Roster discoverability
Booked builds and maintains the machine readable layer for an entire roster, on your own domain, alongside the booking platform you already pay for.
For photographer and director reps, HMUA and stylist agencies, production and crew agencies, casting, and talent boards. R4,700 per head, minimum R50,000, then a monthly retainer from R28,200.
Start with roster sizeThe shift
A brand casting twelve faces for a summer campaign used to start with a booker and a set of boards. Increasingly the first pass runs through a search tool or an assistant before anyone picks up a phone. The system reads what it can parse, returns a shortlist, and a person starts from that shortlist rather than from the whole market.
That only matters if your roster is legible to the thing doing the reading. A page that looks correct to a human can be almost empty to a crawler. Names shortened to an initial are not entities. Credits held inside images are not text. A roster that lives only on Instagram and inside a rented platform has almost nothing a machine can turn into an answer.
Nobody is claiming every brand runs its casting through an assistant today. Some do, most do not yet. The direction is one way, and the fix takes about 90 days to index, so the useful question is whether your roster is readable before that matters rather than after.
What breaks at roster level
These are not hypotheticals. Every one was found on a live South African roster site checked directly in August 2026. The organisations are described rather than named, because the point is the pattern and not a callout.
A premium Cape Town board lists every person as a first name and a surname initial. "Princess S" is not an entity. No system can connect that string to a campaign, an editorial or a credit anywhere else on the web, so the person has no findable identity at all.
On the same site, 22 images carry an identical alt attribute. To a crawler the entire roster is one undifferentiated object repeated 22 times. Nothing distinguishes a womenswear lead from a new face, because nothing describes either.
Roster images served from signed storage URLs that expire after 30 minutes. Nothing can be indexed, Google Images holds nothing permanent, and any system that cached the page is holding a dead link. The photography budget is real and the asset is not durable.
A second board names Cartier, Armani, Chanel and Vogue on its homepage. Not one talent name appears in machine readable text. The credits are real, they belong to specific people, and to any system reading the page those people are anonymous.
The roster layer
Three stages. Each one produces something you keep, and each one is useful on its own if you stop after it.
Your full roster scored against the five dimensions below, per person and aggregated to a board score out of 20. Named findings against your actual site rather than general advice. A baseline of the plain language queries a client would really run, recorded with what comes back today.
The layer itself, built on your domain. Full names restored, credits recovered into text, a permanent image pipeline with descriptive filenames and per person alt text, Person schema for every talent, Organization schema for you, and the crawler files underneath it.
A roster changes and a static fix rots. New signings on within five working days, departures removed the same month, monthly monitoring for broken links, expired images and schema drift, and a quarterly re-run of the same query baseline so movement is measured rather than asserted.
Method
The same five dimensions Booked uses on an individual creative, applied per person across a roster and aggregated. Four points each, twenty in total. Scores reflect what exists rather than what could exist, and nothing is rounded up.
Whether systems can find, understand and recommend this person. Killed by social only presence, credits held in image captions instead of text, and no structured data. Weighted first because it is where the first cut now happens.
Whether someone who finds the person immediately understands role, market and level. A profile that could describe anyone at any level scores low here regardless of how good the work is.
Whether the presence is owned or rented. A domain, a working contact path and pages you control score high. Everything living inside a platform you licence scores low, because you do not control what it outputs.
Whether the calibre and range of the work is actually legible. Strong images with no named client or publication in text score in the middle: the work is visible, the context is not, and context is what a search returns.
Whether the person presents consistently everywhere they appear. Different names and different credits across your site, their social and any agency listing fragment the entity, and a fragmented entity gets full weight nowhere.
Aggregated to a board score out of 20, with per person scores underneath so you can see which part of the roster is carrying the loss.
What it costs
Not priced on my hours. If it were, every hour of automation I build would reduce your price, which is the wrong incentive for both of us. The schedule below is published so you can work out your own number before you speak to me.
