report beside the citation count rather than instead of it.
Instrument and window: Searchable, 30 days to September 24, 2026, across ChatGPT, Perplexity and Google AI Overviews. Rank is by share of answers drawing on the source, not raw citation count.
Distinct pages cited in AI answers
What the instruments cannot do. AI citation exports refresh only partially, so a period-on-period change is a floor rather than a measurement. Composite visibility scores are unstable enough that we will not quote them alone. Google withholds low-volume search terms, so some value is always unattributable.
What you will see, and when
Directional. Enough signal to steer, not enough to conclude. We label it that way in the report.
Decision. Four clean weeks, which removes day-of-week bias. This is when we tell you to continue, change or stop.
AI visibility work. Those effects are slower, and a 28-day read alone would mislead you.
Where the method came from
Before advising on this work, our founder ran it. Ten years at Vitec Group, now Videndum, building a global ecommerce infrastructure from nothing: one ERP running nine instances, one product information system, and 130 direct-to-consumer storefronts. And the team that ran it, every one of them recruited, hired and trained by him.
That last part is the whole method. You cannot run twenty-five people across nine countries by doing the work yourself. You make them capable or nothing ships. It turned out to work better than doing it yourself ever had, and it has been how we work ever since.
Which is why we are not a service provider. We train on the job, and the capability stays with your team when we are done. The results our clients get are what expands the work. It is the only sales mechanism we have ever needed.
Who we are not for
We turn work down. It is faster to say so here than in a first call.
- You want a supplier. We work alongside your team and we expect access to it. If the plan is to hand it over and hear from us quarterly, we are the wrong choice.
- Nothing can be measured. No tracked conversion, no baseline we can freeze before starting, no instrument for AI visibility. We do not sell unmeasurable work.
- The scope is too small to justify the method. Fewer than about sixty pages worth optimizing and the program exhausts itself in two batches.
- You want social accounts run. We are the wrong firm.
Questions we get asked
Q. Do we have to replace our current agency?
No. We frequently work alongside an agency rather than instead of one. Our work is usually the part they do not do: architecture, measurement, AI visibility, or deciding what should be built before anyone builds it.
Q. Do you work with our team, or instead of them?
With them. We do the work, and your people watch it happen and question it. If nobody on your side has time to be involved at all, we are the wrong choice and we will say so early.
Q. What does a first engagement usually look like?
One bounded piece of work with a clear question behind it, finite and scoped in weeks. It ends with something you could act on without us, and sometimes that is exactly what happens.
Q. Do we need to change platforms to work with you?
No, and we will usually argue against it. Replatforming is expensive, it puts existing traffic at risk, and most problems we are called in for turn out not to be platform problems.
Q. How long before we see anything?
We report at 14 days as directional and 28 days as a decision point, and we label them that way. AI visibility work is slower, so we set a 60-day read alongside the 28-day one.
Q. What happens if it does not work?
We tell you. We record before starting what would prove the work failed, regressions go into the report with everything else, and on our last on-page program 12 of 20 pages improved while 8 declined.
Q. Do you sign a long contract?
No. Our on-page program typically runs three or four productive batches rather than indefinitely, because the pages carrying the value get done first by design. When the pool is exhausted we say so.
Q. How do you charge?
We price the engagement, not the hours. Most relationships start with one bounded piece of work at a fixed fee, and continue on a monthly retainer only if the first piece earned it. We will not quote before we understand what the problem is costing you now and what solving it would be worth, because a price without that context is just a number. If the only difference between two providers is price, the work has been treated as a commodity, and that is not what we sell.
Q. What size company do you work with?
Size is not our filter. We have worked with a one-person business and with distributors running large catalogs. What matters is whether the work can be measured, whether there is enough of it to justify the method, and whether someone on your side has time to be involved. Those three questions decide fit, not headcount or revenue.
Start with a conversation
A first call to work out where you actually stand, what is worth doing first, and what it would take to measure it. No obligation and no pitch. If we are the wrong fit we will say so on that call, which is cheaper for both of us than finding out in month three.
Pick a time that suits you below. Thirty minutes, straight into the calendar.