Your full roster scored per head and aggregated out of 20. Named findings against your live site. A baseline of five plain language queries a client would really run, put through ChatGPT, Claude, Perplexity and Google AI Mode. Three of your people rebuilt as a working demo.
Pay R25,000Secure Yoco checkout. Credited in full if you go ahead. If your roster is under 25 heads, message me instead and we will skip straight to the engagement.
The layer itself. Full roster mirror on your domain, Person schema per head, ImageObject markup on flagship work, a permanent image pipeline, full names restored, credits recovered into text, category and market signals per head, your Organization schema, the crawler files, Search Console, and a 90 day re-baseline.
New talent on within five working days, departures removed the same month, monthly crawl and drift monitoring, a quarterly re-run of the same query baseline, and a one page visibility report you can actually read.
Every figure below re-derives from that one rate. Nothing here is a package, and nothing is rounded to look friendlier.
| Roster | Engagement | Retainer, monthly | Year one |
|---|---|---|---|
| 12 heads | R56,400 | R28,200 | R394,800 |
| 20 heads | R94,000 | R28,200 | R432,400 |
| 30 heads | R141,000 | R28,200 | R479,400 |
| 45 heads | R211,500 | R30,575 | R578,400 |
| 60 heads | R267,900 | R37,700 | R720,300 |
| 100 heads | R399,500 | R56,700 | R1,079,900 |
Roughly one to two extra booking days per person per year covers it. That is the number worth checking against your own commission, rather than a break even count I could make say anything.
Talent onboarding for anyone added mid term, R550 per head. A 45 head board at 40 percent churn is eighteen builds a year, so this is not padding.
Compliance Layer, R35,000 plus R2,800 a month folded into the retainer. POPIA across the roster, model release audit, likeness and AI usage clauses, operator agreement.
Diagnostic upfront, credited in full. Engagement 50 percent on signature to hold the slot, 50 percent on go live. Retainer billed on the first, head count trued up the same day, three month minimum then 30 days notice.
Net 14, quoted in ZAR. Annual prepay is twelve months for ten. Retainers carry a CPI plus 2 percent escalator from day one. The rate does not move, scope does.
Fit
Procurement questions
Only information you already publish or hold a release for is used. Booked acts as an operator under POPIA, works to a written operator agreement, and does not process special personal information. Talent data stays on your domain and your hosting. Nothing is sold, syndicated or used to train anything.
You do, on handover, in full. The pages, the structured data, the image pipeline and the documentation sit on your domain and your repository. There is no licence to revoke and no platform to be locked into. If you stop working with Booked the layer keeps running.
A current roster export, the highest resolution images you hold, whatever credits exist in any form including spreadsheets and inboxes, and one named person who can confirm facts about talent. Access to the domain and DNS at build stage. No design sign off cycle is required, because this is not a rebrand.
The Scan returns inside two weeks. The Structure build runs roughly four to six weeks depending on roster size and how much credit recovery is needed. Indexing is slower than building: expect meaningful movement in AI and search results from about 90 days, which is when the baseline is re-run.
Departures are removed in the same month and the structured data is updated so search and AI systems stop associating that person with you. New signings are onboarded within five working days. Because maintenance is priced against roster size, the cost moves with the roster instead of being fixed for a year.
No. The layer sits alongside whatever you already run and does not touch it. Those platforms were built for a booker browsing on a laptop and they do that job well. They were never built to output a page a machine can parse, and you do not control the markup they produce.
The diagnostic is R25,000 upfront and is credited in full against the engagement. The engagement is 50 percent on signature to hold the slot and 50 percent on go live. The retainer is billed on the first of the month with head count trued up the same day, three month minimum, then 30 days notice. Net 14, quoted in ZAR, annual prepay is twelve for ten.
Start here
No number can be quoted before roster size, category and timeline are known, so those are the first three questions rather than the last. It takes about a minute and it is the same information the first call would open with.
Rather talk first? WhatsApp or luca@bookednow.co.za.